Development process analysis and AI adoption plan
A measured analysis of a product software company’s development process and a plan for adopting AI: where the time goes and what to do about it.
Code review had concentrated on a few senior people.
A product software company had grown its development substantially over two years, many people had been on the team only briefly and the quality of briefs varied. Code review concentrated on a few senior people, who thereby became the ceiling on throughput. The question was where time is really lost in the process and how to apply AI to those places measurably.
Custom tooling over the data, then workshops and a plan.
We wrote our own tools in Python that pulled the full pull-request history and CI data through the API, computed derived lead-time metrics and generated an interactive HTML report with charts. On that we built two in-person workshops and a closing report.
The data base was 627 pull requests from 13 repositories over six months, cleaned together with the client of release branches and deployment commits. The output was a measured baseline of the process, not an improvement: there is no repeat measurement in the materials.