Techy Surgeon Podcast

This is a free preview of a paid episode. To hear more, visit techysurgeon.substack.comThank you Doug Fullington, MD, Alex Rivero, HealthMind Insights, Eric Burgh, MD, and many others for tuning into my live video! Join me for my next live video in the app.I can publish a paper in an upper-tier orthopedic journal and, if I’m being candid, a few hundred to maybe a few thousand people will read it. Some of those people will cite it in their own papers. A small number will change anything in their practice because of it.Then I share an insight about CMS payment model mechanics on Substack, and a health system CFO I’ve never met emails me because she’s been trying to articulate that exact problem to her board. A policy researcher in Geneva follows the thread. Three orthopedic surgeons I’ve never spoken to start a conversation about what I wrote that turns into an ongoing dialogue that has genuinely shaped how I think.That asymmetry in exposure is what motivated me to start writing in public. Not personal branding, not monetization, not the promise of newsletter revenue (though those have their own logic). The primary driver was reach and dialogue. The feeling that the ideas worth communicating in healthcare were reaching a fraction of the people who needed them, mostly because we’d built a science communication infrastructure that made academic publishing the gold standard and everything else secondary.AI changes the equation a bit. Not by doing the thinking. The thinking is still yours, and it’s still the most important part. But by handling enough of the operational needs that someone running a clinical practice, a research agenda, and a startup can also maintain a consistent presence in public. That’s what I want to share here: how I’ve structured that system, in enough detail that you can build your own version of it if it seems useful.What works for me may not work for you. I can’t promise to optimize your content calendar. But I wanted to describe infrastructure that has made the practice of writing in public sustainable for me, and that has produced a lot of unexpected good in the process.Why This Isn’t Primarily About Content CreationA minor but important framing note before the setup details: the value I’ve gotten from Techy Surgeon isn’t primarily the newsletter metrics.It’s the policymakers who’ve reached out. The research collaborators who found me because of a piece on care coordination. The clinicians who wrote to say that an article crystallized something they’d been trying to explain to administrators for years. The people building interesting companies in healthcare who wanted to connect because we were apparently thinking about similar problems from different angles.None of those connections would have happened if I hadn’t been willing to put my thoughts into a form that others could interact with. And it wasn’t a white paper or journal article that did that. It was something closer to a public conversation, where the format invites response and the distribution reaches people outside the academic bubble.I think medicine, and clinical research more broadly, is underinvested in this kind of communication. Not because clinicians don’t have things worth saying (clearly they do), but because the infrastructure for saying them publicly has been either unavailable, considered taboo, or too costly in time. AI is changing that. The skill of generating multimodal media whether written, visual, or video, is becoming something a motivated clinician can build and maintain without a full production team. We probably need more of us doing this.The Flywheel, BrieflyBefore the setup details, the concept: a content flywheel is a system where each piece of output feeds the next, and where the marginal cost of producing content decreases over time rather than remaining constant.The alternative is what most clinician-writers default to: the brute-force model, where every piece starts from scratch, involves a Sunday afternoon staring at a blank editor, and depends entirely on having energy left over from the clinical and research work. That’s a fragile system. It produces good work occasionally and nothing the rest of the time.The flywheel I’ve built runs on five loops: Ideate (keep a backlog so you never start from nothing), Research (get sources before you start writing, not after), Write (use AI constrained by your voice, not AI in its default state), Distribute (publish once, distribute many times across platforms), and Repurpose (one strong piece seeds two weeks of downstream content).The setup below is how I’ve operationalized each of those loops.

This is a free preview of a paid episode. To hear more, visit techysurgeon.substack.com Thank you to everyone who tuned into my live video! Join me for my next live video in the app.I went live at 6:45 the other morning to open three tabs, ChatGPT for Clinicians, Doximity GPT, and OpenEvidence, and ask them the same questions. A few dozen clinicians and subscribers joined at that hour on a Sunday, which I did not expect, and I’m grateful for.The headline finding isn’t who won. It’s that it seems soon you won’t be able to tell the three tools apart from the navigation bar. Each one now has an ambient scribe (or form of one). Each one tracks CME. Each one has a “skills” or “dot flows” tab that, today, mostly amounts to baked prompts dressed up as workflows. OpenEvidence has a feature literally called the dialer — Doximity has had a dialer for a decade. The product surface is converging fast.A quick disclaimer before we go further: opinions here are mine alone. I have no financial relationship with any of these companies. I selected these three because they appear to be getting the most traction in the marketplace — not because they’re the only ones worth your time. Up-to-Date Expert AI, Glass Health, Abridge’s embedded answering, and others all deserve their own look.What each tool is best at right nowChatGPT for Clinicians is the new entrant. Verification is rigorous — NPI, photo of a driver’s license, a ClearID face match — which I read as a deliberate credibility signal. Underneath, the experience is polished but the clinical answers were the weakest of the three on the queries I ran. There is a skills surface that hints at where this is going, but most of the entries today function as prompts rather than true agentic workflows. I did not see a Business Associate Agreement presented during signup, and I have not yet found a satisfying answer on PHI handling. Doximity GPT quietly has the best one-off clinical answers right now. Not by a wide margin, the others are good, but on a hip arthroplasty question and a DVT prophylaxis report, Doximity surfaced the PREVENT CLOT trial and the CRISTAL trial at the top of the response, where a domain expert would put them. For a clinician, citation prioritization is trust. Doximity also brings a distribution moat the others can’t replicate quickly — the dialer, fax, telehealth, the news network, and Peer Check (where physician experts grade the answers) — and a redesigned interface that’s the cleanest of the three.OpenEvidence has the lowest friction and the fastest latency. They are clearly throwing serious compute at the answer surface. The differentiator most clinicians never find is Deep Consult. Turn it on, answer two or three follow-up questions, and you get a research-grade brief with embedded figures from JAMA and NEJM, made possible by the licensing partnerships OpenEvidence has signed with NEJM Group and other major publishers. When I asked Deep Consult to brief me on secondary fracture prevention for a quality improvement committee, the output was something I could have walked into a department meeting with that morning.Distribution beats product when the products convergeAll three are free. All three answer questions credibly (ChatGPT least so). All three are racing to bolt on the same surrounding capabilities. Doximity wins on installed clinician base. OpenEvidence wins on speed, trajectory, raw capability and on Deep Consult. ChatGPT for Clinicians wins, today, on almost nothing — but the verification gate suggests they intend to be taken seriously, and they hold a foundational model and patient facing asset the others don’t.The chat interface is no longer the moat. The moat is whoever first connects grounded clinical evidence to native multimodal output, real workflow extensibility, and physician-earned trust without forcing the clinician to play copy-paste between four tabs to get there.This is the worst these tools will ever be. That should change how we evaluate them: I’m less interested in the question “does it work today?” , and more gravitating to “how do we shape what it becomes?” Date several. Marry none. Use the tool that fits the question in front of you. Send the teams behind them living, breathing feedback.On the HorizonNone of these tools is built for patients. The updated guidelines on incidental hepatic steatosis answer ChatGPT gave me this morning was reasonable, and I am a bone surgeon. I read it the way a layperson would, and I would not stake decisions on it without help. The literacy gap between clinical outputs and patient comprehension is not a UX problem. It is a safety problem. Tearing down the gate before we have built tools that respect that gap is how we get harm.The administrative arms race — prior authorization letters, denial appeals, faster note-writing — is a symptom, not a cure. We went deep and fast on the workflows where the money lives, which are the workflows our payer infrastructure forces clinicians to spend their evenings on. That work is valuable, and it is not patient-facing. The places where AI could actually move outcomes — secondary fracture prevention, fall prevention, post-op care navigation, osteoporosis treatment rates that sit around 20% after a fragility fracture when the evidence base for treatment is overwhelming — are still under-resourced.Trust and co-design with clinicians is the unlock. OpenAI has not earned it yet for clinical use. Doximity and OpenEvidence have, in different ways, by being physician-forward from day one. That posture is not optional going forward — it is the moat.The path from clinical intelligence in your pocket to democratized, evidence-based care that actually moves quality and outcomes runs straight through clinicians willing to show up and iterate.That is the dream. We are not there. We can get there.Christian Péan, MD, MHS, is an orthopedic trauma surgeon in Durham, North Carolina. He is core faculty at the Duke-Margolis Institute for Health Policy and CEO and co-founder of RevelAi Health, an AI care management platform for value-based care. Opinions are his own.🔒 For paid subscribers — the full demo and the operator’s notesThe complete screen-share from the morning’s livestream. Side-by-side queries across all three tools, the Deep Consult walkthrough, the prior authorization generation, the acetabular fracture surgical-plan comparison, the live multimodal handoff into a branded HTML committee deck, the connector detour inside Claude, and the clinical trials map I discovered on stage.

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Thanks to everyone who joined this morning to tune in. I have below linked to my other techy flows, and please stay tuned for the follow-up article with a walkthrough of shorter clips of how you can use AI to make images and PowerPoints. Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.https://techysurgeon.substack.com/t/techy-flows This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Thank you Edward M. DelSole, MD, Danny Goldenberg, Audley Mackel III, Darren Michael, and many others for tuning into my live video with Hadi Javeed! Join me for my next live video in the app.The Skill Is the New WorkflowClinical AI won’t scale through better models. It will scale through better instructions.Interested in deploying clinical AI for your practice, value-based care organization, or health system?RevelAi Health partners with clinics and health systems to build AI workflows for CMS models (TEAM, ASM, ACCESS), care coordination, and clinical operations. We bring the software, the clinical expertise, and the AI-fluent staff to deliver outcomes, not just tools.Schedule a demo or reach us directly at [email protected] billboard on Market Street in San Francisco advertises “skills,” the hot new paradigm in AI development. Walk three blocks in any direction and you’ll find someone who can explain, in considerable detail, what a skill is, why it matters, and which framework implements it best. Fly to any hospital in the country and ask the same question. You’ll get a blank stare.This gap between what AI can do and what healthcare is doing with it has become the defining tension of clinical AI’s current era. The models are smart. Sixty-six percent of physicians now report using health AI tools, a 78% increase from 2023. Billions have been invested. Yet nearly four years after the ChatGPT moment, most large health systems still haven’t deployed a single patient-facing AI application beyond ambient documentation.The question is why. And the answer, increasingly, points not to the intelligence of the models but to the architecture around them: the instructions, the context, the workflows that translate raw capability into clinical utility.If you like deep dives on clinical AI and health policy, consider becoming a free or paid subscriber to Techy Surgeon!The Context Problem Nobody Wants to AdmitThis past weekend, my co-founder Hadi and I sat down for what we’ve been calling Founders Coffee, a live conversation on Substack about what we’re seeing in clinical AI, what’s working, what isn’t, and what comes next. Hadi brings a particular vantage point: before we started RevelAi Health together, he was one of the earliest applied AI engineers at Capital One, building voice AI for banking back in 2016, when the technology was, as he puts it, “not that cool and less practical.”The lessons from that era are uncomfortably relevant now.At Capital One, text-based chatbots found product-market fit. Voice did not. It got dates of birth wrong. It misread credit card numbers. And the core insight that emerged, one that the current wave of healthcare AI companies would do well to internalize, was deceptively simple: people hate chatbots. Not because the technology is bad, but because it fails to deliver unique value. Empathy for the sake of empathy, as Hadi noted, does not work. People engage with AI when it solves their problem. They disengage quickly, permanently, when it doesn’t.“People only would chat to a chatbot if it solves their needs,” Hadi said. “As long as the chatbot is not providing unique value, it does not work.”This observation lands differently in 2026 than it would have in 2016. Today, the models are dramatically more capable. But capability without context is just expensive latency. And in healthcare, the context lives behind a walled garden.The data gravity (the patient charts, the encounter histories, the medication lists, the imaging orders) sits in electronic health records. Epic. Cerner. Athena. And without that context flowing securely into AI systems, even the most sophisticated models are left prompting in the dark. As one survey found, hospitals on Epic had roughly 90% AI usage, while those on smaller EHR platforms averaged just 50%, a disparity that reveals how tightly AI adoption is coupled to infrastructure access.“AI is not the bottleneck,” Hadi argued. “It’s the context that’s the bottleneck right now. Models are pretty smart. But if you cannot get patient chart information securely into AI, you can do only enough.”What an AI “Skill” Means for HealthcareA skill, in this context, is a structured set of instructions that teaches AI how to perform a specific task when triggered by specific conditions. Think of it less as a prompt and more as a protocol manual for a very capable but context-dependent assistant.A prompt says: summarize this note. A skill says: whenever a patient mentions diabetes in an encounter, trigger a downstream workflow. Draft dietary counseling documentation for the staff. Generate a glucose monitoring plan. Prepare a patient-facing message at an appropriate reading level. Format all outputs according to this template. Ground clinical recommendations in these evidence-based guidelines.Hadi framed the clinical application nicely: “Healthcare workflows are very if-then-else logic. If BMI is 30, do this. If they have diabetes, go on this path. And traditionally with software systems, it was so hard to scale healthcare because who’s going to build this if-then-else logic? You’re going to rely on your dev team or maybe Epic consultants, and that takes forever.”Skills collapse that timeline. They translate clinical protocols (the ones that live in binders, in the heads of experienced nurses, in institutional memory that evaporates with staff turnover) into executable AI instructions. And critically, they can be built by clinicians, not engineers. You describe your workflow conversationally. The AI interviews you, iterates, produces the skill. You test it against real examples and refine. Looking to understand Claude’s skills better and see real-life examples? Check out this article below on meta-prompting (full article with in depth walkthrough)Consider the practical applications that emerged from our conversation: a pre-clinic screening skill that reviews a panel of patients before Monday morning, flags missing imaging orders, and surfaces relevant history in a style you specify. A prior authorization appeal skill that ingests a denial letter and produces a structured response matching the format that has historically succeeded with a specific payer. An independent medical examination skill that parses 6,000 pages of records into a timeline of treatment, imaging, and interventions, work that currently requires hours of manual review or a dedicated team.These aren’t hypothetical. We’re building and deploying versions of these at RevelAi Health right now, integrated with EHR data through FHIR resources, with the clinical team able to customize and test skills through a user interface rather than filing engineering tickets.The Compliance ReckoningThere’s another thread from our conversation worth pulling. Earlier this month, allegations surfaced that Delve, a Y Combinator-backed compliance startup that had raised $32 million, allegedly generated 494 fabricated SOC 2 Type II reports for its clients. The reports were 99.8% identical boilerplate, with pre-written auditor conclusions filed before companies even submitted their evidence. The auditors Delve marketed as “US-based CPA firms” were traced to offshore operations using virtual addresses.The revelation emerged, almost poetically, because someone left a Google spreadsheet open to the internet.For health tech, this extends beyond a compliance scandal to become an ecosystem problem. Hundreds of companies, including health tech startups handling protected health information, may now hold invalid security certifications. The ripple effects will tighten an already rigorous procurement environment at a moment when health system CIOs were only beginning to open the door to smaller vendors.“You can’t outsource security responsibility,” Hadi said. “If someone is trusting you with their patient data, you have a huge responsibility to protect it. Security and compliance is not a cost center. It’s the most important foundational thing you have to do.”We felt the FOMO ourselves at RevelAi. We went through Vanta, checked every box, invested heavily in governance, and watched competitors claim they completed SOC 2 Type II in three weeks. The temptation to move faster was real. But in healthcare, the “move fast and break things” mantra will also break your company. We’ve watched it happen. Babylon, once valued at $4.2 billion, collapsed in 2023. Olive AI, valued at $4 billion, shut down the same year. The outward appearance of success, it turns out, is often inversely correlated with the rigor underneath. Curious about the tools that I use to put together Techy Surgeon and leverage AI to improve my personal productivity? Check out my article below — The Clinician Founder’s AI Stack.Where the Bridges Are Being BuiltNot everything is stalled. The interoperability landscape is shifting, unevenly but meaningfully.Athena has emerged as an unlikely leader. At HIMSS 2026, the company previewed an industry-first Model Context Protocol server, infrastructure that allows AI agents to securely access patient chart data in real time. They’re building athenaConnect, an intelligent interoperability layer connecting 170,000 providers serving 20% of the U.S. population.This matters enormously. Model Context Protocol (MCP) is what makes skills practical at scale. It’s the plumbing that lets an AI agent not just follow instructions but access the clinical context those instructions require. When Hadi built a FHIR integration with Cerner’s proprietary APIs, it took him one hour using skill-based development. Previously, that work took two weeks. That’s the offline version, engineers using skills to accelerate code. The online version, where skills execute in real time against live patient data, is coming but isn’t here yet in production.Anthropic, notably, has published a FHIR skill on their marketplace, their engineers spending considerable time understanding HL7 and FHIR data exchange standards. The infrastructure players are taking healthcare seriously. The question is whether the healthcare systems will meet them halfway.Hadi offered what he acknowledged was a controversial take: “In a year or two, we’re going to see engineers or startups not writing code for integrating with EHR. There will only be a skill. The skill is not just a prompt. There could also be scripts, how you should be pulling patient chart data, how you should be looking at observation resources, medications, encounters. You give that skill to AI and the integration pathway, and AI just executes the command without engineers writing the code.”If you like deep dives on clinical AI and health policy, consider becoming a free or paid subscriber to Techy Surgeon!The Agentic Clinic, and Who Gets AccessThere’s one more implication worth sitting with. A listener named Jared asked a question during our live session that I’ve been thinking about since: With all the AI-native tooling coming out for documentation, billing, prior auth, RCM, do you think independent private practice could actually make a comeback?The answer is genuinely uncertain, but the possibility is real. This wave of AI is making individuals more productive faster than it’s making organizations productive. The procurement cycles that slow down a 500-bed academic medical center don’t apply to a four-physician orthopedic practice. One of our design partners, Dr. Kofi Agyeman at ASAP Orthopedics and Avors Medical Group , has adopted AI workflows at a pace that would take a large health system eighteen months of committee review.The equity implications cut both ways. Smaller practices may leapfrog larger systems in AI fluency, a reversal of the usual resource advantage. But the tools require literacy, and that literacy requires investment. The adoption of AI skews strongly toward well-resourced hospitals, and policymakers have noted this as an important equity issue: without deliberate support, high-quality AI may deepen rather than narrow the urban-rural divide.Outcomes, Not SoftwareThe thread that ran through our entire conversation, from Hadi’s Capital One days through Vincere Health’s Medicaid tobacco cessation work through what we’re building now, is that the technology is never the product. The outcome is the product.Hadi’s first company, Vincere Health, deployed a mobile app with a breathalyzer sensor to help Medicaid patients quit smoking using financial rewards. The technology was novel. But the real innovation was the outcome: measurable behavior change in a population that every public health expert will tell you is among the hardest to engage. They started with a peer-reviewed study at Washington Medical Center before scaling. They partnered with the existing infrastructure, the 1-800-QUIT-NOW quitlines that every state operates, funded by state tax dollars and contracted to carriers like Optum. And when Optum, which operates about 30 state tobacco quitlines, eventually acquired Vincere, it wasn’t a technology acquisition. It was an outcomes acquisition.That trajectory informs how we think about RevelAi’s direction. The next phase of clinical AI isn’t software as a service. It’s outcomes as a service. The distinction is more than branding. Software companies sell features: here’s what our chatbot can do, here’s our dashboard, here’s our analytics module. Outcomes companies sell results: your appointment volume will increase, your care gaps will close, you’ll stop losing money on these episodes of care.To get there, you need AI-fluent staff on platforms designed for clinical workflow, connected to systems of record through FHIR and emerging interoperability standards, executing skills that encode clinical protocols. You need the software to be ambient, running in the background, reading context from the EHR, writing results back as patient notes, in-basket notifications, task lists. Not a second platform clinicians have to learn, but an invisible layer that makes the existing one dramatically more capable.“We don’t want people living in two tools,” Hadi said. “Your view layer should remain the system of record, the EHR. You use separate tools to configure AI, how to run ambiently on your workflows.”What Comes NextWe’re launching a few things on Techy Surgeon worth mentioning. The Clinical AI Leadership Institute, a selective cohort of 30 clinicians and healthcare leaders for intensive virtual sessions on AI implementation, will be announced soon with a waitlist on Techy Surgeon. We’re also building out Claude for Clinicians, a resource hub with pre-made skills, prompts, and video tutorials for clinicians who want to start experimenting. You can find this at agentic-clinic.com, a growing resource hub I’ve started that will be a fully dedicated tutorial and learning site for Claude and other clinical AI tools and productivity tools. Because the gap on Market Street, between the people building these tools and the people who could most benefit from them, won’t close itself. At RevelAi Health, we’re building agentic clinic workflows that truly scale, embedding what we believe to be the clinical AI platform of the future, not just to automate tasks and workflows but to deliver population health for musculoskeletal conditions and beyond to close care gaps for millions of patients.The models will keep improving. That’s the easy part. The hard part is the architecture: the FHIR connections, the compliance infrastructure, the skills that encode clinical judgment, the change management that makes adoption stick, and the interoperability reforms that unlock the context AI desperately needs.The billboard in San Francisco advertises skills to people who already know what they are. The work that matters is translating those skills into languages that clinicians, practice managers, and patients understand, and into workflows that make Monday morning in a clinic measurably better than it was last week.The infrastructure is being built. The question is whether it will reach the people who need it most, or only those who can afford to build it themselves.Christian Pean, MD, MS is CEO and Co-Founder of RevelAi Health, Executive Director of AI & IT Innovation at Duke Health, and Assistant Professor of Orthopaedic Surgery at Duke University. He writes the Techy Surgeon newsletter on clinical AI and health policy for surgeons and health system leaders.This article is adapted from the inaugural Founders Coffee live session with Hadi, CTO and Co-Founder of RevelAi Health. You can watch the full conversation on Substack.If you like deep dives on clinical AI and health policy, consider becoming a free or paid subscriber to Techy Surgeon! This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Thank you Hadi Javeed, Mike Logan, MD, Rachel, and many others for tuning into my live video with Sarah Habeeb! Join me for my next live video in the app.🗓️ TEAM Connect Virtual Summit — April 30, 2026 A full-day virtual event for hospital leaders navigating TEAM implementation. Speakers from Mass General Brigham, CommonSpirit, AdventHealth, Duke-Margolis, and more. I’ll be presenting on technology-driven care coordination. Early bird registration closes March 20th. → Reserve your spot at apmconnect.com/virtual-summitSarah Habeeb is System Director for Medicare Value-Based Care Products at Baylor Scott & White Health and co-founder of APM Connect. Connect with her on LinkedIn or atDr. Christian Pean is an orthopedic trauma surgeon and faculty member at Duke University School of Medicine, core faculty at the Duke-Margolis Institute for Health Policy, and CEO and Co-Founder of RevelAi Health. He writes Techy Surgeon at the intersection of clinical AI, health policy, and care coordination. You can find him at the TEAM Connect Virtual Summit on April 30, 2026.Sarah Habeeb started pre-med at Texas A&M. Then she drove an ambulance, decided she didn’t want to touch patients, and pivoted into health administration. She was sitting in a grad school classroom around 2014 when someone started explaining the Medicare Shared Savings Program—a program that was, at the time, so new that almost no one understood it. She listened, thought it made intuitive sense, and made a career bet on it.That bet paid off. Today, she oversees a program that generates roughly $110 million in annual savings against a CMS benchmark, retaining about 75 cents of every dollar through MSSP’s Enhanced Track. That’s approximately $77 million flowing back out to physicians annually, for a program covering 120,000–125,000 Medicare beneficiaries across an entire major health system.She also co-founded APM Connect—a free community for hospitals navigating mandatory payment models—and will be speaking at the TEAM Connect Virtual Summit on April 30, 2026. I’ll be there too. If you’re a hospital leader trying to make sense of what TEAM actually requires in practice, this event is where you want to be.But before the summit, here’s what an hour with Sarah Habeeb taught me about what value-based care actually looks like from the inside—and what that means for every clinician and operator who is about to be pulled into it, whether they’re ready or not.The Mechanics Most Clinicians Never LearnRisk, in the value-based care sense, is a word that gets thrown around like clinicians should already know what it means. Most don’t. Here’s the actual structure:When you enter a total cost of care contract, a payer assigns you a benchmark—an expected spend per member, per year, risk-adjusted based on patient complexity. If you keep costs under that benchmark, you share in the savings. If you don’t, you may owe back a portion of the overage. The HCC risk adjustment model is how CMS calibrates those benchmarks—it’s essentially a formula that assigns a risk score to each patient based on documented diagnoses. A patient with diabetes, COPD, and heart failure carries a higher score than one with no documented conditions, so their expected cost is set higher.This is where documentation integrity enters the picture. Sarah is direct about it: “If you’re a cardiologist in the heart failure cohort, you need to be sure that you’re getting credit for the risk of your patients because that directly affects your benchmark, which then directly affects your performance in the contract, which affects your Part B adjustments.”Physicians often experience risk coding conversations as administrative irritation—another box to check, another form to sign. The translation layer is missing. If your patients are genuinely sicker than their documented diagnoses suggest, you’re being benchmarked against a population that looks healthier than yours. The comparison doesn’t hold. You look like you’re mismanaging costs when you’re actually managing a high-acuity panel with inadequate documentation.The fix isn’t gaming the system. It’s accuracy. And it starts with clinicians understanding why it matters.What “Operator” Actually Means on a Monday MorningBaylor Scott & White’s ACO is structured around what Sarah calls product owners—people responsible for specific contracts (MSSP, Medicare Advantage risk agreements, direct-to-employer deals) who identify their contract’s cost drivers and design remediation plans. Care management, quality teams, and marketing function as internal vendors to those product owners, not parallel departments chasing their own initiative lists.This sounds obvious until you’ve seen the alternative, which is how most health systems actually operate: multiple teams, multiple initiatives, no clear accountability, and no way to know what’s actually moving the needle.“Because if everybody’s working on a bunch of things and we’re not talking to each other,” Sarah explains, “you can’t figure out what actually made the difference.”The answer for Baylor was forced prioritization. Pick three initiatives. Measure them monthly. Make every stakeholder meeting about those three things. It’s not sophisticated—but discipline consistently outperforms sophistication in operations.The Inpatient Rehab Problem (Which Is Probably Your Problem Too)If you work in orthopedics, you already know what Sarah is about to say. If you don’t, here’s the version that will make you understand it.At Baylor Scott & White, inpatient rehab facility utilization is—in Sarah’s words—”completely unmanaged.” Against Milliman benchmarks, they’re over-utilizing inpatient rehab while simultaneously running near-zero skilled nursing facility use. They’re tracking 55% of hip fracture patients going to inpatient rehab, with a 30-something percent readmission rate that isn’t meaningfully better than SNF. “So was that the right decision? Were they even ready to discharge from the hospital?” she asks. And the honest answer is: often, no.This is the central tension in any ACO that’s embedded in a health system with joint ventures. The health system may have financial interests in keeping patients flowing through high-cost post-acute settings. The ACO’s job is to reduce unnecessary utilization of those same settings. Getting alignment between those two forces is genuinely hard. Sarah doesn’t pretend otherwise. “You have to find a way.”What Baylor has built: a six-to-eight-person team of post-acute care nurses who follow ACO patients through preferred SNFs, with direct EMR access (a requirement of network participation), weekly interdisciplinary calls with each facility, expected discharge dates set within seven days, and a target average stay under 28 days. They track return-to-acute rates and share scorecards with SNF partners quarterly. It’s brute-force infrastructure, but it’s working better than the alternative—which is patients in a black box beyond hospital discharge.From the orthopedic side, I’ve been doing telephone visits at two weeks for all my hip fracture patients, closing them with absorbable sutures so the visit doesn’t require a trip in. Half the time, I’m chasing down whether they’re still in the SNF, whether they’ve bounced back to a different ED, whether anyone has even talked to them since discharge. The clinical relationship doesn’t just end at the door of the facility. But the information infrastructure does.TEAM, ASM, and the Art of Not PanickingHere’s Sarah’s read on the two mandatory models that are consuming health system bandwidth right now:On TEAM: Most hospitals are in year one, upside only, and either don’t know they’re in it or aren’t taking it seriously. This is a mistake. The regional benchmarking structure means your performance is being measured against peers in your geography. “If your regional peers are paying attention and you’re not,” Sarah says, “that affects your benchmark.” By the time year two arrives with downside risk up to 20%, you’ll be starting from behind, not from neutral. Only four hospitals voluntarily opted into TEAM early from prior CJR participation—which tells you something about the expected economics. But ignoring it isn’t a viable option.On ASM: CMS’s Ambulatory Specialty Model is designed as physician-level accountability—NPI-specific participation, measuring cardiologists on heart failure costs and a range of specialists on low back pain. The design is conceptually provocative: specialists competing against each other within the same market for Part B adjustments. The implementation, however, came in lighter than expected. At Baylor, the initial projection was 200–400 physicians selected. The actual list: 51. Nationally, organizations are reporting five or six physicians per system. The cost of the required quality reporting infrastructure may exceed the penalty exposure for smaller lists.Baylor’s response: build a shadow bundle internally. Treat ASM as an MSSP workstream. Develop heart failure and low back pain strategies that produce dividends in 2027 regardless of what the formal adjustment looks like. It’s the right call. These conditions were selected because they represent genuine opportunities to improve care and reduce low-value utilization—guideline-directed medical therapy for heart failure, fewer unnecessary MRIs and high-risk opioid prescriptions for low back pain. The model may be imperfect, but the clinical direction is sound.Where AI Actually Fits—and Where It Doesn’tSarah told me upfront: she doesn’t spend much of her day thinking about AI. Her ACO generates north of $100 million in savings the old-fashioned way, through data discipline, care management operations, and physician engagement. “I try not to [use AI],” she said. “I’m like, don’t be the reason we need to build data centers.”That grounded skepticism makes her view of where AI could help more credible, not less.Her highest-value use case: discharge disposition prediction. A tool that could tell a case manager at the bedside—in real time—that this specific hip fracture patient, given their comorbidities and functional baseline, has a 38% readmission rate if sent to inpatient rehab versus a projected outcome curve if they went home with home health instead. “That sounds like magic to me. If that can be populated, that would be really helpful.” The connection between value-based care economics and bedside decision-making is currently made by humans who don’t have time to make it well.Her second use case: home health appropriateness review. Some patients are on home health for two years. The authorization review documents run ten pages of wall-to-wall all-caps text. Physicians batch-sign them without time to read them. AI that could synthesize that documentation into a structured summary—flagging patients who are likely no longer appropriate for the benefit—and send it to the physician for a judgment call, not a rubber stamp, would be genuinely valuable. “There needs to be some type of tool rather than a pre-review.”Both use cases share a structure: AI doing the parsing, pattern recognition, and synthesis; humans making the clinical judgment call. Augmented intelligence, as she put it. Not a replacement for the nurse who calls the SNF. A better brief before the nurse calls.The Summit You Should Know AboutAPM Connect—which Sarah co-founded with two former Common Spirit colleagues—has built the closest thing to a peer community for TEAM participants. It’s free for hospitals and clinicians. About 200 of the 729 mandatory TEAM hospitals have already joined. The resources are real: best-practice webinars, forums, a participant community, and expertise Sarah and her co-founders have spent years developing.On April 30, 2026, APM Connect is hosting the TEAM Connect Virtual Summit—a full-day virtual event for hospital leaders working through TEAM implementation. The speaker roster includes colleagues from Mass General Brigham, CommonSpirit, AdventHealth, Duke-Margolis, and Bass, Berry & Sims. I’ll be presenting as well, focused on technology-driven approaches to TEAM care coordination.Early bird registration closes March 20th. TEAM Connect members register at $250; non-members at $375. Group rates available for teams of three or more.If your hospital is in TEAM and you haven’t found your footing yet, this is a low-cost way to learn from people who are further along than you. The fact that only four hospitals voluntarily opted in early tells you something about the economics. But it also tells you that most of the 725 hospitals that didn’t raise their hand are now navigating this without a roadmap. APM Connect is trying to build that roadmap in public. Register here.What $110M Means in ContextBefore getting into ACCESS strategy, it’s worth grounding the Baylor Scott & White number in the broader MSSP landscape—because “top-performing ACO” is a phrase that gets used loosely, and the data tells a more interesting story than the label suggests.See Data Visualization of ACO savings hereIn 2024, the MSSP delivered $6.6 billion in gross savings—a record, and more than 20 times what the program generated in its first full year. Of 476 participating ACOs, 75% earned performance payments totaling $4.1 billion, while Medicare netted $2.4 billion in savings relative to benchmarks—the highest share of ACOs generating savings since the program’s inception.Here’s what makes BSW unusual: the median ACO generated roughly $8M in earned savings in 2024; top-quartile performers averaged closer to $20–25M. BSW’s ~$110M in gross savings against benchmark places it in a category occupied by fewer than a handful of systems nationally. That’s not a function of luck or favorable geography. It’s 10-plus years of compound operational discipline—the product owner model, the disciplined SNF network, the monthly initiative tracking—applied consistently at scale.ACOs that entered the program in 2012 or 2013 and sustained participation earned an average of $7.6 million net to the Medicare Trust Fund—versus $4.6 million for ACOs that started in 2019 or 2020. Time in program matters. There is no shortcut. Which is precisely why hospitals in TEAM that are treating year one as a grace period are setting themselves up poorly for year two.Two-sided risk is the engine: ACOs in Level E and Enhanced tracks generated more than two-thirds of all MSSP savings in 2024 ($5.4 billion of the $6.6 billion total) despite representing a minority of total ACO participation. The lesson is structural: downside risk changes behavior in ways upside-only contracts don’t. TEAM’s mandatory downside exposure beginning in year two isn’t punitive policy—it’s the mechanism that produces results.The Data Pipeline ProblemOne of the most practically useful things Sarah said, almost as an aside, is that the data infrastructure underneath a high-performing ACO is invisible to most clinicians and is harder to build than the clinical programs sitting on top of it.At BSW, claims data comes in through CCLF files from CMS and through bespoke payer feeds for Medicare Advantage contracts. A dedicated analytics team ingests all of it into a centralized repository, then layers in a vendor like Milliman for external benchmarking—so Sarah can see not just how BSW is performing against its own benchmark, but how it compares to peers nationally. The result is the “contracting grid” she described: a single view of every risk contract, every provision, every performance lever, with monthly measurement against the initiatives that matter.What this requires: a data team that speaks both ACO contract language and engineering. A vendor relationship for external benchmarking. A governance model that connects product owners to that data without creating a meeting-intensive bureaucracy that slows down action. And institutional patience—because the ROI on this infrastructure takes two to three years to materialize.For smaller hospitals entering TEAM for the first time, almost none of this exists. They’re getting quarterly CMS files, trying to reconcile them with what’s in their EHR, and hoping someone has time to run a query. The gap between those organizations and BSW isn’t primarily clinical. It’s informational.The ACO-ACCESS Playbook: What Operators Need to KnowThe CMS ACCESS Model generated an underappreciated moment in health system boardrooms. As Sarah put it, her health system went through “the five stages of grief once they figured out that they could not apply as an ACCESS participant themselves.”That reaction reveals something structurally important. Health systems are excluded from ACCESS participation because their financial incentives—keeping patients within their network, returning them for follow-up care—run counter to what the model is designed to do. An ACO, by contrast, is measured on keeping patients out of high-cost settings. These are genuinely different businesses, even when they share a tax ID.For ACOs considering whether to engage with ACCESS participants proactively, three operating principles apply:Move before the OAPs hit your total cost of care. Outcome Aligned Payments(OAP) from the ACCESS Service will count toward total cost of care calculations for attributed ACO patients after the grace period. The model is designed so that ACCESS participants substitute for low-value care such as imaging, injections, and certain surgeries that conservative management makes unnecessary. If it works as intended, OAP reimbursement costs should be more than offset by reductions in downstream utilization. But an ACO that hasn’t established a measurement relationship with an ACCESS participant before those services start appearing in claims data is flying blind. The time to start measuring is now, before the formal performance period creates noise in the baseline.Come with validated outcomes, not a vendor ROI deck. ACOs have actuaries. They have their own analytical infrastructure. The last thing a Sarah Habeeb wants to receive is a projection her team has to spend two weeks independently validating. ACCESS participants approaching ACOs should arrive with outcomes already assessed by a credible third party—something like a Validation Institute report—rather than a proprietary model the ACO has no way to audit. Shift the burden of proof to the participant, where it belongs.Understand what “substitute spend adjustment” means for your ACO partner. The ACCESS model financially penalizes participants who generate OAP utilization without a corresponding reduction in what CMS calls “substitute spend”—the downstream care that better conservative management was supposed to prevent. ACOs have a direct stake in this math. An ACCESS participant whose patients continue to receive the imaging, injections, or surgeries they would have gotten anyway is a cost increase to the ACO, not a savings lever. The right question for any ACO evaluating an ACCESS partnership: can this organization demonstrate that their model bends the utilization curve downstream? And can they do it with real claims data, not projections?The structural alignment between ACOs and ACCESS participants is genuine. But it only materializes if the ACCESS participant can prove it.The Physician Champion ArbitrageOne more thing Sarah said that deserves its own heading: if you’re a clinician who wants to have influence in value-based care, the path is easier than most people think, because there’s a scarcity of engaged physicians on the ACO side.“If you are going to be the physician in value-based care that’s engaged, you’re going to have a best friend in the ACO who are always looking for this—the physician that gets it, they’re engaged. And all of our ideas are going to be filtered through that person before we go take it to this broad group.”ACOs run on data and initiatives, but initiatives require physician behavior change, and behavior change requires trust. The ACO administrator who has a cardiologist champion who actually reads the quality reports is in a fundamentally different position than the one who doesn’t. If you’re an orthopedic surgeon, a cardiologist, or a low back pain specialist in a TEAM or ASM hospital, you have more leverage than you probably realize, and the ACO on the other end of that relationship is actively looking for you.Dr. Christian Pean is an orthopedic trauma surgeon and faculty member at Duke University School of Medicine, core faculty at the Duke-Margolis Institute for Health Policy, and CEO and Co-Founder of RevelAi Health. He writes Techy Surgeon at the intersection of clinical AI, health policy, and care coordination. Find him at the TEAM Connect Virtual Summit on April 30, 2026.Sarah Habeeb is System Director for Medicare Value-Based Care Products at Baylor Scott & White Health and co-founder of APM Connect. Connect on LinkedIn or at apmconnect.com. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Thank you Doctor Steven Murphy, Edward M. DelSole, MD, The Counterweight Pod, John Lee, Rez, and many others for tuning into my live video with Mike Logan, MD! Join me for my next live video in the app.Fourteen payers representing 165 million Americans just pledged to align with CMS’s ACCESS model by 2028. Venture capital funds are spinning up dedicated vehicles. Digital health companies are scrambling to retool their pitch decks. And the most interesting reaction I’ve heard so far came from an emergency medicine physician running government programs for one of the largest ACOs in the country.“This is potentially a windfall for ACOs.”Not the take you expected.I sat down with Dr. Mike Logan—medical director of government programs at Advocate Physician Partners, one of the nation’s largest clinically integrated networks covering over a million lives and redistributing $170 million in shared savings back to physicians and hospital partners this year—for a live conversation about accountable care, clinical AI, and what happens when CMS decides to pay for outcomes instead of activities. What emerged was a strategic framework that cuts against most of the ACCESS model commentary I’ve seen.The Payment RealityLet’s get the numbers out of the way, because they’ve dominated the conversation.CMS released the outcome-aligned payment amounts for the ACCESS model, and the reaction from the digital health world ranged from disappointment to alarm. For one year of managing a musculoskeletal patient: $180. Behavioral health: $180. Early cardio-kidney-metabolic disease: $360. Advanced CKM: $420. Waive the 20% coinsurance—which most participants will do as a beneficiary engagement incentive—and the MSK track delivers roughly $6 per patient per month in actual cash flow, with another $6 contingent on hitting a 50% outcome attainment threshold.That’s not going to sustain a care team of pharmacists and nurses managing a panel of 500 patients. And if you’re a digital health company that raised venture capital on the promise of this model as a standalone business, you have a problem.But here’s where the strategic calculus shifts entirely if you’re sitting where Mike Logan sits.The View from Inside the ACOAdvocate Physician Partners runs one of the larger MSSP programs in the country. They have roughly 5,000 participating clinicians, a clinical integration scorecard built on standard HEDIS measures, value-based care advisers embedded with every practice, and a data infrastructure that lets them identify patients with uncontrolled diabetes, uncontrolled hypertension, and chronic back pain in a matter of hours.The economics Mike described are inverted from how most people are thinking about ACCESS.If you’re already accountable for total cost of care—Part A, Part B, everything—and you’re spending tens of millions annually on care management, outreach, RPM, and CCM programs, then ACCESS doesn’t need to be profitable on its own. It needs to be cheaper than what you’re already doing, while potentially bending the cost curve on your most expensive clinical populations.“If you’re spending tens of millions on legacy RPM and CCM per year, you can do the same thing now potentially for a tenth of the cost,” Mike told me. And then the savings compound: reduced MACE events, fewer hospitalizations, even delaying dialysis by six months represents significant avoided spend for an organization bearing total cost of care risk.This is the strategic insight that most ACCESS commentary misses. The model wasn’t designed for a digital health startup to build a business around $180 per patient per year. It was designed to give organizations already managing populations a low-friction way to extend their reach—and a new reason to invest in the technology-enabled workflows they should have been building anyway, through partners.The Free SalesforceOne of the most striking things Mike said was also the simplest: “You basically have a legion of people that will be your salesforce for free.”He’s talking about the existing infrastructure of ACO quality teams, care managers, and value-based care advisers whose job is already to get patients on wraparound services. They’re already running patients through clinical filters in Snowflake. They’re already identifying who has uncontrolled A1c, who has chronic low back pain, who hasn’t had a Medicare wellness visit.For a digital health company partnering with an ACO, the most expensive problem in healthcare—patient acquisition—is already solved. The ACO has the attributed lives, the clinical data, the referring relationships, and the physician buy-in. The referring clinician earns roughly $100 per patient per year for coordinating with the ACCESS participant. The friction is intentionally low.Compare this to the alternative many digital health companies are considering: direct-to-beneficiary advertising and enrollment. Spray-and-pray outreach to Medicare beneficiaries, most of whom will be deeply skeptical of unsolicited health technology offers, all for $6 per month before outcomes reconciliation.The ACO pathway isn’t just cheaper. It’s the only one that makes operational sense at these payment levels.Where to Start: MSK and Behavioral HealthBoth Mike and I converged on the same conclusion, and it’s supported by the independent evidence.The Peterson Health Technology Institute’s 2024 evaluation found that virtual MSK solutions deliver clinically meaningful improvements in pain and function, with economic evidence supporting broader adoption. Behavioral health digital interventions have a similarly robust evidence base. The CKM and eCKM tracks, by contrast, require devices like continuous glucose monitors—which can run over $150 per month—and involve medication titration decisions that raise real questions about clinical workflows and liability at these payment levels.The MSK and behavioral health tracks also have the lightest clinical complexity. A patient-reported outcome measure for back pain or a PHQ-9 for depression is a fundamentally different data stream than real-time glucose monitoring with titration alerts. And if you pair a musculoskeletal patient who also has comorbid anxiety or depression—and the overlap in these populations is substantial—CMS applies only a 5% discount on the lower-cost track during overlapping months. That’s $360 for managing two conditions in a single patient, with an evidence base that actually supports the intervention.The harder tracks aren’t impossible. But they require a different level of clinical infrastructure, a closer partnership with the referring physician, and probably a willingness to subsidize pharmacist oversight from shared savings rather than from the OAP alone.The RPM ReckoningThere’s a harder conversation happening underneath all of this, and Mike didn’t shy away from it.Legacy RPM and CCM companies are facing what may be an existential pivot. The payer pledge commits commercial insurers to adopt outcome-aligned payment approaches by January 2028. If that holds, the arbitrage of measuring a patient’s blood pressure sixteen times a month and billing fee-for-service codes gets replaced by a requirement to demonstrate that you actually improved someone’s blood pressure.“I don’t really see how it’s not going to significantly affect RPM companies,” Mike said. And he’s right. The entire payment philosophy is shifting from reimbursing activity to rewarding outcomes. Companies built around time-based billing for remote monitoring minutes are facing a structural change in how the work they do is valued.But this isn’t purely a story of decline. For RPM companies willing to adapt, the transition period offers an opportunity: partner with an ACO for ACCESS-eligible Medicare patients while continuing to provide traditional RPM services for the organization’s commercial population. It’s a bridge, not a cliff—but it requires moving now, not waiting to see what happens.The Agentic AI QuestionMike and I both build things with Claude. He wrote an IRB proposal for a wearable-based sepsis readmission study using Claude that required only two edits from a seasoned researcher. He’s vibe-coded gamified GDMT compliance dashboards for heart failure patients. I’ve built interactive evidence spotlights and care coordination prototypes.The underlying insight we share: generative AI is going to deliver on the promise of value-based care by solving its scalability problem.If you’re an ACCESS participant managing 3,000 MSK patients, you need agentic workflows that can segment your panel, conduct text-based outreach at hours when patients actually respond, collect PROMs through conversational interfaces, and surface the 300 patients who need human attention this week. The math simply does not work if every patient interaction requires a licensed clinician.This is what CMS is betting on. Not that a company will hire an army of care coordinators at $6 per patient per month. But that AI-augmented workflows will compress the cost of engagement to a point where outcome-aligned payments become viable—and that the organizations who figure this out first will have a structural advantage that compounds over the ten-year life of the model.What Remains UnsettledThere are genuine open questions. Can ACCESS participants actually hit the 50% outcome attainment threshold across a diverse Medicare population? Will the payer pledge translate into real commercial contracts, or is it aspirational signaling before the OAPs were released? How will CMS handle the inevitable friction when ACCESS care reports land in PCP inboxes alongside the forty other notifications competing for their attention?And perhaps most importantly: will the clinical workflow for the CKM and eCKM tracks prove tractable at these payment levels, or will the MSK and behavioral health tracks become the only tracks that achieve meaningful enrollment?We don’t know yet. And anyone who tells you they do is selling something.The Path ForwardACCESS launches July 5, 2026, with applications due by April 1 for the first cohort. The window for building ACO partnerships isn’t months away—it’s right now.If I had to distill the strategic playbook from this conversation into a single sentence, it would be this: ACCESS is a wedge, not a business model—and the organizations that treat it as an invitation to embed within the infrastructure of accountable care will be the ones still standing when the payment amounts eventually increase.Because they will increase. CMS has built a ten-year runway. The payer pledge extends the addressable market beyond Medicare. And the organizations that demonstrate outcomes in year one will have data, relationships, and operational muscle that no late entrant can replicate.The question isn’t whether the payments are enough. The question is whether you can build something at these payments that becomes indispensable.Dr. Christian Pean is an orthopedic surgeon, health policy researcher, and CEO of RevelAi Health. He writes about clinical AI, healthcare policy, and care transformation at Techy Surgeon. The conversation referenced here was recorded live on Substack with Dr. Mike Logan, medical director of government programs at Advocate Physician Partners.Disclosure: RevelAi Health is developing tools for the ACCESS model MSK and behavioral health tracks. The author’s analysis reflects his own views and does not constitute financial or investment advice. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Thank you Shaleen Vira, MD, MBA, Audley Mackel III, Nick Lella, Darren Michael, and many others for tuning into my live video with Edward M. DelSole, MD! Join me for my next live video in the app.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.A spine surgeon in rural Pennsylvania just found out his entire practice was selected for a mandatory CMS payment model he’d barely heard of. His Medicare Part B reimbursement—all of it, not just the episodes in question—is now subject to adjustments of up to 9% based on how his low back pain patients do over time. He didn’t opt in. There is no opt out.This is not a hypothetical. It’s happening now to an estimated 8,600 physicians across orthopedic surgery, neurosurgery, pain management, anesthesiology, and physical medicine and rehabilitation. And it’s just one of five policy forces converging simultaneously on spine care economics.I sat down with Dr. Edward Del Sole—spine surgeon, former Geisinger faculty, and author of the new Substack newsletter The Spinal Column—for a live conversation about what’s coming. Ed and I trained together at NYU, back when we were the guys pulling paper charts at Jamaica Hospital. The world has changed. What follows is a distillation of that conversation, supplemented with primary sources, because the policy landscape facing spine surgeons right now deserves more than casual awareness.The Convergence No One Is Talking AboutEd’s debut article on The Spinal Column identified five policy forces reshaping spine surgery economics simultaneously. It’s worth naming them together, because their compound effect is greater than any single policy suggests:The Transforming Episode Accountability Model (TEAM)—a mandatory bundled payment program launched January 1, 2026, holding roughly 741 acute care hospitals in 188 markets accountable for surgical episodes including spinal fusion, from admission through 30 days post-discharge. Site neutrality expansion, which continues to compress the payment differential between hospital outpatient departments and freestanding offices. The inpatient-only list elimination, migrating more procedures to outpatient settings. Physician fee schedule erosion—a 33% decline in real-dollar Medicare physician payment since 2001, according to the AMA. And the Ambulatory Specialty Model (ASM)—the one that caught Ed’s practice by surprise.Any single one of these warrants strategic attention. Together, they represent a structural transformation of how spine care is financed, measured, and delivered under Medicare.The Ambulatory Specialty Model: Why the Employment Shield Is DissolvingASM is the policy that should keep spine surgeons awake at night—and that most haven’t yet internalized. Here’s why.Unlike TEAM, which operates at the hospital level, ASM is evaluated at the individual TIN/NPI level. Every physician. Individually scored. No hiding behind institutional averages. The model launches January 1, 2027, with payment adjustments beginning in 2029, ranging from –9% to +9% of all Medicare Part B payments—scaling to –12% to +12% by 2033. CMS will retain 15% of the total risk pool as built-in program savings, meaning this isn’t budget-neutral like MIPS. The government is designed to win.Ed put it bluntly during our conversation: “If you are a practicing clinician and you’ve been selected for ASM, you would be a fool to push it off even for a minute.”He’s right. The requirements are concrete and operational. For low back pain episodes, clinicians must collect Oswestry Disability Index scores and demonstrate functional improvement over time. They must screen for health-related social needs. They must establish collaborative care agreements with primary care physicians. And they’re scored on episode costs—spending they may have limited ability to control once a patient leaves their office.What struck Ed most—and what should strike every employed surgeon reading this—is what he called the dissolving “employment shield.” Historically, surgeons employed by health systems could rely on institutional infrastructure to manage quality reporting, MIPS compliance, and payment model participation. ASM breaks that pattern. It evaluates you, the individual clinician, regardless of your employment arrangement. As Ed noted: “Even as an employee, you are going to be responsible for the result here.”The Scorecard Gap: What You Don’t Know Is Costing YouOne of the most revealing moments in our conversation came when we discussed a question from Darren Michael of Forte Analytics: Do physicians actually look at their MIPS scores?Ed’s answer was striking in its candor. At Geisinger, across five years of academic practice, MIPS was a word he “never really heard.” The institution handled it in the background. When he transitioned to private practice, he discovered he needed to personally enroll. And even now, enrolled in MIPS, he described a system that delivers no usable scorecard to his inbox, no integration with his daily workflow, no actionable intelligence about how his patients are actually doing relative to peers.This is the gap that ASM will expose. Under ASM, clinicians will be scored relative to other specialists in their geographic region using decile-based benchmarking. If you don’t know your current performance on episode costs, patient-reported outcomes, and functional status improvement, you’re operating blind in a game where the stakes are real and the clock is already running.The infrastructure to deliver that intelligence—to surface a clinician’s performance on the measures that actually matter and then connect it to guideline-concordant recommendations for improvement—doesn’t exist natively in most EHR platforms. It’s one of the areas where purpose-built tools could make the difference between a 9% penalty and a 9% bonus, but only if clinicians engage with the data before 2029.The Social Determinants ParadoxHere’s an inconsistency worth scrutinizing. ASM mandates screening for health-related social needs—an acknowledgment that non-medical factors drive outcomes and costs. Yet the model includes no meaningful risk adjustment for social determinants in its cost benchmarks.TEAM, by contrast, incorporates a community deprivation index decile into its target price risk adjustment. ASM does not. CMS is effectively telling clinicians: identify the social factors affecting your patients, but don’t expect us to account for them when we score you on cost.Ed and I spent time on this tension. The research is clear—patients with adverse social determinants are less likely to complete patient-reported outcome measures, more likely to have ED utilization, and more likely to experience complications that drive episode costs. Penalizing clinicians who serve these populations without adjusting benchmarks accordingly is a policy choice with equity implications that deserve continued advocacy.The practical response, in the interim, is to both screen and act. Organizations like Understood Care—which partners with practices to deploy medical advocates who help patients navigate insurance, transportation, and care access using the community health integration CPT code (G0136)—represent one model for extending the care team without adding burden to the clinician.The Fee Schedule Erosion You Can’t IgnoreBehind all these model-specific dynamics sits a structural reality that Ed described in terms any practice manager would recognize: “You cannot really earn your money and your lifestyle by just seeing patients and even just doing surgery anymore.”The numbers support the claim. The AMA reports that Medicare physician payment has declined 33% in inflation-adjusted terms since 2001, while practice costs have risen approximately 39% over a similar period. The 2025 physician fee schedule applied an additional 2.83% cut. The 2026 schedule avoided further cuts but did not close the gap. Medicare physician reimbursement remains the only major healthcare payment category not tied to an annual inflation index.Meanwhile, physician compensation represents roughly 6–9% of total healthcare expenditure. The United States spent $5.3 trillion on healthcare in recent years, concentrated in hospital care and clinical services—categories where spending has continued to grow. Cutting physician pay while administrative and facility costs remain largely untouched doesn’t address the primary cost drivers. As Ed put it: “Trust in God, and all others must come with data.”For private practices, this fee schedule erosion compounds the other four forces. Margins compress from the revenue side while compliance requirements expand from the regulatory side. The strategic calculus for many independent surgeons is increasingly stark: invest in infrastructure to win under value-based models, or accept slow financial attrition.The Clinical AI Question: Where Technology Actually HelpsOur conversation pivoted from policy to technology—and to the question every clinician is quietly asking: where does AI actually make a difference in my practice?Ed identified three domains. Surgical planning, where AI-driven preoperative modeling and robotic assistance are already transforming complex spine reconstruction. Patient communication, where asynchronous, AI-augmented care coordination can extend the care team without adding staff. And clinical triage—the “AI-powered front door” that routes patients to the right clinician or care pathway before wasting a visit, a copay, and a referral.That last point deserves emphasis. Ed described a scenario any orthopedic surgeon would recognize: a patient calls their health system complaining of knee pain, gets routed to a joint reconstruction surgeon expecting a 70-year-old with arthritis, and instead encounters a 22-year-old with an ACL tear who needs a sports medicine physician. Wasted visit. Wasted copay. Delayed care. An intelligent triage system—one that understands clinical context before scheduling—could prevent this entirely. The technology exists. The integration challenge is real but not insurmountable.The elephant in the room, as always, is the electronic health record. Epic’s recent ambient scribe rollout and pre-charting tools signal that the dominant system of record is moving into territory previously occupied by startups. The pattern is familiar: incumbents arrive late with good-enough solutions that benefit from built-in distribution. For innovators, the path forward requires solving problems the EHR won’t solve well—care coordination beyond the four walls, regulatory compliance workflows, outcomes intelligence, AI-powered patient engagement across episodes. The startups that survive will be those that don’t compete with the system of record on documentation but instead extend care delivery into the spaces where the EHR has no presence.What Comes NextThe convergence of these five forces isn’t theoretical. TEAM is live. ASM participant lists arrive this summer. Fee schedule erosion continues. Site neutrality and IPO list changes are accelerating the shift to outpatient care. And the infrastructure to succeed under these models—the outcomes tracking, the care coordination, the intelligent patient engagement—takes time to build.Ed’s closing thought captures the urgency: “To win the ASM game, you need to be strategizing like yesterday about how to get there, because it’s going to require not just your thoughts. It’s going to require active implementation, development of a legitimate strategy, potentially capital expense.”The surgeons who prepare now—who understand their episode costs, establish care coordination agreements, invest in outcomes infrastructure, and begin collecting the measures that will determine their payment adjustments—will have a structural advantage over those who wait. The employment shield is dissolving. The scorecard is coming. The question is whether you’ll be ready to read it.Christian Pean, MD, MS is CEO and Co-Founder of RevelAi Health, Executive Director of AI & IT Innovation at Duke Health, and Assistant Professor of Orthopaedic Surgery at Duke University. He writes the Techy Surgeon newsletter on clinical AI and health policy for surgeons and health system leaders.Dr. Edward Del Sole is a spine surgeon at Keystone Spine and Pain Management in Reading, Pennsylvania and author of The Spinal Column on Substack.Disclosure: Dr. Del Sole serves in an advisory capacity to RevelAi Health and uses its platform in his clinical practice. The policy analysis in this article reflects the authors’ independent perspectives. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

I first became aware of Doximity GPT the way many clinicians do—through a peer's recommendation. My brother who was an internal medicine resident at the time, (now a hospitalist), was using their AI scribe over a year ago when it was still in beta. “Have you seen this thing? It’s free,” he texted me. That early signal—clinicians organically discovering and adopting a tool—turns out to be core to Doximity’s entire philosophy.When my three-year-old recently may have swallowed a plastic fork tine at a restaurant (he didn’t, thankfully), I found myself cycling through the emerging ecosystem of clinical AI tools: Open Evidence, Doximity GPT, even poison control. Each gave slightly different guidance, each had its own approach to surfacing evidence. It crystallized a question I wanted to explore: in this rapidly converging landscape of clinical AI, what actually differentiates one tool from another?Visual Prompt 1: Split-screen comparison showing a worried father with phone at restaurant table on left (warm editorial illustration style), transitioning to floating interface elements of multiple clinical AI apps (Open Evidence, Doximity GPT, Up-to-Date) on right, with subtle connecting lines between them suggesting the question of convergence—modern editorial style, muted teals and warm graysThe User-First PhilosophyDr. Amit Phull, Doximity’s Chief Clinical Experience Officer and an emergency medicine physician, joined the company in 2014 through what he calls “equal parts luck and serendipity.” With a background in computer engineering before medical school, he was part of Doximity’s early physician advisory panel, beta testing their software. At a 2014 medical advisory board meeting, when leadership mentioned they were looking for technically-minded clinicians, Phull essentially raised his hand and said, “What about me?”That origin story matters because it encapsulates Doximity’s approach to product development. “Our process is what differentiates us,” Phull explained. “We have a very broad network of clinicians that comprise the membership of Doximity, and we have the capability to engage those connections and directly ask them how we might improve this product to be most useful in their clinical practice.”The emphasis on “improve” is deliberate. “Our approach is not that medicine was solved last Tuesday,” Phull said, pushing back against what he sees as hubris in healthcare AI. “Anyone who presents any technology as having been the solution to everything that ails healthcare—there’s a healthy amount of hubris there.”From Administrative Assistant to Clinical ReferenceDoximity GPT’s evolution reflects this user-driven philosophy. In early 2023, the company surveyed hundreds of clinicians about where AI could have the most impact. The answer was overwhelming: administrative burden. Well-designed AI tools, clinicians estimated, could save 12-13 hours per week—”a ridiculous amount of time,” as Phull put it. You could do two additional operations or see 20-30 more patients.So Doximity GPT launched focused on being an administrative assistant—helping with documentation, prior authorizations, and appeals. The “unsexy part of AI,” Phull called it, acknowledging data showing that for every hour of clinical work, the average clinician does two hours of “other stuff.”But then something interesting happened. Despite being fashioned as an administrative tool, a significant percentage of queries were actually clinical reference questions. Users were asking for clinical decision support.Visual Prompt 2: Flow diagram showing the evolution of Doximity GPT—left side showing “Administrative Burden” (prior auths, documentation, appeals) with bar graph showing 12-13 hours saved, flowing through user feedback loop in center, emerging as “Clinical Reference” (clinical questions, decision support, peer review) on right—clean modern editorial style with flowing connection linesAt their March 2025 medical advisory board, Doximity floated the idea of leaning into clinical reference. Advisors made their requirements clear: HIPAA compliance, instant lookup capability, peer review, and transparent citations. By summer’s end, Doximity had acquired Pathway, a company co-founded by clinicians that had built a massive semantic graph of medical knowledge—guidelines, peer-reviewed drug monographs, and interconnected citations that could interface with large language models to improve reliability.“In about nine weeks, we did nine months or nine years worth of work,” Phull joked. They integrated Pathway’s entire tech stack and re-released Doximity GPT with clinical reference capabilities.The Convergence QuestionThis brings us to the elephant in the room: convergence. UpToDate, Open Evidence, Glass Health, Expert AI—they’re all racing toward similar functionality, all surfacing citations, all claiming to reduce hallucinations. What differentiates them?Phull’s answer was nuanced. He positioned Doximity as occupying the middle ground in a spectrum. On one end sits UpToDate—humans creating consensus, then layering AI on top. On the other end are tools that lean more heavily into pure AI capabilities. “We’re kind of right in the middle,” Phull explained. “We have technology that enables us to move at a speed that the UpToDate process probably could not. And we also have this human layer—a very broad and extensive network of folks who have domain expertise.”The key difference is the “order of operations”—technology for speed, human expertise for verification, but sequenced differently than traditional approaches.This philosophy manifests in PeerCheck, an initiative Doximity announced recently, co-chaired by Dr. Eric Topol and Dr. Benjamin (readers can learn more at the Doximity blog). The program leverages hundreds of thousands of Doximity members who are primary authors of cited literature, bringing them “behind the curtain” to verify AI representations of their own work.“When there are areas that a pure technology might give variable answers—you ask the same question 100 times and get some variation—we can solve for some of those problems by maintaining human expertise right at the center of clinical decision making,” Phull said.The Patient-Facing QuestionPerhaps the most interesting tension in clinical AI is whether these tools should be patient-facing. As Phull noted, “Even if we didn’t want to put these tools in the hands of patients, they’re gonna be in the hands of patients.” He trained in the “WebMD or Dr. Google generation,” where patients arrived armed with information and pre-formed decisions.But Doximity’s answer is “not yet”—and maybe never. “The way we’re structured as a company, we’re not inherently patient-focused,” Phull explained. Their focus remains on empowering clinicians alongside their patients. They’ve made docs.doximity.com publicly accessible (with caps after a few queries), primarily to make it easier for clinician members to access.“Our summary position is not that we’re trying to make Doximity GPT restricted from patients. It’s just not our personal point of focus as a company—at least not yet,” Phull said. “We hope that the more empowered we make our clinician user base to leverage these tools to advance patient care, the patients are gonna be right along with them for that ride.”Visual Prompt 3: Conceptual illustration showing clinician and patient together looking at screen/interface (suggesting collaborative use rather than separate patient tool)—warm editorial style showing partnership, with subtle Doximity interface elements floating nearby, modern healthcare settingBeyond the Hype: Practical AIWhat stuck with me most from our conversation was Phull’s repeated emphasis on “practical AI”—technology that’s actually usable in the real world, not just impressive in demos. This means navigating health system approvals, ensuring HIPAA compliance, building EHR integrations, and maintaining trust.“The technology is just the technology,” Phull said. “It’s not gonna show up and suddenly be operating on your patients. The way this technology actually does anything is clinicians participating in its development and its deployment. That’s where we think we win. And frankly, it’s not even about winning—it’s about making sure healthcare doesn’t miss this boat.”When I asked what we should be excited about on Doximity’s roadmap, Phull reframed the question. “Viewing artificial intelligence as a product unto itself permits folks to maybe wrap their mind around it in one way, potentially get caught up in hype around it, and think about it as this wholly separate element. The actual unlock of Doximity GPT is infusing its capabilities into the entire ecosystem we’ve built for clinicians over the last fifteen years.”That ecosystem includes one of the most broadly used telemedicine platforms in the United States, a comprehensive news and content machine, and fundamentally, a network that connects clinicians across the healthcare system. AI woven through that infrastructure, rather than sitting as a standalone product, represents Doximity’s vision.The Real DifferentiatorAs clinical AI tools converge on similar features—citations, peer review, specialty-specific training—what actually matters may not be the technology itself, but the process of development and the infrastructure for deployment. Doximity’s advantage isn’t just in what they’ve built, but in how they’re building it: listening to hundreds of thousands of clinician users, iterating based on real-world usage patterns, and integrating AI into workflows clinicians already trust.The race isn’t really to build the smartest AI. It’s to build the AI that clinicians will actually use—and that patients will actually benefit from. That requires something more than algorithms. It requires understanding that medicine wasn’t solved last Tuesday, and won’t be solved next Tuesday either. It requires the humility to let users shape the product, and the infrastructure to deploy it where it matters.“We’re just at the tip of the iceberg,” Phull said. Given how quickly the landscape is shifting—Doximity went from administrative assistant to clinical reference powerhouse in under a year—he’s probably right.For clinicians interested in participating in PeerCheck or learning more about Doximity GPT’s clinical reference capabilities, visit the Doximity blog or reach out to Amit at [email protected]. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Knowtex is a 2022 women-founded company led by Stanford AI scientists that is headquartered in San Francisco, building ambient clinical intelligence to transform how clinicians capture and use medical information. Designed to be EHR-agnostic, specialty-specific, and deeply integrated into workflows, Knowtex enables providers to generate complete, accurate notes, codes, and orders in real time. By combining clinical-grade AI with enterprise-grade security and speed, Knowtex helps health systems and providers reclaim time, reduce burnout, and deliver comprehensive patient care.Knowtex is backed by Y Combinator, Amazon Web Services (AWS), the UCSF Rosenman Institute, and MedTech Innovators among others. To learn more, visit www.knowtex.aiRecent Press Release on their rollout in Kansas below as well!Knowtex Successfully Launches Ambient Clinical AI at VA Kansas City Site, Advancing Multi-Phase Rollout Across VA Primary Care This is a founder story that announces itself through a single anecdote. Caroline Zhang, CEO of Knowtex, tells hers about a 75-year-old oncologist who had spent his entire career resisting technology.“He had never used a scribe,” Zhang told me during a recent conversation at JPM Healthcare Conference. “He would print out twenty pages of paper every day for chart review and pre-charting. He liked to write up the notes too. He saw notes as his legacy.”His verdict on ambient AI? “I’ll have to die before this technology can help me.”Six months later, that same physician was using Knowtex for every patient visit—going from fifteen patients daily to over twenty—and spending his reclaimed hours reading clinical research papers rather than documenting encounters. His new position: “Do not take it away from me.”The conversion of a single skeptic might seem like thin evidence. But it captures something essential about where clinical AI has landed in early 2026: the technology has crossed from “promising but unproven” to “infrastructure I cannot practice without.” And Zhang believes we’ve only glimpsed the beginning of what ambient technology can become.Enjoy this article and podcast? Consider becoming a free or paid subscriber to Techy Surgeon!The Operating System ThesisKnowtex occupies an increasingly crowded market. Ambient scribes—tools that listen to clinical encounters and generate documentation—have become healthcare’s fastest-adopted AI category. Abridge recently raised $300 million at a $5.3 billion valuation. Nuance DAX, now owned by Microsoft, has deployed across thousands of sites. Epic announced native ambient capabilities at its 2025 User Group Meeting, sending a shiver through every startup in the space.Against this backdrop, Zhang makes a distinction that shapes how she thinks about the market: scribing is a feature; ambient is a platform.“When it comes down to it, I think workflows stem from the doctor-patient visit,” she explained. “Even intake and everything else…we want to work in the center of everything.”If you enjoy reading founder stories and staying on the cutting edge of where clinical AI and health policy are going, consider becoming a paid or free subscriber to Techy Surgeon. The argument is that whoever captures the clinical conversation captures the most valuable data in healthcare. Not claims data, which is downstream and filtered for billing. Not chart data, which is shaped by documentation requirements and time constraints. The actual conversation—what the patient said, how the physician responded, the nuances that never make it into a structured note.“We have over a million clinical, real clinical conversations in our dataset,” Zhang noted. “That’s a novel dataset, and that informs the rest of the product and workflow design.”The vision extends beyond documentation. Zhang describes a “voice AI operating system”—a command center that can surface fifteen years of patient history in a face sheet, alert clinicians to clinical trial eligibility mid-visit, complete coding and orders within thirty seconds of encounter end, and push everything back into the medical record for downstream billing and quality teams.“It really becomes the central focal point for upstream and downstream workflows,” she said. “That can eliminate the reliance on EHRs that historically have been the data repository, the billing system, the reminder, the alert—all those types of things. Ambient can abstract all that away.”The VA BreakthroughThe scale of Zhang’s ambition became tangible in October 2025, when Knowtex was selected alongside Abridge as one of two ambient AI vendors for deployment across the Department of Veterans Affairs—170 medical centers and 1,193 outpatient clinics serving 9.1 million veterans.The contract, valued at $15 million for Knowtex’s portion, represents more than revenue. It represents federal validation for a company that emerged from Y Combinator’s Summer 2022 batch with modest seed funding. And it required solving a problem that has defeated larger, better-funded competitors: integrating with CPRS, the VA’s notoriously antiquated electronic health record.“The VA and the government side—they’ve allocated CTO resources, so their team has been a true partner in that sense,” Zhang said. “That’s also how we approach integration as a whole. If we can work with the provider and we can work with the health system and they are willing to think beyond EHR limitations, then we can build out API connectivity, FHIR API connectivity, HL7—all the things.”The subtext is important. CPRS, the legacy VistA system, is not known for its interoperability. That Knowtex achieved integration where others have struggled speaks either to unusual technical capability or unusual government partnership—likely both.Zhang frames it as a product philosophy rather than a technical achievement. “I would recommend to other folks—still makes sense to be in EHR marketplaces. We’ve worked with Epic, Cerner, Athena, OncoEMR, Flatiron. We definitely have earned our stripes in EHR integration. But we believe there’s more that you can do with innovating directly with the provider.”The Metrics That MatterWhen I asked Zhang about the KPIs she tracks internally and shares with health system buyers, the list was specific:Technology performance: 97.3% medical accuracy for audio processing, voice activity detection, and custom speech recognition. Over 99% accuracy for medical entity extraction—the translation of spoken information into notes, coding, and orders.Clinician outcomes: Two hours saved per day for adopters. A 29% decrease in “pajama time”—the evening and weekend hours physicians spend completing documentation—documented through internal Epic Signal data.Financial impact: Over $92,000 per month in additional revenue capture at implemented sites, primarily through surfacing ICD-10 codes and evaluation and management suggestions that busy clinicians would otherwise miss.Adoption: Above 80-90% utilization across specialties, which Zhang considers the leading indicator. “Initially, when you roll out AI, it’s a lot to tell the hospital stakeholders that you’re going to get XYZ results for your system—that’s just not something that we can guarantee for an organization. But when you see good utilization, then we hear all the happiness and the time saved and the burnout reduction.”The University of Rochester Medical Center case study, published in February 2025, offers independent verification of similar claims. URMC physicians reported reducing documentation time from ten hours weekly to two hours weekly. The revenue capture calculation—16,000 appointments monthly, 10% undercoding rate, approximately $58 revenue leakage per appointment—arrives at figures consistent with Zhang’s claims.The First CustomerThe conversion of the 75-year-old oncologist took six months. Zhang’s team was developing the product in his clinic, iterating on feedback from a user who represented the hardest possible case.“Someone who delivered such quality of care but in his lifetime told us that the EHR had only increased his workload,” Zhang recalled. “He didn’t feel like he was operating at his competency level. He was spending two to three hours, four hours every night, just working on documentation.”The breakthrough wasn’t a feature or a demo. It was sustained exposure to technology that actually worked—that didn’t require him to change his workflow but rather eliminated the parts of his workflow he had always resented.“I think that’s the biggest win we can have with anybody,” Zhang said. “Now our technology is something that is a daily reliance in your life and that is adding value and that is making the joy of medicine possible for you again.”The observation echoes across ambient scribe implementations: physicians often don’t realize how dependent they’ve become until the tool is unavailable. The ultimate test of product-market fit may be whether users panic when it goes down.The Patient-Facing QuestionOur conversation turned to a topic generating considerable anxiety in clinical circles: patient-facing AI. OpenAI launched ChatGPT Health in early January 2026, offering medical record integration through b.well and enterprise rollouts to HCA Healthcare, Cedars-Sinai, and Stanford Medicine. Anthropic followed days later with Claude for Healthcare, emphasizing HIPAA-ready infrastructure and partnerships with Banner Health, Sanofi, and Novo Nordisk.Both products target patients directly, positioning AI as a first-line resource for health questions, a development that evokes memories of “Dr. Google” but with vastly more sophisticated capabilities.Zhang’s response was measured but firm.“I’m going through some of a journey here too,” she acknowledged, describing a personal shoulder injury that led her to run her own MRI report through GPT and Claude. “Ultimately when my decision-making process comes down to it, I’m still trusting the expert. I’m still trusting the orthopedic surgeon and the PT as the person quarterbacking my medical care.”The concern isn’t that AI can’t provide useful information. It clearly can. The concern is about accountability, confidence calibration, and the subtle ways AI can speak authoritatively about matters where uncertainty should dominate.“When we’re in healthcare, we believe people still want humans being the face of the interaction,” Zhang said. “We believe that people want expertise. It’s really about supercharging the clinician, not replacing them. We believe that it’s about making the experience more human and not less.”This framing positions ambient scribes as fundamentally different from patient-facing AI. The clinician remains in the loop, signing off on notes, reviewing coded suggestions, approving orders. The AI augments rather than replaces clinical judgment.But the lines may blur faster than anyone anticipates. If ambient technology evolves into an “operating system” that surfaces recommendations, flags clinical trial eligibility, and alerts clinicians to missed diagnoses, how different is that from patient-facing AI that does similar things for the patient directly?“It’s about being hallucination-proof,” Zhang emphasized. “Let me say best-in-class model. Even in the ambient side, we can say there’s a spectrum of tools that have different performance. Contextual linked evidence from the chart history—those are metrics we’ve co-designed with clinician leaders and clinician stakeholders.”The distinction between clinician-facing and patient-facing AI may ultimately depend less on the technology’s capability than on who bears responsibility when it’s wrong.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.The Epic QuestionNo conversation about ambient AI in 2026 can avoid the elephant in the room: Epic’s entry into the market.At the August 2025 User Group Meeting, Epic unveiled three AI assistants—Art for clinicians, Emmie for patients, Penny for revenue cycle—with licensing beginning in 2026. The announcement represented a fundamental shift. EHR vendors had historically been platforms for third-party innovation; now Epic was building competing products directly.Survey data suggests 67% of outpatient providers may switch ambient vendors within three years. The question is whether startups can differentiate fast enough to survive commoditization.Zhang’s answer is speed.“That’s also the thing—velocity is going to be their Achilles heel,” she argued. “You can’t wait for another decade of a clinician’s life, or a year of a clinician’s life, to have tools that can save hours today.”The thesis is that Epic, constrained by its incumbent customer base and enterprise release cycles, cannot innovate as rapidly as focused startups. Whether that proves true over a multi-year horizon remains uncertain. But it explains why Knowtex and peers are racing to expand from documentation into ordering, clinical trial matching, and full-workflow integration—building defensible positions before the giant catches up.The AdviceWhen I asked Zhang for the best advice she’d received as a founder, she returned to relationships.“Build long-term relationships in the industry,” she said. “Things move at speed of trust. Today I was meeting with somebody I’d met two JPMs ago that told me they were not ready for ambient in 2023. It would have to be an RFP in 2025. And lo and behold, it’s 2025. They’re ready.”The worst advice? “That there is a set playbook that you have to follow for adoption of technology.”Her counter-lesson: find the customers who understand what you’re building. Some hospitals aren’t ready for ambient AI—and no amount of persuasion will change their procurement timeline. Others are actively searching for innovation. The founder’s job is to identify which is which without wasting months on the former.“You’ll talk to those individuals—they’ll say this is not possible,” Zhang reflected. “But then you just have to go find somebody who understands the magic you’re creating and the value you’re generating.”What Comes NextZhang’s 2026 roadmap centers on order entry—the logical extension of ambient documentation. If the AI can capture the clinical conversation accurately enough to generate a note, it can capture it accurately enough to propose the orders implied by that conversation.“I think this is a year where having order entry is going to be table stakes for ambient solutions,” she predicted. “You go into a room, you walk out. Now it’s not just a note, but you are going to expect that your ambient tool can do all those things—notes, coding, orders from the conversation.”Beyond that, she envisions embedded research access, clinical trial matching, and the gradual absorption of workflows that currently require context-switching between systems.“The clinician can be able to have evidence-backed research inside the ambient platform or inside their patient visit workflow,” she said. “Not having to switch screens—not having to look up things in GPT or Open Evidence or Doximity—but to actually do this right in the ambient workflow.”Whether this vision succeeds depends on factors beyond any single company’s control: regulatory evolution, payer acceptance, Epic’s competitive response, and the uncertain boundary between clinician augmentation and patient-facing AI. But the direction is clear. The clinical encounter—the conversation itself—is becoming the center of gravity for healthcare AI.The 75-year-old oncologist who thought he would die before AI could help him may have been the leading indicator. The question now is whether the technology can scale fast enough to reach the millions of clinicians still waiting—before the landscape shifts again.Knowtex is backed by Y Combinator, Amazon Web Services (AWS), the UCSF Rosenman Institute, and MedTech Innovators among others. To learn more, visit www.knowtex.aiChristian Pean, MD, MS is CEO and Co-Founder of RevelAi Health, Executive Director of AI & IT Innovation at Duke Health, and Assistant Professor of Orthopaedic Surgery at Duke University. He writes the Techy Surgeon newsletter on clinical AI and health policy for surgeons and health system leaders.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

We’re taking a short break from health policy and the 12 days of ACCESS to rehash a clinical AI story. In 2025, ambient scribes proliferated and they got squeezed as well. I anticipate as native EHRs roll out more of their ambient scribes, the market is going to contract and contort in interesting ways. Ambient scribes have become the default “first use case” for clinical AI.Vendors promise hours back to doctors, reclaimed evenings, and a future where the note writes itself while we focus entirely on the patient.A recent STAT News piece pulled together the best evidence we have so far on whether that vision is actually coming true. I turned it into a short Techy Surgeon video explainer you can watch above (love making these videos, it’s just very gratifying), but I also wanted to lay out the numbers here.The short version: ambient scribes don’t yet deliver the clean financial ROI many people imagine…but they DO seem to help clinicians feel less crushed by documentation. And that matters.Enjoy articles on clinical AI, health policy, and venture capital? Subscribe to Techy Surgeon to receive new posts and support this work.What the studies show on “time saved”Across the early randomized and observational studies, the story is remarkably consistent:* In a randomized trial at UCLA, clinicians using AI-powered scribes saved on the order of 20–30 seconds per visit compared with usual care.* In a study from Wisconsin using Abridge, ambient documentation reduced total documentation time by about 22 minutes per day.Those are not trivial numbers, but they are also not the “two hours back every clinic day and an extra visit to squeeze in” narrative that dominates marketing decks.That’s a big, sobering gap on expectations if you’re a CFO evaluating ROI purely on time savings.So why are clinicians (and investors) still excited?Because something else is showing up consistently in the data:Even when the time savings are modest, burnout scores improve.In these early studies, clinicians using ambient scribes report:* Less emotional exhaustion* Lower mental load from documentation* A generally better experience of their clinic dayThe hands on the clock don’t move much, but the experience of work does. That’s hard to price in a traditional ROI model…and impossible to ignore if you’re actually practicing medicine.The missing layers for true financial ROIIf ambient scribes only replace human scribes or take a small bite out of after-hours charting, the math is shaky.For these tools to generate clear financial ROI, they’ll need to plug into a broader automation workflow:* Upstream: smarter intake, templated plans, and structured data capture.* Downstream: automatic coding, quality reporting, and care coordination tasks triggered from the note.* Across the continuum: reuse of the same structured data for value-based programs (ACCESS, TEAM, PRO-PMs, etc.).Right now, most ambient scribe deployments still stop at “nice note, slightly faster.” The next wave of use cases must look more like “note → data → automated work that someone used to do manually.”Emotional ROI is still ROIAs a surgeon and founder, I’m wary of tools that are all story and no signal. But I love my ambient scribe. And I’m also honest about the fact that feeling less ground down by documentation has its own incredible value:* It can help keep people in practice.* It can make full clinical days sustainable.* It can create the mental space needed to actually engage with patients, learners, and quality improvement.If a tool barely moves the clock but helps clinicians feel less burned out, I wouldn’t call that a failure. I’d call it baseline value, and then challenge the ecosystem to build the automation around it that makes the financials work too.🎥 Video explainer:I walk through these studies and the ROI question in more detail in the Techy Surgeon explainer video embedded above. If you want to see how I make these video explainers in under an hour using Gemini 3 pro, Claude, and CapCut, consider becoming a paid subscriber to get access to my Techy Surgeon tutorials!If you’re experimenting with ambient scribes in your clinic or health system, I’d love to hear what you’re seeing…both on the clock and in how it feels to practice.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Part 2 of a 12-part operator series on the CMS ACCESS ModelToday, we’re covering how ACCESS fundamentally restructures Medicare payment for chronic disease management. This is truly a different financial architecture than fee for service. Here’s how the mechanics work.The 50% Withhold StructureThe payment model uses a split disbursement approach with backend reconciliation. During the 12-month care period, CMS distributes up to 50% of the total annual Outcome-Aligned Payment (OAP) through quarterly installments. You submit monthly claims using track-specific G-codes to your Medicare Administrative Contractor (MAC), which processes them as “zero-paid.” The Innovation Payment Contractor (IPC) then issues actual payments quarterly based on validated claims.The remaining 50% sits in escrow until the care period concludes. This isn’t arbitrary—it creates the financial mechanism for outcome accountability while providing operational cash flow during active care delivery.Clinical Outcome Adjustment: The Performance ThresholdAt reconciliation, CMS calculates your Outcome Attainment Rate (OAR)—the percentage of completed 12-month care periods where patients met all required outcome measure targets for that track. This gets compared against the Outcome Attainment Threshold (OAT).In Year 1, the OAT is 50%. If your OAR equals or exceeds 50%, you receive full payment with no adjustment. If your OAR falls below 50%, you receive proportional payment calculated as (OAR ÷ OAT) × full OAP amount.Example: If 40% of your completed care periods met all outcome targets:* OAR = 40%* Calculation: 40 ÷ 50 = 0.80* You earn 80% of full OAP* Clinical Outcome Adjustment = 20% reductionThe adjustment caps at 50% reduction. If only 20% of patients met outcomes (which would yield 40% payment under the formula), you still receive 50% of the gross OAP. However, consistently performing below minimum thresholds subjects you to termination under the Participation Agreement.The OAT will increase in subsequent participation years beyond 50%, creating a ramp that balances initial accessibility with long-term accountability. CMS will publish the year-by-year OAT schedule in 2026 before the first application deadline.Substitute Spend Adjustment: Preventing Duplicative CareThe second adjustment mechanism addresses care coordination and duplicate spending. CMS calculates your Substitute Spend Rate (SSR)—the percentage of aligned patients who did not receive specified substitute services from other Medicare providers for the same qualifying condition during their ACCESS care period.Each track includes a defined Substitute Spend List identifying services that represent new care initiation for the same diagnosis. For eCKM/CKM, this includes ambulatory blood pressure monitoring setup (CPT 93784, 93786, 93788, 93790), remote physiologic monitoring device setup (99453, 99473), diabetes self-management training (G0108), intensive behavioral therapy for cardiovascular disease (G0446) or obesity (G0447), medical nutrition therapy initial visits (97802), and MDPP enrollment (G9880, G9881, G9886, G9887).For MSK: physical therapy evaluations (97161-97163), occupational therapy evaluations (97165-97167), and remote therapeutic monitoring setup (98975). For BH: digital health medical treatment device supply (G0552-G0553), initial individual psychotherapy (90832-90834, 90836-90838), and RTM setup (98975).The Substitute Spend Threshold (SST) in Year 1 is 90%. If your SSR equals or exceeds 90%, no adjustment. Below 90%, proportional payment as (SSR ÷ SST) × full OAP.Example: If 85% of your patients avoided substitute services:* SSR = 85%* Calculation: 85 ÷ 90 ≈ 0.944* You earn 94.4% of full OAP* Substitute Spend Adjustment = 5.6% reductionThis adjustment caps at 25% reduction.The Single Adjustment RuleCritical detail: CMS applies only the larger of the two adjustments during semi-annual reconciliation. If you face a 20% Clinical Outcome Adjustment and a 5% Substitute Spend Adjustment, you lose 20% total—not 25%. This prevents compounding penalties while maintaining accountability across both dimensions.The reconciliation occurs twice per year, assessing all patients whose 12-month care periods ended during the trailing 6-month window. CMS nets the reduction against your withheld payments. Any remaining balance owed to you gets released through the next quarterly payment. If additional recovery is needed beyond withheld amounts, CMS may offset against future payments or initiate standard Medicare overpayment recovery procedures.Enjoy timely and insightful deep dives on health policy and clinical AI? Subscribe to Techy Surgeon for more!Payment Rates and Rural AdjustmentsCMS hasn’t published specific dollar amounts yet—those come in 2026 before the April application deadline. What we know: payment rates vary by track, with higher rates for Initial Periods (first 12 months or when baseline measures aren’t at target) and lower rates for Follow-On Periods (maintenance care for established patients or those starting at target).For eCKM and CKM tracks specifically, the payment includes expected device cost for a cellular network-connected blood pressure cuff used for condition management and outcome reporting. Rural patients in these tracks receive a fixed add-on payment to account for higher distribution costs. Other tracks don’t include device costs and therefore have no rural adjustment.Rates may update annually based on Medicare Physician Fee Schedule (PFS) updates, regulatory changes, or an efficiency adjustment tied to the Medicare Economic Index (MEI) productivity adjustment percentage.Multi-Track DiscountWhen a beneficiary enrolls in multiple tracks with the same participant, CMS applies a payment discount to the total OAP amount. This reflects administrative and operational efficiencies from delivering integrated care—shared intake, unified patient records, coordinated communication with PCPs. The specific discount percentage will be published with payment rates in 2026.Cost-Sharing PolicyParticipants must establish a uniform cost-sharing policy per track: either collect standard Medicare Part B cost-sharing or forego collection under the CMS-sponsored model patient incentive safe harbor (42 CFR § 1001.952(ii)(2)). This policy must apply consistently to all aligned beneficiaries within each track and can only be revised with CMS approval on a go-forward basis.If you elect to collect cost-sharing, you must clearly disclose expected beneficiary payment amounts before enrollment. Most participants are expected to waive cost-sharing as a beneficiary engagement strategy, but the choice remains at the organizational level.What the Payment EnablesThis structure creates financial predictability—recurring revenue based on panel size—while tying full payment to measurable outcomes. The quarterly disbursement provides operational cash flow. The split threshold approach (50% for outcomes, 90% for substitute spend) balances accessibility for new participants against long-term accountability for results and coordinated care.Unlike shared savings models where attribution complexity makes condition-specific savings hard to isolate, ACCESS directly measures clinical outcomes and duplicative spending. Unlike fee-for-service where reimbursement requires specific billing codes, ACCESS gives flexibility in care delivery modalities as long as outcomes are achieved.The model explicitly accommodates technology-enabled care—virtual visits, remote monitoring, asynchronous engagement, FDA-cleared digital therapeutics—under appropriate clinical oversight. Payment doesn’t depend on time-based billing or in-person visit requirements. It depends on moving the clinical measures.Next module, we’ll cover the six-stage beneficiary journey from discovery through reconciliation. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

The ACCESS Model, ExplainedPart 1 of a 12-part operator series on the CMS ACCESS ModelThere has been no shortage of commentary since CMS released the 61-page Request for Applications for the ACCESS Model.The early reactions were predictable. Medicare finally opens to digital health. A direct threat to fee-for-service. A gold rush for chronic disease management.I shared my initial impressions shortly after the RFA dropped and created a landing page at TechyPolicy.com for readers who wanted to track the model more closely as details emerged.Those headlines are not wrong. But they are incomplete.What CMS has not yet released are the numbers that usually dominate these conversations. We still do not know the per-beneficiary outcome-aligned payment amounts. CMS has said those figures will come in 2026. What the agency has provided instead is something more revealing: a detailed architecture of participation that shows who can realistically enter the model, what participation requires, and where the economic tensions sit.That architecture will matter as much as the payment itself.The short video above walks through how I am thinking about ACCESS at a structural level. Not whether the model is ambitious or disruptive, but whether it can function as designed once real organizations with real incentives attempt to operate inside it.What the explainer covers* Why ACCESS is being interpreted as a direct Medicare entry point for digital health companies* The Part B enrollment requirement that separates interest from eligibility* The fee-for-service exclusion rule and why it reshapes health system participation* The coordination problem created when patients receive care across overlapping payment modelsThe central tensionACCESS introduces something genuinely new for Medicare. Chronic care delivered through technology-enabled models, paid on outcomes rather than visits.But the constraint that defines the model is easy to miss.Buried in the RFA is a clear rule: ACCESS participants and their affiliates cannot bill Medicare fee-for-service for aligned beneficiaries during active care periods.For organizations without existing fee-for-service revenue, this is not an obstacle. It is the design feature that makes the model viable. Outcome-aligned payment replaces visit-based billing rather than competing with it.For health systems whose economics still depend heavily on fee-for-service billing, the math looks very different. Participation does not require regulatory exclusion. It creates financial exclusion on its own.The likely result is selective participation rather than universal adoption. Patients may be enrolled with one organization for ACCESS conditions while continuing to receive primary care or specialty services from others operating under entirely different incentives.In that environment, coordination is not a secondary consideration. It is the model.Subscribe to Techy Surgeon to follow the ACCESS Model Series.What we still do not knowSeveral core elements of ACCESS remain unresolved.CMS has not yet published per-beneficiary payment amounts by clinical track. The details of the risk adjustment methodology remain unclear. It is also not yet known how outcome attainment thresholds will evolve beyond the first performance year.What we do know are the timelines.Applications open in January 2026. The initial deadline is April 1. The first cohort launches July 5, 2026.That leaves organizations roughly a year to decide whether ACCESS fits their strategy, and if so, how to structure participation without creating unintended exposure.About this seriesThis post is the first in a 12-part series examining the ACCESS Model from an operator’s perspective.Each installment focuses on one specific mechanic in the RFA and the practical implications CMS does not always state explicitly. The goal is not advocacy for or against the model, but clarity around how it functions once contracts are signed and patients are enrolled.New posts will publish on TechyPolicy.com throughout the series.If you want a brief downloadable overview of the model, download the PDF here. I’ll have more operator-specific and mechanics-specific resources and assets throughout the course.Follow the 12-part ACCESS seriesI’ll be releasing an article a day and explaining topics. Paid subscribers receive all posts, diagrams, and decision frameworks.A note on coordination infrastructureFor transparency, I am co-founder and CEO of RevelAi Health. We build care coordination and patient engagement infrastructure designed for value-based and outcome-aligned models.ACCESS will require reliable outcome measurement, sustained patient engagement, and real communication across organizational boundaries. Those capabilities will determine whether coordination succeeds or breaks down.If you are evaluating ACCESS and want to talk through the operational requirements, I am open to those conversations.📧 [email protected]🌐 www.revelaihealth.comSubscribe to Techy Surgeon. Independent analysis of CMS models, clinical AI, and care delivery design. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Hey everyone,I know 2025 was tough. New regulations dropping without warning. AI tools promising everything and explaining nothing. Value-based care models that sounded great until you tried to implement them.So I’m doing an end of year discount run. Some of the items on the agenda for 2026.* Exclusive interviews with leaders from Doximity, OpenEvidence, and Notex* AI tool guidance to help you cut through the marketing noise and pick what actually works for your workflow* Value-based care playbooks for executives wrestling with ACCESS, BPCI-A, and CMS quality programs* The ACCESS Model breakdown you’ve been asking for—what it means, who can participate, and how to prepare* Behind-the-scenes Techy Flows showing how I actually create this content while managing clinical work and familyAnd because it’s the holidays…🎁 CHRISTMAS SPECIALThrough New Year’s Day:Annual subscriptions: $130 → $100/year (save $30, lock in forever)Subscribe here: https://techysurgeon.substack.com/Christmas2025Here’s the deal:* This price holds as long as you stay subscribed (never increases)* Full access to everything: interviews, playbooks, tutorials, monthly Q&As* If you’re already a paid subscriber, I’m retroactively applying this discount to your next renewal—you don’t need to do anythingWhy I’m doing this: I just got a nice bump in paid subscribers, and frankly, you all deserve this. You show up, you engage, you ask the hard questions that make this newsletter better. Consider this my thank you.What You GetStarting January 2026:* Deep dives on the ACCESS Model (participants, exclusions, coordination challenges)* Clinical AI product reviews (the real pros/cons, not the marketing pitch)* Policy-to-playbook conversions (CMS regulations → actual implementation guides)* AI building tutorials (how I make landing pages, videos, and content with AI)* Executive interviews with the people actually building clinical AI* Monthly Q&A sessions where we tackle your specific questionsFree subscribers still get weekly policy analysis and clinical AI insights. But paid subscribers get the full playbooks, templates, and direct access.No coal in YOUR stocking this year.Lock in $100/year forever →Offer expires Thursday, January 1st at midnight EST.Let’s make 2026 the year we actually figure this stuff out together.— ChristianP.S. — If you’re already subscribed at the monthly rate and want to switch to annual to lock in this price forever, just reply to this email and I’ll help you swap over. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.Osteoporosis is one of those conditions that stays invisible until it suddenly isn’t. People often feel perfectly fine, their daily routines unchanged, and then a fall from standing height leads to a broken hip or spine. The diagnosis arrives way after the damage is done.My goal with this explainer is to help shift the conversation earlier so more people understand the condition before the first fracture occurs. I also want to address a fear I hear often: the risk of osteonecrosis of the jaw (ONJ) from osteoporosis medications. ONJ is real, but the risk at standard osteoporosis doses is much smaller than most people think, especially compared with the risk of untreated osteoporosis.This is not personal medical advice, but it reflects how I think about the evidence as a surgeon who treats fragility fractures regularly.Osteoporosis Is a Fracture DiseaseOsteoporosis occurs when the internal architecture of bone becomes weaker over time. You cannot feel this happening. There is no early warning signal. The first sign is often a fracture.Hip fractures are particularly serious. Large studies estimate that about one in five older adults who experience a hip fracture die within the following year. Many survivors lose mobility, struggle with activities of daily living, or require long-term support.¹–³This is why clinicians worry about osteoporosis. We are not treating a number on a bone density report. We are trying to prevent the events that follow a hip or spine fracture: hospitalization, surgery, prolonged recovery, loss of independence, and increased mortality.How Treatment HelpsOsteoporosis treatments do more than slightly improve bone density scores. They lower the chance of having a future fracture.Medications like bisphosphonates and denosumab have been shown in large trials to reduce vertebral fractures by roughly 40 to 70 percent and hip fractures by about 40 to 50 percent in people at higher risk.⁴–⁷Nothing eliminates risk completely, but these reductions matter. They are the difference between falling and bruising your hip versus falling and breaking it.Understanding the ONJ ConcernOsteonecrosis of the jaw is a complication in which part of the jawbone does not heal properly. It can be painful and concerning. The fear often comes from mixing together two very different dosing scenarios.High-dose intravenous bisphosphonates used for cancer carry a much higher risk, sometimes 1 to 10 percent depending on treatment duration.⁸–⁹ These numbers are frequently cited online without specifying that they apply to cancer therapy, not osteoporosis treatment.For the doses used to treat osteoporosis, the data tell a different story. Studies estimate the risk of ONJ somewhere between 1 and 10 cases per 10,000 people receiving treatment.¹⁰–¹³ That means at least 9,990 out of 10,000 people taking these medications for osteoporosis do not develop ONJ.The key point is scale. A hip fracture is far more common than ONJ at osteoporosis dosing, and the consequences of hip fractures are significantly more severe.How I Frame the DecisionWhen I talk with patients, I try to simplify the comparison:* Fragility fractures are common in osteoporosis, especially hip and spine fractures. They are associated with loss of independence and increased mortality.* ONJ at osteoporosis treatment doses is rare, and we have ways to reduce the risk further, such as coordinating around major dental work and maintaining good oral hygiene.* Medications offer substantial reductions in fracture risk for patients who are already at high risk.So the real question becomes: Given a patient’s fracture risk, do the benefits outweigh the small risk of ONJ or other side effects? For many high-risk individuals, the answer is yes. For others at lower risk, the conversation may lead to delaying medication or choosing a different strategy.This decision is individualized, but it should be anchored in accurate information.What You Can Do NowRegardless of whether medication is appropriate for you, a few steps always help:* Understand your fracture risk.* Get a bone density test if you meet criteria.* Strength train and do weight-bearing activities to stimulate bone.* Ensure adequate calcium and vitamin D intake.* Discuss treatment options if your risk is elevated.If you watched the accompanying animated explainer, I hope it helped put the ONJ conversation in perspective. Osteoporosis treatment is not about chasing a lab value. It is about preventing the fracture that can alter your life.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.References* Brauer CA, Coca-Perraillon M, Cutler DM, Rosen AB. Incidence and mortality of hip fractures in the United States. JAMA. 2009;302(14):1573-1579.* Schnell S, Friedman SM, Mendelson DA, et al. The 1-year mortality of patients treated in a hip fracture program for elders. Osteoporos Int. 2010;21(11):2007-2014.* Guo J, et al. A nomogram for one-year risk of death after hip fracture. Front Med. 2025;12:123456.* Byun JH, et al. The efficacy of bisphosphonates for prevention of osteoporotic fracture. Endocrinol Metab (Seoul). 2017;32(4):437-445.* Wu CH, et al. Pharmacologic intervention for prevention of fractures in osteoporosis: a systematic review and network meta-analysis. Bone. 2020;133:115247.* Deardorff WJ, et al. Time to benefit of bisphosphonate therapy for the prevention of nonvertebral fractures. JAMA Intern Med. 2022;182(1):33-41.* Goodman CW, et al. Reconsidering the benefits of osteoporosis treatment. Am J Med. 2024;137(5):555-563.* Khosla S, et al. Bisphosphonate-associated osteonecrosis of the jaw: report of a task force of the American Society for Bone and Mineral Research. J Bone Miner Res. 2007;22(10):1479-1491.* American Dental Association. Osteoporosis Medications and Osteonecrosis of the Jaw. ADA; 2023.* Huang YF, et al. Impact of bisphosphonate-related osteonecrosis of the jaw on osteoporotic patients: a nationwide cohort study. PLoS One. 2015;10(7):e0120756.* Tennis P, et al. Incidence of osteonecrosis of the jaw among users of bisphosphonates. J Rheumatol. 2012;39(6):996-1003.* Liu FC, et al. Risk comparison of osteonecrosis of the jaw in osteoporotic patients treated with bisphosphonates vs denosumab. J Formos Med Assoc. 2023;122(5):340-349.* Royal Osteoporosis Society. Osteonecrosis of the jaw (ONJ). ROS; 2023. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.Parenting a toddler while juggling surgery, a startup, and a six-month-old is…a lot. There were nights when my son was screaming over bedtime or dinner and I found myself sneaking into the kitchen, typing frantic prompts into ChatGPT: “He’s jealous of the baby and throwing toys—what do I say right now?”I was doing it so often that I realized I basically had an invisible parenting coach on call—if only I didn’t have to re-prompt from scratch every time.So I pulled together the best of those AI prompts and responses, layered in evidence-based parenting strategies, and turned it into a structured guide for “big feelings” moments. From there, I used Gemini to draft the first pass of a site, then built a fully functional web app in Lovable (no coding) and wired it up to an OpenAI-powered chat coach.That’s how BigFeelingsAI.com was born: a free parenting copilot with* ready-to-use scripts for ~12 common meltdown scenarios* a Calm Nights section that builds a bedtime routine for you* a “I’m losing it” button that walks you through a 10-breath reset with ambient ocean sounds* and a built-in chat coach that gives step-by-step, evidence-based responses in the moment.It’s not a replacement for real parenting, but it is the tool I wish I had the first time my toddler melted down over the blue cup versus the red one. This is just a fun hobby project I made for myself, but I’m free, and I’m sharing it with anyone who needs that on-demand advice in a way that is actionable, easy, and evidence-based. If nothing else, check out the mini meditation “parent pulse” for when you need a “Time Out” yourself, in the right corner!If you want the full behind-the-scenes on how I built it (prompts, tools, and all), I wrote it up here: “How I Use AI to Raise My Toddler”.Try the site, use the chat coach next time you’re in the trenches, and let me know what works—or what you’d want it to handle next! If you are a paid subscriber and have other ideas for vibe coded landing pages, functional apps, clinical workflows and more, but aren’t sure where to start, send me a message and I’d love to work through it with you!Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

This is a free preview of a paid episode. To hear more, visit techysurgeon.substack.comI hit “record” on the very first Techy Surgeon podcast with all the swagger of a resident doing their first solo case: a little shaky, very excited, ready to learn in public. My guest, Dr. Alex Crespo—orthopaedic trauma surgeon and Director of Orthopaedic Research at Endeavor Health in Chicago—showed up with something rarer than a perfect fluoroscopy shot: a simple, repeatable way to move hip‑fracture care from SNF by default to home by design.Dr. Alex Crespo and I chatted about hip fracture care and whether or not AI is written in the stars of value based care.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.If you live anywhere near value‑based care, you already feel the pressure building. CMS’s Transforming Episode Accountability Model (TEAM) looms, mandatory in 2026, and hip fracture care sits right in the crosshairs. The episode math is fairly unforgiving: outcomes divided by cost. SNF days dominate the denominator. We can argue for better risk adjustment (we should), but the systems that win will do something far more basic: build a reliable path to home for the patients who can safely make it, and keep a clean line of sight on those who can’t.“Road Home,” explained the way a family needs to hear itCrespo’s team calls their approach Road Home. At first glance, it’s almost laughably analog: a large laminated poster next to the TV in every hip‑fracture room—think Candyland track, not clinical dashboard. Day 0, we align on intent (“Home is best if safe.”). Day 1, the goal is ~10 steps—and we set expectations that this is the worst day. Day 2, ~20 steps. Day 3, ~40. Day 4, 75–100 feet and a stair or two, and—if the whole picture supports it—home.That’s it. Not an app. Not a sensor. A shared, visible plan that everyone references: surgeon, nurses, PT, social work, PA, patient, family. The genius isn’t the numbers; it’s the psychology. Hip fractures are an emotional injury as much as a mechanical one. A scheduled total hip gives patients months to prepare; a fragility fracture ambushes them before breakfast. The Road Home poster lowers the temperature in the room. It makes progress legible. Families can point to something and say, “We’re getting there.”A simple intervention and visual help guide patients from injury to home post hip fracture. Selection without cynicism“Send everyone home” isn’t the point. The team starts with a simple, human question—“What do you want for your care plan?”—and then does the mundane, important work: pre‑injury function, stairs at home, assistive devices, local support. By Day 2, they decide which track a patient is truly on. If SNF is the right answer, they move decisively rather than lingering in wishful purgatory on the ward. If home is possible, they coach like it matters—because it does.What changed when they did this for a yearOver the first 12 months, across >600 hip‑fracture patients, their SNF utilization dropped from ~78% to ~52%. Length of stay didn’t bump (my biggest fear going in), which tells you the early triage cut out the “maybe tomorrow” drift. The surprise win: HCAHPS (Hospital Consumer Assessment of Healthcare Providers and Systems survey) shot up, especially in physician communication, to the best the trauma service (and even the system) has seen. Families felt like the team was aligned. Patients felt coached, not bounced.On returns: their team sees far fewer revisits from home than from SNF (on the rough order of ~3% vs ~30%, ED revisits included). That stat is more nuanced than you would thin, SNF patients are sicker by definition, but it punctures the common fear that “home equals dangerous.” Risk follows the patient, not the home discharge destination. What we control is the plan.The uncomfortable truth about tech (from an AI guy)I run an AI company. I love sensors, stratification, and agentic workflows. But here’s the quiet part we have to say out loud: the first unlock wasn’t a fancy model…it was a poster. Consistent communication beats clever code at the bedside, especially in the first 96 hours.That doesn’t mean tech has no role. It means sequence matters:* First, prove the loop with paper and change management: shared goals, daily coaching, early triage.* Then, digitize what already works: give patients and caregivers a mobile view of that same four‑day curve; let staff tap in steps and stairs in 10 seconds, not two minutes.* Next, watch the places we lose people: a lightweight SNF surveillance cadence (Day 7/14/21) that tracks barrier codes—pain control, placement, DME, therapy tolerance, social—and gives the team a reason to act today, not next week.* Finally, layer smart help: guideline surfacing for osteoporosis/secondary prevention, fall‑risk nudges, an after‑hours agent that answers FAQs with empathy and evidence and escalates when it should.None of that requires a moonshot. It requires discipline, UI that respects clinical time, and a willingness to brand the pathway so people actually remember it. Road Home is memorable. That matters.Why this works (and why it is repeatable, if not scalable)What Crespo’s team built is portable because it attacks a universal and undertreated problem in hip fracture: fear and uncertainty. Patients are frightened of falling again. Families picture staircases like cliffs. Staff want to help, but the plan can feel abstract, especially when everyone is rotating. The poster makes the plan tangigfble. The four‑day arc makes the struggle predictable. The early Day‑2 fork keeps the service honest. And—this is key—it gives administrators something to invest in that doesn’t require a procurement saga.If you’re reading this in a hospital that’s bundle‑exposed, you don’t need permission to start. You need a printer, a name, and a daily huddle. When the data begin to move, that’s the time to bring in a product team and wire up the inexpensive pieces of automation that turn a local habit into a system behavior.From chalkboard to electronic chart? Sometimes just tallying up the wins is better than trying to force a technology that is superfluous down the team’s throatsMonday‑morning versionCall it Road Home. Put the poster up. Tell patients Day 1 will be rough and that it’s part of the plan. Decide by Day 2 who’s going home and who isn’t. Treat families as members of the care team. Track a few outcomes you care about: SNF rate, SNF days, HCAHPS communication, ED/return. If you have a quality analyst, ask for a quick‑and‑dirty run chart. Celebrate the slope, not a single number.Listen inIn the episode, Crespo walks through how the idea grew out of NYU’s early arthroplasty value work, what changed in the room on Day 1, why the family matters as much as the patient, and where technology should and shouldn’t show up. It’s candid, practical, and, I think, a blueprint for the TEAM era.For subscribers: the money, the model, and the SNF‑surveillance spec 🔒Below the line, I break down a rounded, site‑agnostic ROI frame you can adapt for your internal deck: how to estimate contribution from avoided SNF days, what staffing it actually takes to run Road Home, and how to stand up a minimal SNF Day‑7/14/21 tracking loop with clear owners and barrier codes. I’m also including a one‑page “build‑with” brief for digital partners who want to help without boiling the ocean.Figures below are intentionally aggregated, with institution‑specific numbers generalized to protect local details and avoid overfitting.Disclosures & disclaimers: The views here are mine and Dr. Crespo’s, not our employers’. This is not medical advice. Program results reflect one system’s experience over ~600 patients and should be validated locally with your case mix, staffing, and market rates.If your team is spinning up a Road Home‑style pathway, I’d love to hear what you’re seeing. Hit reply, or—if you want a deeper dive—become a paid subscriber and grab the implementation packet.If you'd rather interact with an infographic landing page that I made rather than read through the guide below, click the following gemini link It comes complete with the ability to generate talking points for the program and some really great visualizations of the data from the program.

In a year thick with demos, headlines, and understandable caution, it’s hard to separate signal from noise in health AI. Duke is taking a different approach: pair scientific rigor with real operational build, and invite the people who will live with these systems to co‑design them. On October 8–9 at the North Carolina Biotechnology Center (RTP), the Duke Summit on AI for Health Innovation—hosted by Duke AI Health, the Center for Computational & Digital Health Innovation, and the Duke Clinical Research Institute (DCRI)—will convene that conversation in the open.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.I’ll be speaking about how AI-driven workflows close care gaps and improve patient experience, with a focus on agentic systems that escalate intelligently, create auditable loop closure, and support value-based care without adding burden. I’m honored to share the stage with leaders who collectively shape how a health system evaluates, deploys, and governs technology: Richard Shannon, MD (Chief Medical & Quality Officer), Matthew Roman, MMCi (Chief Digital Strategy Officer), and Eric Poon, MD, MPH (Chief Health Information Officer). If you’ve ever wondered how decisions actually get made inside a health system—where safety, equity, and ROI meet—this is that room.What sets this Summit apart is its span. From evaluation science and governance (Michael Pencina, PhD) to enterprise informatics and EHR modernization (Jeffrey Ferranti, MD, MS), from biomedical engineering and data fusion (Jessilyn Dunn, PhD; Amanda Randles, PhD) to computational and clinical machine learning (Monica Agrawal, PhD; Ricardo Henao, PhD), and from surgical AI and clinical translation (Ozanan Meireles, MD; Allan Kirk, MD, PhD) to emergency medicine and global implementation (Catherine Staton, MD, MSc) and biostatistics at scale (Chris Lindsell, PhD; DCRI). We’ll also explore the engineering and commercialization bridge with Jerome Lynch, PhD, Steve McClelland, and Jonathan Owens, MBA, and hear from industry voices such as Kurt Schmidt (NVIDIA)—a reminder that trustworthy AI needs both clinical depth and robust infrastructure. Guiding the ecosystem view is Shelley Rusincovitch, MMCi (Duke AI Health), whose work sits at the intersection of research, operations, and workforce development.The through line is responsible AI with measurable outcomes: surgical innovation that helps in the OR and on the ward; computational medicine that moves beyond novelty to reproducibility; human‑centered design that earns trust; and breakout work that turns into artifacts you can use Monday. Duke is signaling that leadership in this space isn’t about hype or paralysis—it’s about building together with accountability.Who should come? Clinicians and care team leaders trying to reduce avoidable utilization; engineers and data scientists seeking clinical context; operators and product leaders translating pilots into programs; Investors looking to understand how innovation translates to care transformation; students looking for a front-row seat to how health systems decide; and anyone interested in how adoption really happens.If that’s you, I hope you’ll join us in Durham. Registration (with nonprofit and student discounts) is available now. Come for the signal—and help us build what lasts.Register Here for the Duke Summit on AI for Health InnovationTechy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

If you follow me on Instagram, you’ve probably seen me share the little storybooks I’ve been making for my toddler. I started by using GPT image tools—super fun, but a little clunky. My son loved them, but honestly, I was spending way too much time stitching pages together.Turns out, there’s now a much easier way. Google quietly rolled out a Storybook Gem inside Gemini, and it’s almost absurdly simple to use.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.How to Do It* Head to Google GeminiJust log in with your regular Gmail address.* Open the Expanded MenuIn the sidebar, scroll down until you see Gems—these are essentially pre-built mini-agents for specific use cases (you can create your own by the way!).* Select “Storybook”Google has made this one available for everyone. Once you click in, you’ll see a panel of starter prompts (e.g., “Write a bedtime adventure” or “Create a story about a dinosaur and a rocket ship”).* Pick a Prompt (or Write Your Own)Click one of the options, and Gemini will generate both the text and illustrations. You can quickly edit the story, swap styles, or download the finished book.That’s it—you now have a personalized, AI-generated storybook in just a few clicks.Why I Like It* It’s fast. No juggling separate image and text tools.* It’s customizable. You can add your child’s name or favorite animal and watch the story adapt.* And most importantly: my toddler lights up every time we flip through a new book.If you’re curious, open Gemini today and try the Storybook Gem. It’s free, playful, and surprisingly polished. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

The United States spends roughly $380 billion each year on musculoskeletal (MSK) conditions—more than any other diagnostic category. Meanwhile, outpatient shifts have splintered workflows, and care gaps are widening in the very populations that most need coordinated care.That’s why we’re convening the Downeast Digital Summit, September 18–20 at the Westin in Charlotte, NC: an invite-only working meeting for people who want to do the work, not just talk about it. Edmondo Robinson, MD, MBA, MS, FACP—Founder & CEO of Downeast Digital and former SVP/Chief Digital Officer at Moffitt Cancer Center (previously Chief Transformation Officer at ChristianaCare)—is co-organizing the summit with us. If you’re serious about using AI and modern operating models to move the needle on population health in orthopedics, this room is for you.Why nowCMS has finalized the Transforming Episode Accountability Model (TEAM)—a mandatory, five-year episode model launching January 1, 2026. TEAM will reward organized, high-quality, cost-conscious episode care and penalize fragmentation. Translation: the window to redesign perioperative pathways, data flows, and patient engagement at scale is now. What makes this differentThis isn’t a trade-show floor. It’s a build space. We’ll mix short, provocative talks with table-top design sprints—small, cross-functional groups that leave with 90-day pilots, shared metrics, and lightweight governance plans. Think:* Pre-op to 90-day episode playbooks aligned to TEAM quality measures presented by Medisolv* AI-enabled intake and triage that actually reduces cancellations and post-op ED visits by RevelAi Health* PROMs and patient education that patients complete (and clinicians trust).* Closing Care Gaps by design: tracking outcomes by neighborhood deprivation and social driver capture with interventions that close gaps—not just report them.We’ll publish a short post-summit brief summarizing the playbooks, pilots, and partnership commitments that emerge—so the work compounds across institutions.Who’s in the roomWe’re bringing together policy thinkers, health-tech builders, payers, surgeons, and data leaders who have already shipped meaningful work—and want to partner up for the next phase. Among the leaders we’ll be highlighting this week:* Annunziato (“Ned”) Amendola, MD — 2025–2026 President of the American Academy of Orthopaedic Surgeons (AAOS).* Travis Zack, MD, PhD — UCSF oncologist and AI researcher; clinical leader at OpenEvidence. * Shruti Shah — Partner, Symphonic Capital. symphoniccapital.com* Benjamin Schwartz, MD, MBA — Fellow Orthopedic Surgeon, Innovator and Incoming SVP of Care Services & Strategy at Commons Clinic. By the way, check out his substack!* Richard “Chad” Mather III, MD, MBA — Chief Medical Officer, Optum Specialty Care. * Michael Pencina, PhD — Chief Data Scientist, Duke Health; Vice Dean for Data Science, Duke School of Medicine. Duke Health* Erin Heilman, CPHQ and Kristen Beatson, BSN — Quality reporting and value-based care leaders at Medisolv.…and a growing roster of founders, clinicians, and health system operators who are ready to get specific and ship.What you’ll leave with* A TEAM-ready episode blueprint you can adapt at your org.* A shortlist of AI use cases (with guardrails) that cut cancellations, reduce ED revisits, and lift PROMs completion—plus how to measure impact.* New partners—from quality reporting to AI governance, and care coordination—aligned on deliverables, timelines, and shared metrics.* A path to population health grounded reporting, using pragmatic proxies that translate to operational action.Who should applyAI/health-tech founders, quality leaders, researchers, entrepreneurs, surgeons and ASC leaders, service line and population health execs, payer medical directors, analytics and operations teams—anyone accountable for results in MSK care who wants to collaborate.Join usThe Downeast Digital Summit runs September 18–20 at the Westin Charlotte. We have a limited number of additional invitations available for the Techy Surgeon community. Apply to attend here MSK DownEast Digital Summit or message me directly if your team is ready to co-design a use case at the summit. Techy Surgeon Subscribers can also message me directly for limited discounted registration spots.Let’s stop admiring the problem. Let’s build the playbooks, pilot them together, and publish what works—so we can spend the next year scaling solutions that patients, clinicians, and payers can all say yes to.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

I recently got access to ChatGPT’s new Study Mode, a feature that offers step-by-step guidance instead of just giving quick answers. As a techie surgeon and educator, I was eager to see how this could help me learn a topic interactively. To test it out, I picked a concept relevant to healthcare that I don’t know well off the top of my head: moral hazard in healthcare. Here’s how my learning session went and some takeaways on using Study Mode—all in first-person, informal style.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber!Learning a Healthcare Concept with ChatGPT Study ModeFiring up Study Mode (via the “Study and learn” tool in ChatGPT), I told it I wanted to better understand moral hazard in healthcare. Rather than immediately lecturing, ChatGPT Study Mode started by asking me a few questions about my goals and current understanding, almost like a tutor sizing up a new student. This is is similar to how the tool functions when you use the Deep Research Function. I gave a somewhat rambling answer (honestly, I knew only vague bits about moral hazard from memory). Based on that, it tailored the explanation to my level. I then asked it for “a concise refresher” on the topic. It responded with a clear and compact summary of moral hazard – perfect for getting me oriented.At this point, I switched to voice mode. Using voice, I dug deeper: “Can you give me a simple example of moral hazard in healthcare that I could use to teach someone else?” ChatGPT obliged with a relatable scenario. Chat: “Imagine you have a health insurance plan that covers all prescription costs with no copay. When you had a small copay before, you might have chosen a generic drug to save money. But now, with zero out-of-pocket cost, you might always opt for the expensive brand-name drug – even if the cheaper generic works just as well. Why not, if it’s free for you?” The AI model explained that this increased usage of pricier options, without added health benefit, is a classic case of moral hazard. It gave another simple example: patients scheduling more doctor visits for minor issues when there’s no copay, whereas they might have skipped unnecessary visits if they had to pay a little each time.I was impressed by how clearly the AI broke it down. In plain terms, moral hazard in health insurance means that when people aren’t paying the full cost, they tend to use more healthcare (or more expensive healthcare) than they otherwise would. This doesn’t mean patients are acting badly; it’s a natural response to the incentives. As one economics article put it, generous insurance makes healthcare feel “too cheap,” so people consume “too much of it” (like picking the brand-name drug every time if it costs you nothing). The result can be overuse of services or higher costs without better outcomes, which is exactly what insurers worry about.Next, I asked: “How do insurers use the idea of moral hazard when designing networks or pricing?” This steered the conversation into how insurance companies respond to this behavior. ChatGPT explained that insurers build in “friction” for the consumer – not to deny care, but to make you think twice before overusing low-value care. In practice, this means things like deductibles, copays, and coinsurance (forms of cost-sharing) so you have some skin in the game. For example, a deductible is the amount you must pay out-of-pocket before insurance kicks in – essentially a built-in threshold that ensures you bear initial costs. If you have a $500 deductible, you’ll likely be more selective about running to the ER for a sniffle, because you pay that first $500 of care yourself. Insurers also create tiered provider networks, offering lower copays or better coverage if you use “preferred” doctors or hospitals that provide cost-effective care. All these tactics are about balancing access and preventing overuse: insurance should cover you when you truly need care, but not make things so free and easy that the system gets overloaded with unnecessary expenses.I echoed back to ChatGPT my understanding: “So basically, they add a bit of friction for the patient to prevent overutilization – making you pause and decide if that brand-name drug or extra visit is really necessary.” ChatGPT confirmed: “Exactly. It’s about adding just enough friction so patients (and even providers) make mindful decisions about using healthcare resources, without making care inaccessible.” I also had it clarify a term – “Can you succinctly define deductible for someone who’s never heard the word?” The answer was packaged nicely and clearly: a deductible is the amount you pay out-of-pocket for healthcare services before your insurance starts covering costs. Once you’ve paid up to that amount, the insurance begins to pay its share. Simple enough!Throughout this Q&A, Study Mode felt like a cooperative tutor. It prompted me, answered follow-ups, and adjusted explanations based on my questions. This interactive style aligns with Study Mode’s design of guiding learners with Socratic prompts and tailored responses, rather than just handing over an answer key.From Refresher to Lesson PlanAfter this back-and-forth, I wanted to see if ChatGPT could help me turn what I learned into a lesson plan – something I might use to teach others about moral hazard. I asked it to create a “very succinct, easy-to-understand lesson plan for someone learning about moral hazard,” including the key points we had discussed (definitions, how insurers address moral hazard, practical examples) and even a couple of practice problems.It delivered a neat little outline. To summarize that lesson plan, it included:* Definition & Concept: A plain-language definition of moral hazard in healthcare – essentially that people tend to overuse services when they aren’t paying the full cost themselves.* Key Mechanisms: An explanation of how insurers mitigate moral hazard. This covered cost-sharing tools like deductibles, copays, and coinsurance that ensure patients have some cost responsibility, as well as network incentives (using preferred providers, etc.) to encourage cost-effective choices.* Practical Examples: Relatable scenarios (like our prescription drug example and extra doctor visits example) illustrating moral hazard in action. These examples show why insurers put certain rules in place.* Quick Quiz Questions: One or two simple problems for the learner, to reinforce understanding. For instance, one question asked the reader to consider a scenario (e.g. “If Alice’s insurance removes all copays for doctor visits, how might her behavior change, and why is that an example of moral hazard?”). Another might present a cost-sharing scenario (like “Bob has a $1000 deductible – how could that affect his decision to go to the ER for minor issues?”) so the student can apply the concepts.(I’ll include the full lesson plan output below for those interested.) Overall, I found this really useful. As someone who might one day teach this topic to colleagues or students, having an outline generated on the fly – which I can then expand on or fact-check – is a huge time saver. It’s like getting a first draft of a teaching prep done in seconds.Of course, since I’m not an expert on moral hazard myself, I’d still verify details and maybe add nuance (for example, moral hazard isn’t inherently “immoral” – it’s an economics term, and there are also concepts like adverse selection in insurance, though that’s another story). But as a starting point, ChatGPT’s Study Mode provided a solid foundation of knowledge and a structure to pass that knowledge on.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.Study Mode Impressions – A Tutor in Your Pocket?Now that I’ve tried it, how does Study Mode feel compared to a normal ChatGPT session? Study mode feels more purposefully structured to be a personal tutor or study buddy. The mode actively engaged me with questions and checkpoints instead of just responding to my prompts with a long answer. This kept me thinking and involved in the learning process. I can see why OpenAI built it this way – research suggests that active engagement (like answering questions) helps with deeper understanding, and Study Mode is basically ChatGPT nudging you to think, not just absorb answers.One thing I noticed: the difference between Study Mode and a standard ChatGPT conversation can be subtle at times. In a regular chat, I could still ask follow-up questions and get explanations. But Study Mode formalizes that guiding process. For example, right from the start it asked what I knew and what I aimed to learn, calibrating its teaching style. It also structured information into digestible pieces and even offered to quiz me. In a normal session, ChatGPT might not take that initiative unless prompted. Study Mode essentially adds a layer of “teacher intelligence” on top of the AI’s knowledge base. In my case, that meant a more curated learning journey: starting with basics, then drilling into examples, then zooming out to applications – finally culminating in a lesson plan with review questions. Not bad for an AI I just met!I’m still exploring this feature and admittedly curious how it might differ with other subjects or longer study sessions. It’s worth noting that you can toggle Study Mode on or off anytime, even mid-conversation, which is nice if you want to switch back to just getting a straightforward answer. In my moral hazard session, I stayed in study mode and it was an organic conversation flow. The voice interaction capability also made it feel very natural – I was literally talking through a concept with an AI tutor. For an educator like me, that spark of interactivity is exciting.Other Use Cases for ChatGPT’s Study ModeMy experiment was in a healthcare finance concept, but I’m already imagining other ways to use Study Mode day-to-day. Some ideas for clinicians, educators, and anyone curious:* Continuing Medical Education: A clinician could use Study Mode to break down a new medical research paper or guideline. Imagine “studying” a new cancer treatment protocol with ChatGPT prompting you to explain key points, ensuring you truly grasp it rather than just skimming an article.* Academic Study & Exam Prep: Students (or their teachers) might use it for exam reviews. For example, a nursing student could review physiology or pharmacology by having ChatGPT quiz them in Study Mode, or work through practice case studies step by step. It’s like guided 24/7 office hours.* Learning Technical Skills: If you’re learning to code or understand a complex tech concept, Study Mode can walk you through the fundamentals then challenge you with small exercises. For instance, learning a programming language: the AI can teach a concept (loops, algorithms), then ask you to try writing a snippet or predict an output, and give feedback.* General Knowledge & DIY Learning: Really, any topic! You could explore history, economics, or even a new hobby. An educator might use it to generate lesson outlines (like I did) or to get analogy-filled explanations for tough concepts. A curious mind at home might use it to learn personal finance basics – e.g. “Teach me about mortgages” – with the AI ensuring they understand each piece (rates, down payments, etc.) before moving on.In all these cases, the common thread is active learning. Study Mode shines when you genuinely want to understand something, not just get a quick fact. It’s the difference between cramming an answer and actually learning the material. I plan to keep experimenting with it – I can see myself using it to stay sharp on both tech and health topics.Bottom Line: My first foray into ChatGPT’s Study Mode was eye-opening. In a short session, I went from zero to a reasonable renewed grasp of moral hazard, had a mini interactive lecture, and walked away with a teaching outline. All from an AI that guided me in a conversational, low-friction way. It feels like having a patient, knowledgeable tutor on call. I’m excited to see how this evolves and how others use it. If you’ve tried Study Mode yourself – whether for a medical topic, an AI concept, or anything else – I’d love to hear how it worked for you and if it made a difference in your understanding.(Feel free to share your experiences or other use cases in the comments!)Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.Curious what the lesson plan that ChatGPT spat out looks like? Check it out below!Mini-Lesson Plan: Moral Hazard in Health Insurance1 │ Warm-up Prompt“Think of a service—health-related or not—where you behaved differently once someone else picked up the bill. What changed and why?”This primes the idea that cost shields change behavior.2 │ Key Definitions (stick to one sentence each)* Moral Hazard – People or firms use more (or riskier) services after insurance shields them from the full cost.* Ex-ante Moral Hazard – Less effort to stay healthy before coverage (e.g., skipping flu shots).* Ex-post Moral Hazard – Heavier use of care after coverage (e.g., ordering an MRI “just in case”).* Deductible – Up-front amount a member pays each plan year before insurance starts paying.* Copay / Coinsurance – Fixed fee or percentage paid at the point of care to keep “skin in the game.”Ask learners to rephrase each in their own words (Rule 4: check & reinforce).3 │ How Insurers Counter Moral HazardLever (adds just-enough friction) How it works Typical Target Cost-sharing (deductibles, copays) Shifts part of the bill to patients Ex-post (patient side) Tiered / narrow networks Lower cost-share at “preferred” providers who deliver value Ex-post (patient & provider) Utilisation review & prior auth Gatekeepers approve only high-value services Ex-post (provider) Value-based payments (bundles, capitation) Clinicians paid for outcomes, not volume Ex-post (provider) Wellness & incentives Premium discounts or rewards for healthy behaviours Ex-ante (patient)Prompt: “Which lever tackles behaviour before care is needed, and which hit after?”4 │ Quick Example to Teach WithScenario: A plan drops all copays for primary-care visits.Likely ripple: Visits per member rise (some needed, some not). Spending may climb unless early primary-care contact offsets ER or inpatient costs.Debrief question: “How could the insurer keep visits accessible yet discourage low-value appointments?”(Hints: tele-triage first, modest virtual-visit fee, appointment availability caps.)5 │ Homework Problems (no solutions given—designed for discussion)* Generic vs. Brand Game* Facts: Plan A covers generics at $5, brand-name drugs at $35. Next year, Plan B makes both $0.* Tasks:* Predict how prescription patterns change (generic vs. brand mix).* Propose one monetary and one non-monetary tweak Plan B could add to curb low-value brand use without hurting access.* “Back-Up” MRI* Facts: An orthopedic clinic is in a fee-for-service contract that pays per imaging study. The payer sees MRI volume 40 % above peers.* Tasks:* Identify the moral-hazard driver from the provider’s perspective.* Outline a value-based contract term that could realign incentives.* List a patient-facing strategy that might also reduce unnecessary MRIs.Encourage learners to justify answers with one-sentence reasoning—keep it concise.Quick Exit Ticket“In one line, explain why moral hazard is about behaviour, not bad morals.”Collect two or three responses to reinforce the core takeaway.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.This is my first time using ChatGPT’s new agent feature. I’ve been experimenting with different ways to offload the basic stuff I do every day—like making infographics and visually compelling slides to communicate ideas or present (see last week’s video). This time, I wanted to see if this agent feature allowed me to take it a step further and generate a full PowerPoint deck using just a one-liner prompt and the files already sitting in my Google Drive.I recorded the whole thing without rehearsing and the result is above.At 6:13 p.m., I gave the agent access to my Drive and asked it to generate a brief presentation summarizing the RevelAi platform—based on internal documents I already had there.The request was very basic :“Please look inside of my Google Drive for RevelAI Health. Find the user manual information, and create a pleasant, brief PowerPoint presentation that captures the essential elements of the RevelAI platform in a way that is easy for the average clinical user to understand.”Honestly, not an awesomely structured prompt, but I wanted to do this on the fly and see the outcome. The agent immediately started pulling documents, analyzing them, and building slides. There were a few hiccups early on (which I actually liked—it showed me what it was doing step-by-step), and the whole thing took about 45 minutes.What I got back was a solid draft. The branding was not great. The visuals weren’t polished. But the structure and content were correct—and drawn directly from the source material. I could hand this off to a designer or a virtual assistant, or drop it into my next internal meeting with light edits.And the key is: I didn’t touch a single bullet point—I think as a first draft to kickstart it did a good job.A few takeaways from using the agent:* It’s not always faster than doing it yourself.* But it’s faster than starting from scratch.* And it’s way better than letting something sit on your to-do list for two weeks.A few use cases this unlocks (with example prompts + workflows):1. Quick onboarding decksTurn existing materials into usable slideware for new hires or stakeholders💬 Prompt:“Read our onboarding guide in Google Drive and make a 5-slide intro deck covering the most important policies and tools.”🔁 Workflow:Agent fetches the document → summarizes key sections → generates a branded draft → you tweak visuals → done.2. Meeting prep packsSynthesize past notes, briefs, or recordings into a pre-read deck💬 Prompt:“Pull my last 4 weekly updates and create a 3-slide summary of what to review before tomorrow’s strategy sync.”🔁 Workflow:Agent fetches notes by date or tag → summarizes key trends or blockers → assembles slides with bullets + visuals → you review and send.3. One-pagers or client briefsAuto-generate quick-reference materials from long-form docs💬 Prompt:“Make a one-slide summary of our value proposition based on the BD primer and sales playbook in Drive.”🔁 Workflow:Agent compares the two docs → pulls recurring phrases and benefits → formats into a clean, readable layout → you copy into email or deck.The Techy Surgeon TakeCould I have done this faster manually? Maybe.But would I have done it at all without blocking out time? Probably not.In the case above, I used a tool connection to my Google Drive, but you can let the agent take action in Canva, Gmail, and a lot of other applications. So a more complex multi-step workflow may have gotten me a more impressive result, and I'm excited to try these other workflows out. I think the take home is that these ChatGPT agent workflows feel like giving your future self a head start. These AI agent use cases are The output is not meant to be perfect (ye). It’s meant to get you to “good enough to iterate” in one step. And that’s where the leverage comes in.Techy Surgeon is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

🧠 When GPT writes your notes, what’s your brain doing?A new MIT Media Lab experiment wired 54 volunteers to EEG caps while they wrote multiple SAT‑style essays under three conditions: no tools, Google Search, or ChatGPT. Those relying on ChatGPT showed the weakest pre‑frontal and parietal connectivity—the neural circuits tied to working memory and executive control. Even after the researchers took the AI away, those brains stayed under‑engaged, a phenomenon the authors dub “cognitive debt.” MIT Media LabQuality and ownership dipped, too. Teachers rated the ChatGPT essays the least original, and participants later struggled to recall or quote what they had “written,” confirming that speed came at the cost of synthesis and memory. TIMEWhy this matters in clinicAs an orthopedic trauma surgeon I still let an ambient scribe draft every routine fracture‑clinic note—but many in other fields like internal medicine and less procedural specialties have told me that they let generative AI only draft the clerical parts: meds, template language, today’s HPI. The assessment & plan they write line by line. That split keeps paperwork out of a doctor’s head during the visit yet forces them to wrestle with the medical reasoning afterward. Think of AI as a middle pass:* Dictate a scaffold. Voice out the structure in your own words.* Let the model fill. Off‑load rote phrasing and formatting.* Interrogate the draft. Edit until every decision reflects your judgment.If composing the note is how you think , keep writing. But if note‑taking just burns cognitive calories you’d rather spend on the patient, teaching, or research, outsource it—then reinvest that bandwidth where it counts.Open questionWould the same neural slump appear during real‑time clinical reasoning with ambient scribes running? Until someone reruns the study in exam rooms, be intentional: deploy AI where it frees you, guard the zones where insight lives—and don’t let your hard‑won clinical muscle atrophy for a few extra clicks.Link to full study here https://arxiv.org/abs/2506.08872 This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit techysurgeon.substack.com/subscribe