Insurance that watches itself
The right policy for every build, monitored continuously: no one left holding the job of remembering to check.
The situation
Every build needs the right insurance policy. Choosing it was hours of manual work per house. And choosing was only the beginning.
Once a policy was in place, someone had to keep track of it: which build was covered by what, what was active, what needed attention. That tracking was an open-ended job: the constant, easy-to-drop task of remembering to go check, across every active build at once.
What we built
A system that selects the right policy for each build and then monitors it continuously. Instead of a person periodically remembering to look, the system watches every policy and surfaces only what actually needs a decision.
The hard part wasn't the monitoring. It was encoding how this builder thinks about coverage: what the right policy looks like for a given build, and what's worth raising a flag over. That judgment lived in someone's head. Now it's written into the system as rules, locked by automated tests, so a future change can't quietly break it.
The most expensive task here wasn't the work. It was remembering to do the work. So we removed the remembering.
The result
The manual work per house is gone, and so is the mental load of remembering to check. The system flags what needs attention and otherwise just runs, quietly, in the background, all the time. And the judgment behind it stopped being something the company could lose with one departure. It's an asset of the business now.