Agentic AI’s strategic ascent is forcing executives to confront a difficult question: Is AI just improving existing processes, or is it reshaping how their business works?
The IBM Institute for Business Value (IBV), in collaboration with Oracle, finds that more than three-quarters of surveyed executives have invested primarily in using AI to optimize existing operations, even as 78% say that realizing the full benefit of agentic AI will require a new operating model. The study describes a widening split between organizations that chase incremental gains and those that redesign how decisions are made at enterprise scale.
Transformation-led organizations treat agentic AI as a catalyst for net-new capabilities rather than a more efficient version of the status quo. These organizations are building operating models where AI agents take an increasing share of routine and even complex decisions, while humans focus on the decisions that matter most.
Based on the study’s findings, 24% of executives say AI agents currently take independent action in their organizations. By 2027, 67% expect that to be true, with autonomous decision-making in processes and workflows projected to double compared to current levels. Even in highly regulated domains, executives anticipate agentic AI will automate nearly a third of risk and compliance operations by 2027.
New KPIs, Roles, Governance
The study also highlights a sharp divide in how organizations measure progress. While most AI investment remains process-focused, a smaller group of “transformational” organizations have started defining new key performance indicators (KPIs) to evaluate AI agents’ impact on business outcomes, beyond just efficiency. These leaders are using teams of specialized agents to monitor regulatory changes across jurisdictions, adjust workflows in real time, and mitigate risks before they materialize. They are also facing the realities of AI commoditization and moving quickly before early advantages erode.
However, the survey found that workforce readiness and trust emerge as critical constraints to AI adoption. Nearly half of organizations surveyed cite inadequate employee skills as a barrier to agentic AI, yet 79% of leaders say they need to protect and value human critical thinking as AI scales. With that in mind, new roles are appearing around AI orchestration, autonomous system auditing, and human oversight of AI-driven workflows. At the same time, 45% of executives point to a lack of visibility into AI decision-making as a major barrier. Leading organizations are responding with practices such as observability, logging, and systematic testing to make agentic AI traceable, auditable, and accountable.
Data infrastructure and governance can help ground this new operating model approach. Executives flag data privacy and security, integration complexity, and data quality as major challenges, showing how agentic AI is not only a technology problem but an enterprise-wide design problem. The report’s “Transformers” segment, representing 17% of the sample, is 32 times more likely to achieve top-tier performance across KPIs by 2026, linking superior outcomes directly to choices about operating model redesign.
What This Means for ERP Insiders
Agentic AI is an operating-model mandate. The shift from task automation to autonomous decision-making suggests that ERP-centered architectures will need to support AI agents as first-class actors. Processes, controls, and metrics need to be redesigned around continuous machine-led action rather than periodic human-led intervention.
Critical thinking and new roles control the field. As agentic AI moves closer to core finance, risk, and operations workflows, human roles will increasingly focus on orchestrating, challenging, and auditing AI decisions. This places new demands on ERP implementations to surface explanations, interventions, and governance checkpoints rather than just transactions.
Data, trust, and KPIs are competitive differentiators. The emphasis on data quality, transparency, and new AI-specific KPIs indicates that ERP programs that treat data governance and explainability as strategic concerns will be better positioned to harness agentic AI, while others remain trapped in narrow process-optimization use cases.
