Last week, in September 2026, I attended an evening event at MKE-SPIN, the Milwaukee Software Process Improvement Network, a volunteer-run nonprofit that brings people who work in software, process, and data together once a month. The talk that night focused on turning expertise into repeatable skills for AI, and I left thinking about one question: what happens to what we know when it lives only in our heads?
The speaker, Robert M. Orozco, is an AI consultant and Executive in Residence for AI at the University of Wisconsin–Whitewater. As an independent AI consultant, he works with teams to help them move from one-off AI results to enterprise AI systems. His work focuses on architecting proprietary AI systems, AI governance, and enterprise AI implementations. He is a practitioner rather than just a commentator.
Robert made a point that landed with everyone in the room. When you build a great workflow inside your own AI chat, the chat learns your process, while the rest of your organization stays in the dark. The day you take a promotion or move to a new role, your team has to rebuild that knowledge from the beginning.
I see the same pattern in the small businesses I work with, and it shows up long before anyone opens an AI tool. An estimator knows exactly why one bid carries a higher margin than another, and an owner knows which lead deserves the first call on Monday morning. That judgment sits with one person, and the business depends on it every day.
What struck me in the talk is how much AI raises the stakes on documentation. A model follows the instructions you give it, and when your instructions leave a gap, the model fills that gap on its own. Robert offered a simple test that I have been repeating to clients since: if an intern would need to come back to you with questions before they could do the task, then the task still needs the documentation an AI system depends on.
I come to this as a workforce developer who spends a lot of time in data, and I think about it as a missing-data problem. Every undocumented rule is a null value in your process. When a system reaches a null, it can stop and flag the gap, or it can impute a value that looks reasonable. In a spreadsheet, imputation is a choice you make on purpose and label clearly. In a business process, it happens quietly, and you end up with a confident answer built on an assumption that someone in the business would have corrected in a second.
The hardest knowledge to capture is the knowledge experienced people have stopped noticing. After twenty years in a trade, the exceptions feel like common sense, so they rarely make it into a manual. My favorite practical tip from the evening was to narrate your work as you do it: turn on talk-to-text and explain what you are doing and why, in the moment. You end up with your real reasoning, captured while it is fresh, which is far more useful than a tidy summary written from memory a week later.
At Dr. Data, my method for capturing this knowledge is the Skill Threshold Zone, or STZ. It follows a path very close to the one Robert walked through, starting with how the expert actually does the work and building the system from there, and it was good to hear a practitioner arrive at the same place from a different direction.
Documenting expertise also protects what you have built. AI models change every few months, and a written process moves with you to whatever tool comes next. It also keeps your know-how under your control, which matters because your process is a big part of what makes your business different from the one down the street.
My biggest takeaway from the evening is that encoding expertise is where every good AI project starts. I open every engagement by sitting down with the owner and the people who do the work, and writing down how decisions actually get made.
Thank you to Robert for a practical, generous talk, and to MKE-SPIN for creating a space where people who care about doing this work well can learn from each other. If you work in software, process, or data around Milwaukee, I would encourage you to join them at an upcoming meeting.
Want AI built for your actual job? Book a discovery call.