Quentin Lecocq

I build systems that take over repetitive workflows, without giving AI free rein.

I start from a concrete operational task: qualifying incoming requests, checking a file for completeness, preparing a draft response for you to send, or consolidating scattered information. I figure out what can genuinely be automated, where human validation is still needed, and then I build the system and connect it to the existing tools. I work in the form of an "AI System Sprint": a clearly scoped workflow, delivered in 2 to 3 weeks from scoping to handover, covering development, testing and deployment, with code, access, documentation and runbooks handed over at the end.

The systems I build can combine classic automation, calls to language models, and bounded agents: AI is only used where it actually adds value. Permissions, human validation steps, and failure-handling behavior are all defined before going live. As one example, I built a pilot request-qualification system for a joinery/woodwork sales company: it structures incoming requests, flags missing information, and routes atypical cases to a human. It's still in use today.