A process that ran on email
A friend of mine owns several Sport Clips locations. Every store has to be inspected on a schedule, and the whole thing ran on email. Photos attached to threads. Replies confirming which stores had actually been marked complete. Somebody keeping track of which location was due when, across all of them, mostly in their head.
None of that is a technology problem in the exciting sense. It is a coordination problem, and coordination problems are where the time actually goes. The corporate side had been running this way for years and was curious whether anything better existed.
It also felt familiar. Managing many locations from one screen, knowing what is healthy and what needs attention, is the exact shape of the problem I spent years on with OvrC. Different industry, same job: give one person a view across sites and make the exceptions obvious.
What it does
Inspec puts every location on a map and in a list, each with its status and its next due date. Open a store and you get its inspection history, its outstanding items, and the photos attached where they belong instead of buried in a mail thread. The person running the franchise sees the whole portfolio at once and can tell in a glance which store needs them this week.
Taking it to the CTO
We presented it up to the franchise's CTO, and it was well received. It also did not get adopted, for a reason that had nothing to do with the work: they were already partway into implementing an off-the-shelf system, one considerably more expensive than a dashboard their operator's friend built on nights and weekends.
That is a normal outcome and worth saying plainly. Procurement had moved before the demo happened. The door is not closed, and if the other system does not land the way they hope, this is still sitting here.
What I actually learned
This was my first real attempt at building with AI, in March 2025, when the tooling was much rougher than it is now. The lessons were mostly about me rather than the model.
Requirements are the whole job. The quality of what came back tracked almost exactly with how clearly I had described what I wanted. Vague ask, vague result, then twenty minutes of correction. Time spent writing a precise description up front paid for itself several times over, which is a lesson that transfers directly to writing a good spec for any team.
Troubleshooting one thing at a time was more painful than it needed to be. I would change several things at once, or point at the wrong layer, and then spend the next pass undoing my own misdirection. Learning to isolate a single problem and describe only that problem is the skill that separated a frustrating session from a productive one.
Both of those are things I now do reflexively, and both of them made every project after this one faster. Inspec is the least polished thing I have built, and it is the one that taught me how to build the rest.