Stop Doing Your Own HR

Most founders don't think about HR until HR thinks about them. And by then, it's a letter on the desk demanding $38,000 and a lien notice attached for good measure.This week on AI for Founders, John, the founder of CogNet HRO, walks through the quietly catastrophic world of multi-state payroll, the surprise tax bills no one warns you about, and why a guy who spent fifteen years running this thing as a side hustle suddenly grew it from 67 to 600 employees in under five years. He moved offshore back when "doing business in India" still made boardrooms nervous, built a 600-person team in Chennai, and now runs a service operation that lets founders skip the part where they wake up at 2 AM wondering if California changed its overtime laws again. (Spoiler: California changed its overtime laws again.)The conversation goes deep on what AI can actually do for HR right now, what it absolutely cannot, and why CogNet built its own internal ingestion tool called Drive instead of letting client PHI bounce around inside Claude or ChatGPT. John is refreshingly blunt: most of the AI tools the big payroll providers are bragging about are still glorified bots. The real wins are in robotics, document migration, and the unsexy automation work that lets a small founder team punch above its weight.Frameworks discussed:Land and Expand: Solve one acute pain (usually a tax notice), then earn the right to handle payroll, benefits, finance, and HRIS implementation. CogNet is internally organized by practice area, not client, so the expansion is structural.The Bus Theory of Hiring: Don't fire fast, reseat fast. The hard skill is figuring out where someone fits, not deciding they don't. Took John three years to nail this with one senior manager.Predictive Hiring Modeling: CogNet is pulling its own historical hiring data to model who actually thrives, knowing humans are irrational but the patterns aren't.Ingest First, Decide Second: Drive is built to absorb anything (PDFs, registers, JSONs from terminated providers) before any decision gets made about whether AI, robotics, or humans handle it.Robotics Over AI for Repeatable Tasks: When the job is "do these five steps 500,000 times," skip the LLM. Spin up 18 robots on AWS and let them grind 24/7 without exposing data.Multi-State as the Trigger Point: The moment a company hires across more than one state, the compliance math changes. That's the founder's signal it's time to outsource.https://www.cognethro.comhttps://www.linkedin.com/in/john-sansoucie-033b20/https://www.linkedin.com/in/estesryan/⁠⁠https://www.inboxalchemy.cohttps://www.aiforfounders.cohttps://trynina.co/