AI can't be bought. It's learned — and led.
Six things we believe.
These are lessons we learned building and transforming companies in telecom, healthcare and banking — brought into the age of AI. We apply them the same way in consulting, in The AI Academy and in Enterprise Agentic OS.
P.01
The person who does the work knows best what's worth automating.
The person who owns a process knows the exceptions, the unwritten rules and the three customers who are always handled differently. No requirements document captures that. That's why the best automations come from the people who work in the process every day, not from a central department. We prepare them; we don't route around them.
P.02
Speed is of the essence.
People who know their process can build an agent in an afternoon today, without going through IT. It's one of the biggest productivity gains technology has ever brought, and it must not be slowed down. We don't put people in a queue: we let them move fast and put the markers along the road, not barriers in front of it.
P.03
Governance is transparency at the right moment.
Not a committee that approves everything, and not a report that arrives afterwards. Good governance shows each person exactly what they need to know, exactly when they need it: the builder while building, the manager when deciding, the auditor when asking. Low-risk agents pass without friction; high-risk ones reach a human.
P.04
The agents you don't know are the risk you can't explain.
An agent built by a colleague runs on the credentials of someone who has since changed roles. It reads personal data without any assessment. It makes decisions about people. Nobody knows it exists — until an auditor asks. Then it's no longer a technical problem but a question of accountability.
P.05
Strategy only counts once it reaches operations.
A strategy document creates no value on its own. That's why we don't stop at recommendations: together we pick the first cases, take them all the way to production and measure the effect. If people don't work differently after a project, the project didn't happen.
P.06
Adoption is measured, not declared.
The number of licences, people trained or agents built doesn't matter. What matters is how many processes work differently, and how much of the AI running in the company is known, has an owner and was approved before going live. That's the difference between a company adopting AI and one merely accumulating it.