How we built an AI operational brain to reduce founder friction
Built for ourselves (internal), then offered to clients.

- 70%Reduced founder dependency
- 20h+Saved per week
- 99.9%Availability
Where the founder became the bottleneck
Like most growth-stage companies, we hit a point where the founder was the bottleneck. Onboarding decisions, approval routing, exception handling, knowledge lookups, every meaningful path eventually ended at a single inbox. The team grew, but the bottleneck didn't move.
How we designed the operational brain
We built the system we'd want to sell. A custom backend that holds operational state. A vector store that captures institutional knowledge from documents, conversations, and prior decisions. A small fleet of LLM-driven agents, qualifier, router, summariser, scheduler, coordinated by an explicit router with guardrails and retry-loops. A dashboard the team actually uses every day.
What changed: 70% less founder dependency
70% of the operational decisions that previously ran through the founder are now handled or pre-routed by the system. About 20 hours per week reclaimed. Availability tracked at 99.9% over the first six months. Most importantly: the team can answer 'why did the system do that' from the audit log, which is what makes the trust real.
Other systems we've shipped.
- 2026 · independent garage, france
How an AI voice agent answers every call a French garage used to miss
An AI voice agent that answers an independent French garage's phone in natural French, checks live calendar availability, and books the appointment during the call, with RGPD (GDPR) compliance built into the data layer.
- 2025 · innovium
Marketplace pricing intelligence & operations automation for a multi-country seller
A pricing-intelligence and operations automation system for a multi-country marketplace seller. Continuous price monitoring across regions, automated repricing logic, and operations dashboards that replaced nine manual spreadsheets.
