Why your first AI pilot stalled.

Most operators we meet have already tried something. A chatbot on the website, an AI add-on from their software vendor, a few months with a tool someone saw at a conference. It demoed well. Then usage drifted, nobody could say whether it was working, and it quietly stopped mattering.

It's tempting to conclude the technology isn't ready. Usually, three other things went wrong first.

The workflow was chosen for the demo, not the business.

Pilots tend to start where AI looks impressive — a friendly assistant answering questions — rather than where the hours actually go. The best first workflow is boring, frequent and measurable: after-hours maintenance intake, weekend lead response, month-end owner commentary.

Nobody wrote down what “good” meant.

If you can't say what your best leasing agent would have replied, you can't tell whether the AI's reply was better or worse. Without a baseline and a standard, every pilot ends in opinions.

It sat beside the work instead of inside it.

A tool your team has to remember to open is a tool they'll stop opening. AI that works is wired into the queue they already work from, with its output waiting for a decision rather than a visit.

A stalled pilot isn't evidence that AI doesn't work. It's evidence that the pilot wasn't designed to prove anything.

The fix is unglamorous: pick one workflow that matters, measure it before you start, define good with the people who do it best, and put the AI where the work already happens.

— PMStack · October 2026