FOUNDER · EVERY WEDNESDAY · AUGUST 19, 2026
Your teams are already using AI. Without you.
A lunch, a few months ago, with the CEO of a fine industrial SMB. He tells me: "AI? We're not there yet. I'm waiting for things to become clearer." Twenty minutes later, his sales director shows me on his phone how he prepares his client meetings: he pastes the client's history into a free AI and asks it for a meeting plan. Client data, in a tool nobody vetted. The CEO was stunned.
I'm not telling this story to point fingers: I've relived that scene several times since, including at very well-run companies. And I understand the teams. The tool saves an hour, so people use it. It isn't indiscipline, it's individual common sense. The real issue lies elsewhere: AI is the first technology that enters a company through its employees, not through management. While we "wait for things to become clearer," usage settles in, without rules and without protection. A recent study shows the scale of it: 59% of employees use AI tools their company has not approved.
The Spanish bank BBVA did the exact opposite of banning. It opened AI access to those who asked for it, volunteers rather than appointees, with two simple rules set from day one: confidential data never leaves, and an AI answer gets verified before it is used. Then it organized the sharing: everyone publishes their assistants and their uses in an internal library, and teams show each other what works. Within months, employees had created nearly three thousand shared assistants, saving about three hours per week per user.
The most interesting part isn't the number. It's that the early skeptics joined in, not because they were told to, but because a colleague showed them a real use on a real file. And the delicate subjects (confidentiality, made-up answers, what requires human review) were learned right there, in real situations, far better than through any memo. Why did it work? Because the bank didn't try to control the enthusiasm: it channeled it. The rules made usage safe; the sharing made it contagious.
The 3 essentials for you
- Your company is already using AI. The only question is: with you, or without you.
- The lesson from BBVA: you don't control enthusiasm, you channel it. Rules make usage safe; sharing makes it contagious, including for the skeptics.
- Two rules are enough to start: what never leaves (your sensitive data), what always gets verified (any answer that commits you). The rest is learned between colleagues, not through a memo.
What I would do in your shoes…
Take this bank's playbook and bring it to SMB scale: it requires neither a platform nor a budget. First, this week, ask the question in a team meeting, with no sanctions attached: "who uses what, and for what?" You'll be surprised, and it's rather good news: your teams already want to, just like BBVA's volunteers. Then write its two rules, the same ones, on a single page. Then create your own sharing library: thirty minutes every two weeks where everyone shows a use that saves them time. That's exactly the moment that, over there, merged best practices, won over the hesitant, and taught the sensitive topics without a single memo. And as soon as you can, give your teams something better than the free tool: a vetted one, fed with your data, that makes workarounds pointless.
Write to me. Are you living this situation? Did you solve it differently? Write to me: arnaud@yprtnrs.com. I read everything, and I will reply to you personally.
