The SAP Value of AI Report 2026, produced with Oxford Economics from a survey of 2,600 business leaders across 13 countries, shows that enterprise AI has shifted from pilots to practical use: AI now supports about 30% of tasks in the average organization, up from 25% last year.
ROI and agentic AI expectations
Expectations for agentic AI ROI rose sharply, from 10% last year to 17% this year. Many organizations nonetheless believe AI could deliver much more value; SAP Chief AI Strategy Officer Sean Kask says the shortfall is driven more by gaps in strategy, data, and governance than by lack of access to the latest models. According to Kask, AI without contextual inputs — processes, data, or governance — tends to create activity without outcomes and can even introduce risk.
Fragmented investments and piecemeal adoption
Despite accelerating investment, more than half of organizations still fund AI in an ad hoc or piecemeal way. Only 17% report a strategic, holistic approach to prioritization, although that proportion has nearly doubled from 9% a year earlier.
This fragmentation can stem from board-level pressure to deploy AI quickly without a supporting strategy or sufficient AI literacy, producing scattered skunkworks projects. In other cases, lack of board attention lets employees bring their own tools and experiment locally.
Kask says many organic, disconnected AI initiatives struggle because of data quality. Even strategic efforts often operate in silos: they may have consistent data for a single use case but are not yet transforming entire business processes.
The report highlights a paradox: 69% of businesses say they are satisfied with their AI ROI because AI has proven it can generate returns, yet 67% remain unconvinced the technology is delivering its full potential — a reflection of both the value already realized and the challenges to scale it further.
Agents change the economics of enterprise AI
SAP has shipped more than 400 AI use cases across its portfolio, with more in development. Agents are the next expansion because they can plan and reason through multiple steps and tools to reach objectives, mirroring human processes.
Kask gives a concrete example: SAP released, in beta, an agent for accruals accounting that can reduce the monthly accounting time for a mid-size company from about 12 hours to two or three hours. Scaling such reductions across many processes creates substantial potential.
Overall AI ROI rose from 16% to 21% this year, and is projected to grow to $15.9 million in two years, even though only 3% of respondents say they are fully prepared for agentic AI.
Data quality is the main barrier to value
Preparing for agents requires two fundamentals: connecting agents to context-rich data and governing them at scale. Data quality and availability are now the top reason organizations say they are not getting more value from AI (73% of respondents), while 79% report occasional rework, delays, or backlogs caused by low-quality AI outputs.
The nature of the problem has shifted since classic deep learning. Foundation models reduce the need to find, extract, clean, and train bespoke models, but they make preserving business context even more important. Kask points out that extracting data from an ERP often strips out contextual semantics — the part that is most useful for generative AI.
SAP preserves context at scale with a knowledge graph in its cloud ERP that maps 452,000 ABAP tables and 7.3 million data fields. In SAP Business Data Cloud, data products present information such as invoices and suppliers consistently across SAP and non‑SAP systems without losing business meaning.
Governance: the under-recognized enterprise challenge
As AI embeds deeper into business processes, governance is emerging as the next major challenge. Only 12% of businesses say they are fully prepared to govern AI, while 69% acknowledge occasional to frequent use of unapproved shadow AI tools.
Kask warns that as companies roll out AI they often discover shadow agents that can access data they shouldn't or take inappropriate actions — prompting the question of how to audit them. SAP's AI Agent Hub responds by discovering and inventorying agents, large language models, and MCP servers; customers have already surfaced thousands of SAP and non‑SAP agents in their landscapes that they did not know existed. The hub adds lifecycle management, identity and access control, and performance monitoring. Kask likens the discipline to hiring: most firms would not onboard an employee without defining required access rights and permissions.
Governance extends beyond technology. Workforce transformation must progress alongside data work: nearly 80% agree that maximizing AI value requires more than technical upskilling, and 75% are already planning employee reskilling. Discussion is moving away from which jobs AI will replace and toward how people and AI can collaborate effectively, since agents still need human oversight, redesigned workflows, and stronger judgment.
Toward the Autonomous Enterprise
SAP frames these elements as the Autonomous Enterprise: connecting agents to contextually rich data and enterprise governance across functional silos, with Joule serving as a natural-language generative interface between people and systems.
"Realizing real value from AI is not going to be easy because it demands a new approach," Kask concluded. "It is ultimately a human change more than a technical one, because you can only achieve real value if agents, processes, and people work as one."
Further reading
Full findings are available in the SAP Value of AI Report 2026.
(Sponsored content: this report was produced by SAP in collaboration with Oxford Economics.)
