There is an impression — encouraged by a lot of marketing — that hiring an AI studio means someone builds you an artificial intelligence. Almost nobody does that, and the handful who do are training foundation models at a cost measured in the tens of millions.
What everyone else does, including me, is build the system around a model somebody else trained.
Rented versus built
The intelligence is rented. Claude, GPT, a speech provider for voice. You pay per token or per minute and you get a capability that would be irrational to reproduce.
The system is built. The architecture, the integrations, the retrieval over your data, the guardrails, the evaluation suite, the interface, the deployment. That is the part that is specific to your business, and it is the part that determines whether the rented intelligence is useful or embarrassing.
Why say so out loud
Because the client finds out eventually, and it is better coming from the person they are paying.
It also reframes the conversation usefully. If the model is rented, then the model is not the differentiator — your competitor can rent the same one tomorrow. What they cannot rent is a system that knows your pricing, your stock, your policies and your tools. That is where the value sits, and that is what the invoice is actually for.
The question to ask any studio
Not "which model do you use" — that answer changes every few months and should. Ask what happens when the model they chose gets worse, gets expensive, or gets deprecated.
A good answer involves fallbacks, an evaluation suite that would catch the regression, and a model layer that can be swapped without rewriting the product. A bad answer is a brand name said with confidence.
