The challenge
In compensation management, incorrect results can undermine the decisions customers rely on the product to make. OpenComp needed to adopt AI-first development with a testing approach suited to both the speed of agents and the importance of accuracy.
Our work
Give agents the right working context
We helped reorganize repositories and add reusable skills, instructions, and guardrails so agents could work within the team’s conventions.
Train the people directing the work
We trained the product and development team to work with coding agents, define the intended outcome, and assess what the agents produced.
Rework quality assurance
We helped the team adapt QA to check work produced by coding agents, with particular attention to accuracy in compensation management.
The result
OpenComp’s team gained a practical foundation for AI-native development: organized repositories, shared agent skills, and a quality process designed to keep pace with the work.