Enterprise SaaS Vendors Rework Data Access for AI Agents, Shifting Pricing Models

Major enterprise software providers ServiceNow, SAP, and Workday are fundamentally re-architecting how external artificial intelligence (AI) agents access and interact with customer data, signaling a significant shift away from traditional per-seat licensing models.

This strategic realignment introduces new integration layers and policy frameworks that will redefine operational costs and data governance for large organizations leveraging AI.

ServiceNow initiated this change at its Knowledge 2026 event, unveiling Action Fabric, an integration layer through which external AI agents must route to access platform data and execute workflows.

COO Amit Zavery confirmed the company will meter this usage, charging customers based on the number of operations an AI agent completes via the layer, according to The Information.

Anthropic’s Claude is positioned as a launch partner, with Claude Cowork directly connecting into Action Fabric, establishing a precedent for third-party AI integration into core enterprise platforms.

JPMorgan analyst Mark Murphy characterized this charge as an effective “tax” on customers utilizing outside AI agents to interact with their own data within ServiceNow’s applications, highlighting the new financial implications for AI-driven automation.

In contrast, SAP has adopted a more restrictive posture, updating its application programming interface (API) policy in April to prohibit third-party AI agents from interacting with its systems outside of SAP-endorsed architectures.

The company specifically bars autonomous or generative AI systems from planning, selecting, or executing API call sequences without official SAP sanction, reserving such functionality for its own Joule Agents to maintain architectural integrity and control.

This policy drew immediate pushback from the DSAG user group and SAP’s partner ecosystem, impacting vendors who had built connectors for tools like Microsoft Copilot and Salesforce Einstein, as reported by The Register.

SAP CEO Christian Klein publicly stated the company intends to keep its architecture open, yet the policy language has not materially changed, creating ambiguity for partners and customers.

Workday CEO Aneel Bhusri has also indicated a similar strategic direction, noting the considerable financial upside in charging for agent access earlier this year, aligning with the industry trend of monetizing AI interactions.

DataDog has pursued a different operational constraint, capping how often customers can use its model context protocol server, which enables third-party AI agents to interact with its products, to 5,000 daily requests or 50,000 monthly, with room for exceptions, as noted by The Information.

The underlying tension stems from AI agents disrupting the long-standing Software as a Service (SaaS) per-seat pricing model, which traditionally correlated costs with human users and departments.

A single AI agent can generate thousands of API calls daily, creating significant infrastructure load and data processing demands without adding new human user licenses, thereby eroding predictable subscription revenues for vendors.

This shift forces enterprise software providers to recalibrate their revenue models from predictable subscription-based to consumption-based, behaving more like utility invoices that fluctuate with model activity rather than headcount.

Such a transition introduces new complexities for finance teams accustomed to stable, predictable IT expenditures, requiring enhanced cost management and forecasting capabilities for AI workloads.

ServiceNow’s executive vice president for its AI platform, Jon Sigler, framed Action Fabric as a “universal action layer” that third-party systems would inherently call into, asserting architectural control over the integration surface and data flow.

This approach aims to transform potential disruption into a new revenue stream by monetizing API interactions, effectively embedding a tollgate into the enterprise data access architecture.

AWS CEO Matt Garman has publicly warned that incumbent software vendors attempting to protect their installed base risk losing market share as AI redefines how enterprises procure and deploy technology.

These vendor strategies reflect a critical juncture where control over data access and API governance becomes paramount for maintaining competitive advantage and revenue streams in the AI-driven enterprise landscape.

Enterprise architects and CIOs must now rigorously assess the total cost of ownership for AI integration, considering these new metered access charges, potential vendor lock-in effects, and the operational burden of managing fluctuating consumption models.

The varying approaches—from ServiceNow’s metered tollgate to SAP’s stricter architectural enforcement—present a fragmented integration landscape that demands careful strategic planning to avoid unexpected costs and interoperability challenges.

Organizations should closely monitor the general availability of ServiceNow’s expanded AI Control Tower governance features in August, which will further shape how AI agent interactions are managed within their environments, particularly regarding data security and compliance.

The evolving policies underscore a critical need for robust internal governance frameworks to manage fluctuating AI consumption costs, ensure compliance with complex vendor terms, and strategically position the enterprise for future AI adoption without excessive architectural dependency.