New research shows enterprises now re-evaluate AI vendors every six months, and want to pay for outcomes, not tokens

Enterprise AI buying is starting to look very different from traditional software buying. Madrona surveyed 150 enterprise IT buyers and found that fewer than half of AI pilots make it into full production. At the same time, 77% of buyers now reassess AI vendors every six months, or continuously. That is a sharp break from the multi-year contracts and slow switching cycles many SaaS companies have relied on.

The message for vendors is simple. A polished demo and a long contract are no longer enough. Buyers are watching whether a tool works in their actual workflows, with their data, teams and operating constraints. If it does not show clear value quickly, they are prepared to look elsewhere. That puts pressure on AI companies to prove adoption, reliability and commercial value after the sale, not just win the initial pilot.

A companion survey from a16z of 50 technical buyers adds another shift. More than half said they want pricing tied to outcomes rather than usage measures such as tokens. For an operator buying AI, that makes sense. You do not really care how much compute a supplier consumed. You care whether customer response times fell, reporting became faster, sales teams followed up more consistently, or staff stopped doing repetitive work.

Outcome pricing will not be simple. Both sides will need to agree on what counts as an outcome and how it is measured. Still, this is the kind of thing we build into an AI command centre, connecting AI activity to the business measures that matter rather than treating it as another software subscription.