SAP’s AI‑Agent Pricing Model Threatens Unchecked Costs for Enterprise Users

SAP customers warned AI agents could put costs on autopilot

SAP’s latest Autonomous Enterprise platform promises to let businesses build AI agents that perform routine tasks, but the pricing model it introduces could quickly become a financial nightmare for users. Instead of the familiar per‑user or per‑seat licences, SAP will bill customers for each agent action – a vague metric that Gartner says may be defined in ways that dramatically increase the bill.

What happened?

Last week SAP announced that its new AI platform, built around the Joule Studio agent‑builder, will be sold using AI Units. These units are converted into a monetary charge based on the number of actions an agent performs in an “autonomous domain”. SAP says it will provide “Autonomous Domain Blueprints” with T‑shirt size guidance to help customers estimate usage, but the details remain scarce.

Legal and regulatory backdrop

The lack of transparency raises red flags under data‑protection legislation. The General Data Protection Regulation (GDPR) requires controllers to demonstrate purpose limitation and data minimisation – principles that are hard to uphold when an AI service can be billed for every automated decision or data‑processing step without clear limits. Likewise, the California Consumer Privacy Act (CCPA) obliges businesses to provide consumers with clear information about how their data is used and to allow them to opt‑out of certain automated profiling. If SAP agents trigger fees for each profiling action, enterprises may struggle to give users the required disclosures, risking enforcement actions and fines that can reach up to €20 million or 4 % of global turnover under GDPR.

Impact on users and companies

Stakeholder	Risk / Concern
Finance teams	Unpredictable operational expenditure; contracts allow SAP to change conversion factors during the term, potentially raising fees at renewal.
IT & DevOps	No clear metric for “action”; difficulty in budgeting and capacity planning for AI workloads.
Compliance officers	Harder to prove GDPR/CCPA compliance when every automated step may be billed, creating a conflict between cost control and legal obligations.
End‑users	Potential for hidden fees to be passed down, leading to higher product prices or reduced service quality.

Gartner’s senior principal analyst Victoria Rowan warns that if SAP defines an action as any API call, data retrieval, or even a rule‑evaluation, the number of billable events could explode. The report also notes that SAP’s contracts give the vendor the right to alter conversion factors, meaning a customer could see a price hike without renegotiating the agreement.

What changes are needed?

1. Clear definition of “action” – SAP must publish a concrete list of events that trigger billing (e.g., a completed transaction, a data‑write, a decision output). This aligns with GDPR’s requirement for transparent processing.
2. Cap on unit price escalation – Contracts should include a clause that fixes conversion rates for the duration of the agreement or caps any increase to a reasonable percentage.
3. Cost‑prediction tools – The promised “Autonomous Domain Blueprints” should be released as an interactive calculator, allowing enterprises to model worst‑case scenarios before committing.
4. Compliance audit hooks – SAP should expose logs that map each billed action to the underlying data record, enabling auditors to verify that processing respects data‑subject rights.
5. Price‑protection clauses – Customers should negotiate terms that prevent retroactive price changes and ensure any adjustments only apply at renewal, as SAP claims it will do.

How companies can protect themselves now

• Review existing SAP contracts for price‑protection language and for any clauses that allow unilateral conversion‑rate changes.
• Request the current AI Services List from the SAP Trust Center and request a copy of the conversion‑factor table.
• Run a pilot with a limited set of agents, instrumenting detailed usage logs to understand how many actions are generated per business process.
• Engage legal counsel to assess GDPR/CCPA implications of the agent‑driven processing, especially where personal data is involved.
• Consider alternative AI platforms that charge on a more predictable basis (e.g., flat‑rate subscription or per‑instance pricing) until SAP clarifies its model.

Outlook

SAP’s push for autonomous AI agents reflects a broader industry trend: vendors are moving from seat‑based licences to usage‑based pricing. While this can reward efficiency, it also opens the door to opaque cost structures that clash with both fiscal discipline and data‑protection law. Until SAP publishes concrete definitions and safeguards, enterprises should treat the new model as a high‑risk proposition and take proactive steps to lock in price‑stability and compliance assurances.

For further reading, see Gartner’s full report “First Take: SAP Moves to Higher‑Value‑Based AI Pricing, but Potential Cautions Remain” and SAP’s AI Units order form on the SAP Trust Center.