As Enterprise AI Starts Charging For Work, Owning The Outcomes Becomes The Contest

Enterprise AI is evolving beyond simply charging for usage to a focus on the cost of delivering successful tasks. OpenAI's CFO has shifted the metric to "useful intelligence per dollar", suggesting AI should be judged by the work it accomplishes rather than the number of tokens used. By 2026, enterprises will favor routine tasks for cheap models while reserving top-tier models for justifying the cost.

This shift is driven by the fact that effort and value do not always align. Agent products demonstrate that agents achieving a tenth of deals with a hundredth of tokens can still be more valuable than less efficient alternatives. Pricing is shifting towards a value-based model, meaning a seller constructs the price based on the outcome achieved, not the usage alone.

This outcome-based pricing creates a new product - the agent - that is no longer just the interface, but the skills, documentation and rules that enable the system to extract value. The harness around the agent becomes the valuable object, with Salesforce's AIforce being a prime example. This move toward outcome pricing and agent-based solutions could transform enterprise AI, turning the traditional software apocalypse narrative into a more sustainable reality.