Name the decision first
If we cannot identify who decides, how often, and what changes when they decide differently, we will say so before you spend a budget on it.
About Occam Labs
Small by design. Built for consequential work.
Occam Labs is an independent AI and data consultancy founded in 2016. The person you meet is the person who does the work — there is no delivery team behind the pitch.
Our premise
Most failed AI programmes were never technically wrong.
RAND studied why AI projects fail at twice the rate of regular IT projects. The root causes were not model quality or data volume — they were missing ownership, absent governance and no change management. In other words: design faults that are fixable before a line of code is written, if you treat them as first-order work rather than afterthoughts.
Who you work with
Alain Tourtet
Founder and Director
Alain has spent thirteen years on the same problem from three different sides: building enterprise data products at IBM, advising on them at Tableau, and since 2016 running Occam Labs.
At IBM he deployed analytics products used by 1,000+ people at Shell, and learned early that the change programme mattered more than the model. At Tableau he delivered more than 100 enterprise analytics solutions and trained 200+ partner consultants at Deloitte, Accenture and PwC — which is where the pattern became obvious: the technical work was rarely the constraint.
Occam Labs has since worked on an EMEA healthcare analytics platform that reached 4,000+ users across 20+ markets, a risk and scenario platform for a Lloyd’s-backed reinsurer, and CTO-level portfolio governance across 35+ concurrent technology initiatives.
He still writes code. Recent work includes a multi-agent LLM platform built on LangGraph with Claude and Gemini, rebuilt from v1 to v2 for a 15× reduction in cost per task — which is where the views on agentic unit economics on this site come from.
How we work
Three working rules we will hold you to.
These shape what we agree to do, and occasionally what we advise you not to do.
Every pilot has a kill condition
We agree in advance what result would mean stopping. A pilot that cannot fail is a procurement exercise wearing a lab coat.
Your engineers own the repository
From week one, not at handover. If we disappeared mid-engagement your team could continue — that is the test we design against.
The Occam principle
Prefer the explanation that needs fewer assumptions.
Applied to systems, that means we look for the smallest intervention that changes the decision. Often the honest answer is that a workflow needs a data contract and two hours of someone's authority, not an agent. We would rather tell you that than build the agent.
Work to date spans enterprise analytics platforms, risk and scenario intelligence, technology portfolio governance, and hands-on AI product development across healthcare, insurance, entertainment and consumer businesses.
See selected workStart a conversation
Bring the strategy and the system into the same conversation.
Tell us what is changing, where the difficulty sits and what a good outcome would look like. We reply within two working days.