Enterprise AI Shifts to Outcome-Based Pricing and Bounded Domains

Enterprise AI is undergoing a significant shift in its pricing structure. Instead of paying for usage (e.g., "tokens"), clients will now pay for work accomplished and outcomes achieved. OpenAI's CFO has championed this change, proposing "useful intelligence per dollar" as the metric, emphasizing AI's value based on work done rather than mere consumption.

This evolution is a logical response to the decoupling of effort and value in AI. Industry observers note that an AI agent capable of resolving an issue with less consumption but still achieving the desired result is more valuable. For instance, customer service agent Sierra charges a fixed rate per resolved conversation, and Salesforce is reporting "Agentic Work Units" that measure completed tasks.

The user interface is becoming less critical as AI focuses on delivering outcomes. The application of the future is not the screen or interface, but the AI agent itself that executes tasks. Salesforce has launched "Claudeforce," integrating Anthropic's Claude model into its platform. This allows for sales tasks, such as pipeline reviews, to be completed without direct access to the Salesforce application.

AI agents often operate within existing organizational boundaries. Major systems like Salesforce for sales and SAP for procurement specialize in their respective domains. Consequently, a single universal agent controlling the entire enterprise presents a significant challenge. In the near term, teams that can build effective agents and achieve results within defined, high-value business areas are likely to succeed.

Outcome-based pricing may evolve into claiming a share of revenue. As AI becomes deeply embedded in workflows, it may also claim a portion of the generated profits. OpenAI has indicated that as AI enters fields like scientific research, drug discovery, and financial modeling, pricing could incorporate a share of the value created.