The four models and how each fails
Hourly billing prices attendance rather than outcome, and it makes every clarifying question a billable event. Time and materials with a cap improves the ceiling and keeps the incentive.
Fixed fee per outcome transfers estimation risk to the firm, which is where it belongs, and it forces scope to be written precisely. Retained capacity works for ownership roles such as a fractional CAIO and works badly for delivery.
What moves a build price
Three variables account for most of the range in an enterprise AI build.
- Number of integrated systems. Each system of record adds a permission model, a failure mode, and a reconciliation path.
- Regulatory review requirement. Evidence obligations change what ships, not just what is documented.
- Whether an eval harness is required. Where a model is in the loop on a decision that matters, it is required.
Why we publish bands
Total price opacity is the category norm. We publish bands because a buyer comparing three proposals should be able to compare shapes before they compare firms, and because an unstated price gets estimated for you by a directory.
Our commercial policy is on [how we price](/how-we-price), in full, including the payment schedule and what does not trigger a change order.
Next step
Read our pricing logic, then take the free NATIVE Audit if you want a baseline before any conversation.