Executives describe consumption-based pricing and broader system access as part of the company’s agent strategy.

The Register reports that at Workday’s annual conference, executives told investors AI agents could expand its market “beyond the software budgets into the labor and services spend,” where customers are “spending tens of millions of dollars on outsourcing,” in the words of Chief AI Officer Joel Hellermark. He claimed self-service AI agents had increased resolution rates for common tasks by 25 percent. CEO Aneel Bhusri said he expects customers to pay for AI agents by consumption, so bills grow as usage grows, and said AI helps Workday grow without a matching rise in headcount.

The page adds that Workday announced broader “headless” access this week, letting AI agents read data and act in its HR and finance systems through Model Context Protocol (MCP) tools, subject to its permissions, approvals and audit trails. The 25% figure and the growth claims are executives’ statements to investors; the article gives no independent measurement.

In plain English

“Usage-based billing” means paying per amount used rather than per user. “Headless” access means software or agents use a product without its screen. “MCP” is a standard way for AI agents to connect to tools and data.

Why it matters

It is an explicit pricing shift from per-seat software to metered agents, which changes how a buyer budgets, a point that applies to marketing tools that add agents too.

The shift in pricing logic is the central business story. A software subscription is usually compared with other software spending. A service sold as completing work invites comparison with the cost of delivering that work through employees or outside providers. Those are different purchasing discussions.

For a buyer, a usage-based bill also raises a measurement question: what exactly is being counted? A request, a workflow step and a successfully resolved case are not equivalent units of value. The reported ambition to charge for consumption does not by itself tell customers which unit will appear in a contract.

A useful pilot would connect the bill to completed outcomes and the review effort those outcomes require. If a task is counted as resolved but later reopened, the initial completion rate can overstate the benefit. Similarly, a higher volume of automated actions can increase charges without creating a proportionate improvement in service.

Access to HR and finance systems makes permissions particularly consequential. The ability to read data should be distinguished from authority to change records or approve transactions. Existing approvals and audit trails remain meaningful parts of the workflow. The executives' statements describe a commercial direction; whether it produces savings for a particular customer depends on the actual work, pricing and controls.

Sources