Forrester has published research on agentic AI adoption among enterprise leaders, finding that only a small minority have delivered meaningful production impact.
The study points to a gap between reported adoption and operational results: three-quarters of enterprise leaders say they are adopting agentic AI.
The research suggests the technology has reached technical viability, with long-horizon agents now able to operate for days or months. Yet many organisations remain in pilot mode because they lack the orchestration maturity, governance structures and nonhuman identity controls needed to run such systems at scale.
That mismatch is central to the report's conclusions. Businesses are experimenting with agentic AI, but many are struggling to move beyond limited deployments because of uncertainty over returns, weak governance and confusion over platforms.
Production gap
This has created a widening maturity gap, with ambition running ahead of execution. In practice, organisations may be testing agents in isolated use cases without establishing the systems, oversight and co-ordination needed to support broader deployment.
Long-horizon agents are a particular focus. Vendors have shown that agents can persist over extended periods, but enterprises still face operational fragility when trying to manage them safely and reliably. Orchestration, identity management and context discipline remain key barriers.
Without stronger operating frameworks, the risks can build over time. Weak orchestration and governance can lead to agent sprawl, brittle systems and greater exposure to failures or misuse.
Many companies also lack central registries, control planes and automated guardrails, which the report identifies as necessary to oversee agent behaviour across an organisation. That shortfall is one reason initiatives often fail to progress beyond pilot programmes.
Risk focus
Security and oversight are emerging as central concerns. Nearly half of security decision-makers cite agentic AI as a concern, reflecting worries about how autonomous or semi-autonomous systems are identified, monitored and constrained in live environments.
The report argues that identity-driven oversight, runtime enforcement and continuous governance will define how organisations manage agentic risk. Companies that do not address those issues could face security weaknesses, compliance problems and reputational damage as adoption expands.
At the same time, the market around the technology is shifting. Software vendors, hyperscalers and systems integrators are redesigning products, pricing structures and services to support autonomous and multi-agent systems.
That change is increasing pressure on enterprise buyers to make long-term platform choices carefully. Decisions made now, the report suggests, will shape how easily organisations can scale agentic AI systems and how dependent they become on particular suppliers.
Platform shift
The findings indicate that the challenge is no longer whether the underlying technology works. The main issue is whether organisations can build the operational structures needed to use it in a controlled and measurable way.
For large companies, that means moving away from treating agentic AI as a narrow feature test and integrating it into wider operating models. The report emphasises standardisation, co-ordination across the business and governance that can be executed in practice rather than stated in principle.
Brian Hopkins, vice president and principal analyst at Forrester, outlined that view in comments accompanying the research.
"Agentic AI is no longer theoretical - it's technically real in 2026 - but most enterprises are still unprepared to operationalize it. The gap we see isn't about models or ambition; it's about orchestration, control, and trust. Companies that treat agentic AI as a feature experiment will stay stuck in pilots, while those that invest in agent‐native design, executable governance, and nonhuman identity will be the ones that actually capture value at scale," Hopkins said.