Ask the team at any business you're looking at what AI they use at work. Most will say ChatGPT, on their own phone. Then ask what they put into it. Menlo Security's 2025 data found 68% of employees using free personal AI accounts, and 57% of those putting in sensitive company data. In a small service business, that's the customer list, the pricing and the job notes. Insurers have noticed: Verisk now offers standard generative AI exclusions for general liability policies, and carriers have started adding them. The usual post-close move is to buy everyone a seat. That gets you the model, not the business. MIT found about 95% of enterprise generative AI pilots showed no measurable impact on P&L, and traced it to how the tools were fitted into the business, not to the models. What worked for us: I'm the GM of a multi-base outdoor company in Idaho. We put the operation in one place first: guides, qualifications, gear, vans, trips and the rules connecting them. Then the AI could read and write there, inside limits a person sets, with an audit trail on every change. Weekly planning went from 20 to 40 hours to 2 to 4. Guests went from 1,279 a year to more than 4,700, with a three-person office. Before close, I'd check: 1. What AI do people use, and on whose account? Ask the people doing the work. 2. How many apps hold the schedule, the customers and the invoices? 3. Which of those apps let you get your own data out, and which let anything write back in? 4. Who decides what an AI is allowed to change, and who approves it? 5. Could you show an insurer what the AI did last week? None of these is a reason to walk away. They tell you what the first hundred days look like. Across a portfolio, doing it the same way in every company is where it compounds. Curious what others are seeing: when you get into a business, is the team already using AI on their own? The long version, with sources: justinsmith.build/writing/everyone-bought-the-ai/