How do we avoid becoming dependent on an AI supplier?

Dependence sets in quietly. You adopt a supplier for a first use, then a second, and you build on its tools, its formats and its own features. One day, it changes its prices, its terms or its model, and you discover it is very hard to leave.

You cannot avoid all dependence: you have to rely on someone. But you can choose what you keep. Your data, your processes, your decision criteria, the instructions given to the AI and the tests that check its answers must stay with you, in a format you control. The model then becomes a part that can be replaced.

• Dependence is measured by the cost of leaving.If changing supplier means rebuilding everything, you are already dependent.
• A supplier’s own features are the stickiest.They save time at the start, and cost a great deal of it when the time comes to leave.
• For everyday uses, models are increasingly alike.Being able to switch becomes a bargaining point.

1. 01For each use, ask what would have to be redone if you changed supplier tomorrow.
2. 02Keep the instructions, examples and test sets of your AI uses on your side.
3. 03Require in your contracts the right to retrieve your data in a usable format.

We help companies keep control of what belongs to them, and make the rest replaceable.

• Should we host our own models or use a major provider?→
• How do you keep up with a technology that changes every month?→
• ‘The bigger the model, the better the result.’→
• Is our data safe with generative AI?→