Cognition pledges up to $10 million if AI coding tool fails to deliver

The AI startup says customers could receive up to $10 million in service credits if its AI coding agent, Devin, fails to deliver engineering value equivalent to what they pay for the product.

As businesses scrutinize the rising cost of artificial intelligence tools, US-based AI startup Cognition has introduced a guarantee aimed at addressing concerns over return on investment.

The company said customers using its AI software engineering agent, Devin, could receive up to $10 million in credits if the tool does not generate enough engineering value to justify its cost. The offer, described by Cognition as an “AI Productivity Guarantee,” was announced on the social platform X.

AI adoption has accelerated across corporate workplaces, but many organizations are increasingly questioning whether the productivity gains match the growing expense of deploying and scaling AI systems. Industry reports in recent months have highlighted tighter spending controls, usage limits, and closer monitoring of AI-related budgets at several large companies.

Guarantee tied to productivity

Cognition says the guarantee is designed to measure value based on work completed rather than traditional AI usage metrics such as tokens, prompts, or compute consumption.

According to the company, an internal evaluation system reviews Devin sessions and estimates how much engineering time was saved. The assessment considers whether the AI-generated output would be useful to a human engineer and how long a comparable task might have taken to complete manually.

Those estimated hours are then translated into a monetary value using a standard engineering rate. At the end of a customer’s annual contract period, Cognition compares that calculated value with the customer’s spending on the service.

If the estimated value falls below the amount paid, the company says it will provide additional usage credits, up to the stated $10 million limit.

Credits instead of cash

While the headline figure is attention-grabbing, the compensation is not a cash refund. Customers would receive credits that can be applied toward future use of Cognition’s products and services.

The arrangement has also drawn attention because the performance assessment is conducted by Cognition itself. The company’s internal system determines whether the AI delivered sufficient value, effectively making Cognition responsible for evaluating its own product’s performance.

Even so, the announcement underscores a broader shift in the AI market. As enterprises move from experimentation to large-scale deployment, vendors are increasingly being asked to demonstrate measurable business outcomes rather than simply offering access to advanced AI capabilities.

Cognition appears to be betting that stronger accountability around productivity could become a competitive differentiator as organizations demand clearer evidence that AI investments are producing tangible returns.