Form Intel
Making equipment issues in inspection reports visible.
Finding the field reports that need attention.
Form Intel. Crane crews submit inspections and daily reports, but the office doesn't have time to read every one. Form Intel uses AI to flag entries that may need attention and summarize why. We shipped it in September 2026.
My part
I built the demo with an engineer, then owned the requirements, design, evaluations, and cost limits for the production feature. Engineering built it. I narrowed the scope to flagging notable entries and summarizing those, rather than generating a report for every submission.
More about my work at Ribbiot
The problem
We could already email every form to the office. Admins told us they couldn’t go through all of them. They wanted to know which ones needed a closer look.
What we tested
I built the demo with an engineer using real forms. I then built evaluations so we could compare models on what they flagged and how much they cost to run. Defining what deserved attention was part of defining the product.
What I narrowed
I scoped the production feature to deciding whether a submission contained something notable, then generating a summary only when it did. That kept the work focused on the review list admins had asked for and put a limit on how much analysis we needed to run.
What we’re learning
The inspections are surfacing existing equipment issues that crews already know about and work around. At Ribbiot, we hadn’t had a clear view of how often those issues were being recorded. Form Intel is making that information visible.
What shipped
Form Intel went live in September 2026. I owned the requirements, design, evaluations, scope, and cost limits. Engineering owned the production implementation. Admins can review the alerts and give feedback when one isn’t useful.