SAP AI ROI report finds gains, but data and governance gaps remain

SAP’s AI ROI report says enterprise AI is moving into day-to-day operations, with AI now supporting 30% of tasks in the average organization, up from 25% last year. The SAP Value of AI Report 2026, produced with Oxford Economics and based on a survey of 2,600 business leaders in 13 countries, also found that expected returns from agentic AI rose to 17% from 10% a year earlier.

The numbers give SAP a clear message for enterprise buyers: companies are seeing returns, but many are still far from turning AI into a coordinated operating layer. Sean Kask, SAP’s chief AI strategy officer, said the missing pieces are less about access to newer models and more about strategy, data and governance.

“AI has moved from experiment to execution, and that’s beginning to show real returns, but there’s still a long way to go,” Kask said. He added that AI without business context, including process knowledge, data and oversight, can produce activity without useful outcomes and can also create risk.

What does SAP say is holding back AI ROI?

SAP’s report points to fragmented adoption. More than half of organizations still fund AI in an ad hoc or piecemeal way, while 17% say they use a strategic, organization-wide approach to prioritization. That share is still low, though it has nearly doubled from 9% a year ago.

Kask said some companies pushed workers to adopt AI before setting a strategy or raising AI literacy, which can lead to scattered internal experiments. In other cases, limited board attention leaves employees bringing in their own tools and testing them without central controls. He said even more structured AI programs can remain stuck inside silos, with data that works for one use case but does not reshape a full business process.

That helps explain the report’s split finding: 69% of businesses said they are satisfied with AI ROI, while 67% said they are not convinced AI is delivering its full potential. SAP’s read is that early returns have made buyers more aware of how much more value might be available if deployment improves.

Agents raise the stakes for data and controls

SAP said it has shipped more than 400 AI use cases across its portfolio, with additional use cases in development. Kask said agents are the next expansion because they can plan and work through multiple steps and tools toward a defined objective.

As an example, Kask said SAP has released in beta an accruals accounting agent that can reduce a monthly task for a mid-size company from about 12 hours for an accountant to two or three hours. SAP did not provide broader customer benchmarks for that beta deployment.

The report said general AI ROI increased to 21% from 16% this year and should grow to $15.9 million in two years. At the same time, only 3% of respondents said they are fully prepared for that shift.

Data is the main constraint identified by respondents. According to SAP, 73% said data quality and availability are the top reasons they are not getting more value from AI, and 79% reported at least occasional rework, delays or backlogs caused by low-quality outputs.

Kask said foundation models reduce some of the work involved in finding, extracting, cleaning and training on data, but make business context more important. He said extracting data from an ERP system can strip away semantics that generative AI needs. SAP said its cloud ERP uses a knowledge graph covering 452,000 ABAP tables and 7.3 million data fields, while SAP Business Data Cloud is designed to keep information such as invoices and suppliers consistent across SAP and non-SAP systems.

Governance is becoming the next enterprise AI problem

The report found that only 12% of businesses say they are fully prepared to govern AI. SAP also said 69% acknowledge occasional to frequent use of unapproved shadow AI tools.

Kask said companies are finding “shadow agents” that may be able to access data or take actions they should not. SAP said its AI Agent Hub is intended to discover and inventory agents, large language models and MCP servers, then add lifecycle management, identity and access controls, and performance monitoring. The company said customers have already found thousands of SAP and non-SAP agents they did not know they had.

The report also frames AI as a workforce issue. Nearly 80% of respondents agreed that maximizing AI value requires more than technical upskilling, and 75% are planning to reskill employees. SAP’s broader pitch is what it calls the Autonomous Enterprise, connecting agents, enterprise data and governance across functions with Joule as a natural-language interface.

This story draws on original reporting from VentureBeat.