AI
Is artificial intelligence accessible to businesses of all sizes?
Yes — but the cost has moved from the model to the data feeding it. Here is where the budget actually goes.
JS
Julián Soto28 Jul 2026 · 6 min read
Access to capable models is no longer the constraint. A small company can call the same frontier model as a bank, priced per token, with no contract. What separates a useful AI feature from an expensive demo is whether the model is looking at your data or at the open internet.
That is where the real budget sits: collecting, cleaning and keeping current the information the model reasons over.
Where the money actually goes
On the projects we run, model inference is rarely the largest line. The distribution looks closer to this.
If the answer depends on your data, the project is a data project with a model attached.
Data access and preparation: the pipelines, permissions and cleanup that make your own records usable.
Evaluation: a test set that tells you whether a change made the output better or just different.
Integration: putting the result inside the tool your team already uses, not in a separate chat window.
/ AI // Business // Cost /