IT leaders who spent the last two years building AI agent workflows into their enterprise stacks may be about to discover those workflows come with a new line item attached.
ServiceNow, SAP and Workday – three platforms that between them underpin the operations of a significant share of the Global 2000 – have each moved to assert control over how external AI agents interact with their systems.
For IT decision-makers, the practical consequences range from unexpected consumption charges to integrations that no longer work as intended.
The Access Problem
The issue starts with how AI agents actually function inside enterprise platforms. Unlike a human user who logs in, completes a task and logs out, an AI agent can execute thousands of API calls in a single session – querying data, triggering workflows, updating records – without ever appearing on a headcount report or a seat count.
That gap between how AI agents behave and how enterprise software is licensed is now being closed, unilaterally, by the vendors themselves.
ServiceNow moved first. At its Knowledge 2026 conference in Las Vegas, the company introduced Action Fabric – a mandatory intermediary layer through which any external AI agent must now pass to access data and workflows inside its platform.
According to The Information, COO Amit Zavery confirmed the company will charge for that access on a consumption basis, billing customers per operation an agent completes rather than per user on a contract.
For IT teams that have built ServiceNow integrations using third-party AI tools, that means existing workflows now carry a cost that wasn’t there when they were designed.
Anthropic’s Claude is the first external AI to be formally supported through the layer, suggesting ServiceNow has a clear view of which agents it expects to be running inside its environment.
What SAP’s Policy Change Actually Means
SAP’s approach creates a different kind of problem. An API policy update issued in April bars third-party AI agents from autonomously executing sequences of calls within its systems unless they do so through SAP-approved architectures, according to The Register.
SAP’s own Joule Agents are approved. Connectors that partners built for Microsoft Copilot and Salesforce Einstein are not.
For IT teams, the immediate question is whether integrations they are currently running – or that system integrators built on their behalf – are now in breach of SAP’s terms. The answer, based on the current policy language, is that many of them are.
SAP CEO Christian Klein has told investors the company has no intention of charging customers to access their own data, and that its architecture will remain open.
But Klein’s assurances and the written policy are not currently aligned, and until the language changes, IT and procurement teams have limited grounds to rely on the verbal commitment.
Workday’s Position
Workday has not yet announced a specific product or pricing change, but CEO Aneel Bhusri said earlier this year that monetising agent access represented meaningful financial upside for the company.
For organisations running Workday across HR and finance – functions where AI agent adoption is accelerating rapidly – that signal is worth taking seriously in forward planning.
What IT Leaders Should Do Now
The immediate priority is audit. IT teams need to know which AI agents are currently interacting with ServiceNow, SAP and Workday environments, through what mechanisms, and under what contractual terms.
Integrations built before these policy changes may now sit in a grey area at best and a breach of terms at worst.
The broader issue is one of negotiating leverage. Enterprise software contracts were not written with AI agent consumption in mind. As vendors move to meter or restrict agent access, the organisations best placed to push back will be those that understand exactly what usage they are generating – and can negotiate from a position of data rather than assumption.
