The Anatole shutdown
When I decided to stop Anatole, our AI product, it wasn't a failure. It was an act of lucidity.
A good product sometimes deserves to be stopped.
Anatole was our AI assistant for tourism. For eighteen months, we built it, tested it, deployed it with industry professionals. It worked. Users found it useful. Feedback was positive.
And yet, I stopped it.
Lucidity as a skill
Stopping a product that works is counterintuitive. But the question wasn't "does it work?" The question was "is this the right fight?" The answer was no.
The market for conversational AI assistants is commoditizing at a speed no one anticipated. What we were building with Anatole, the large models were going to do natively within six months. Not as well, not as specialized. But well enough that our advantage would shrink to nothing.
What the field taught me
350 professionals trained in one year. Hundreds of hours in the room. What I learned in the field wasn't that AI works (everyone knows that). It's that the real value lies in supporting the transformation, not in the tool.
The tool is replaceable. The ability to read an organization, to identify where AI creates value and where it destroys it, to support the shift from level 1 (relief) to level 3 (reconception), that ability is irreplaceable.
Anatole taught me one thing: building a good product and building the right thing are two different skills.
What comes next
KiXiT continues. Not as a software company, but as a consulting and training firm. The field remains the same: AI in tourism and territories. But the angle shifts. We support the transformation. We no longer sell the tool.
It's a choice that costs in the short term. But it's the lucid choice.