I’ve seen it dozens of times: a company buys a ChatGPT Enterprise license, sends out an email saying “from now on, we use AI” and waits for a miracle. The miracle never comes.
Why implementations fall apart
The most common mistakes I see:
- Starting with the tool instead of the problem
- No champions: nobody to drive the change forward
- Too broad a scope: trying to change everything at once
- Zero measurement: no way to tell whether anything has improved
My method: 6 steps
After dozens of implementations, I’ve developed a process that works:
1. Conversation
45 minutes with the decision-maker. No pitch. Just questions about people, processes and goals.
2. Diagnosis
I identify 3–5 places where AI will genuinely change results. Everything else gets set aside.
3. Plan
A concrete document: what, who, when and at what cost. The plan is yours, even if we don’t work together afterward.
4. Pilot
Training for one team + tool configuration + a pilot with a single process.
5. Scaling
Documentation, templates, best practices. I teach you how to do it on your own.
6. Independence
After 3 months, you shouldn’t need me day to day. That’s what success looks like.
The key to success
Don’t implement AI. Implement a change in the way you work, with AI as one of its elements.
That’s the difference between “we have AI” and “AI helps us.”