71% of support tickets resolved by an agent.
A RAG agent over catalogue and policies. Median latency 1.2s. Continuous evaluation over 14k real tickets.
Divisions
03 — AI
Artificial intelligence, productized to engineering standards.
AI takes artificial intelligence from demo to production: agents, RAG and custom models, built with evaluation, guardrails and monitoring. Measurable results, not promises.
Capabilities
What we do.
Production agents
Agents that complete real tasks inside your workflows, with clear limits and oversight.
Enterprise RAG
Retrieval over the company's knowledge: sourced answers, not hallucinations.
Custom models
Fine-tuning and proprietary models where the case justifies it; no complexity for fashion's sake.
Evaluation & guardrails
Evaluation harnesses, adversarial testing and safety barriers before anything reaches a user.
Data & AI strategy
Where to invest and where not to; what to build, what to buy, and what can wait.
In-product integration
AI embedded in the software your team already uses, not one more tool in another tab.
Representative work
What a project looks like.
Representative examples of the division's work and the kind of outcomes we aim for.
Internal search the team actually uses.
RAG over years of scattered documentation. Finding an answer went from minutes to seconds.
Demand forecast at 6.4% mean error.
A proprietary model over four years of demand. It replaced three internal heuristics and cut overstock 22%.
Does your case fit this division?
A first 30-minute conversation is usually enough to find out.