Enterprise AI has moved from experimentation to execution, and that shift is beginning to show real returns. According to the SAP Value of AI Report 2026, AI now supports nearly one-third of all tasks in the average organization, rising to 30% from 25% last year. For business leaders, this means the focus is no longer just on whether to adopt AI, but on how to scale it effectively without creating fragmented 'skunkworks' efforts. (source)
The transition is already showing measurable gains, but the path to scalable ROI is becoming more complex. While 69% of businesses report satisfaction with their AI ROI because they have proven the technology can generate returns, a significant gap remains. Roughly 67% of organizations remain unconvinced that the technology is delivering its full potential. This skepticism stems from a realization that scaling requires more than just access to a model; it requires a strategy for data and governance.
The shift toward agentic AI
Agents represent the next expansion of enterprise AI because they can plan and reason through multiple steps to reach an objective. Unlike basic chatbots, agents mirror how people and processes work. For example, SAP has released a beta agent for accruals accounting—a task that typically takes a mid-size company's accountant about 12 hours a month—which reduces the workload to just two or three hours.
While general AI ROI grew from 16% to 21% this year, only 3% of organizations say they are fully prepared for this growth. To capture the projected $52 ft in value over the next two years, companies must move past piecemeal investments. Currently, only 17% of organizations report a strategic, holistic approach to AI prioritization, nearly double the 9% reported a year ago.
Overcoming the data quality barrier
Data quality is currently the number-one reason organizations aren't seeing more value from AI. About 73% of respondents cited data quality and availability as the primary hurdle, with 79% reporting rework or delays caused by low-quality outputs. In the era of foundation models, the challenge has shifted: the goal is no longer just finding data, but preserving business context.
When data is extracted from systems like an ERP, it often loses the semantics that make it useful for generative AI. To solve this, SAP utilizes a knowledge graph in its cloud ERP that maps 452,000 ABAP tables and 7.3 million data fields. This allows data products to maintain their business meaning across both SAP and non-SAP systems, ensuring that agents have the context they need to perform accurately.
Managing the 'shadow AI' risk
As AI becomes deeply embedded, governance is emerging as the critical enterprise challenge. Currently, only 12% of businesses say they are fully prepared to govern AI, even though 69% acknowledge the frequent use of unapproved "shadow AI" tools. These are agents that might access sensitive data or take unauthorized actions without oversight.
To manage this, organizations need to treat AI governance with the same discipline as hiring a new employee. You wouldn't onboard a staff member without defining their access rights; you shouldn't deploy an agent without the same controls. Effective governance requires an inventory of all active agents and LLMs, paired with lifecycle management and identity access control.
Success in the next phase of AI adoption isn't just a technical hurdle—it is a human one. Maximizing value requires more than technical upskilling; it requires redesigning workflows and ensuring that people and agents work as one. To start, audit your current landscape to identify shadow agents and prioritize a data strategy that preserves context over sheer volume. Read more: Prentis raises $100M for AI agents. Here is what to expect from your office workflow.
