Ep. 40 Navigating Automation in Revenue Cycle Management with Michael Laukaitis

Navigating Automation in Revenue Cycle Management with Michael Laukaitis In this episode of 'For The Love of Revenue Cycle', host Vanessa Moldovan dives into the world of automation with healthcare IT expert Michael Laukaitis. With over two decades of experience, Mike shares his journey from rural Montana to UT Southwestern, highlighting the evolution of automation in healthcare. The discussion covers the importance of evaluating processes before automation, the differences between RPA and agentic AI, and the need for human oversight in automated systems. Mike emphasizes the necessity of stable, high-volume processes for successful automation and warns against over-promising vendors. Mike and Vanessa provide practical advice for healthcare leaders on selecting and implementing automation tools, ensuring measurable outcomes, and maintaining system integrity. 00:00 Introduction to the Episode 00:36  Meet Mike Laukaitis  01:00 Early Automation Experiences 02:05 Current Automation Projects 03:53 Challenges and Considerations in Automation 04:47 When Not to Automate 07:15 Human Element in Automation 09:58 Technical Aspects of Automation 17:54 Evaluating Automation Vendors 25:39 Best Practices and Lessons Learned 33:37 Conclusion and Final Thoughts What makes a process a good candidate for automation? Use this quick scorecard (1=low, 5=high). Aim for 24+ out of 35. Volume and frequency: high, daily work. Rule clarity: clear decision rules or prompts a model can follow. Input quality: structured data, stable forms, reliable sources. System stability: few UI changes, APIs available, or predictable portals. Exception rate: historically under 15% and well-defined. Impact: measurable time saved, faster cash, fewer denials, fewer clicks. Risk profile: errors are detectable and reversible. Questions leaders should ask before building or buying automation tools What is the exact business outcome and how will we measure it monthly? Where is the waste today: rework, wait time, over-processing, or motion? Is the process stable and documented, or do we need to fix it first? What is the current exception rate and why do exceptions happen? What data sources are required and who owns their quality? What are the failure modes and how will we detect, alert, and roll back? What access, audit, and compliance controls are required end-to-end? Build vs buy: which option reduces time to value and long-term maintenance risk? What's the realistic total cost of ownership, including rebuilds after system or payer changes? How will staff be trained, and what tasks will they do when the bot takes over the boring parts? Who is the named process owner and who shuts it off if metrics slip? What is the decommission plan if the upstream system adds a native feature? Implementation guardrails that save pain later Fix the process first. Map it, remove steps, standardize inputs, then automate. Start in "shadow mode." Let the bot run in parallel for 2 weeks and compare outcomes. Instrument everything. Log starts, stops, exceptions, and outcomes. Alert on drift. Keep humans in the loop for edge cases and approvals above risk thresholds. Review quarterly. If payer or EHR changes make it brittle, redesign or retire it.