Forrester has published its annual State of Agentic AI report, and the headline is sobering. Three-quarters of enterprise leaders say they are adopting agentic AI. Yet only a small minority have it running in meaningful production beyond "agentish" chatbots, and true scaled multi-agent systems are rarer still. In Forrester's framing, the story of 2026 is the gap between the chase and the catch.
For enterprise buyers and suppliers, the report is a useful reality check. The technology is here. Long-horizon agents that run for hours, days or even months are no longer theoretical. OpenAI has operated internal software development workflows with agents for months. Cursor has deployed long-running coding agents. Anthropic has demonstrated multiday research agents. The proofs are in. But turning those proofs into reliable, governed production systems is proving far harder than most organisations expected.
What Forrester found
The report, based on survey data and direct conversations with the architects building enterprise agentic systems, identifies four core barriers that keep most organisations stuck in pilot mode.
ROI uncertainty. Most companies cannot justify moving beyond narrow efficiency gains because the return on investment is unclear. Pilots that save a few hours per week are easy to fund. Production deployments that reshape workflows and require new governance, infrastructure and talent are not.
Governance gaps. More than half of enterprises report agentic sprawl even after adopting the NIST AI Risk Management Framework. A policy document cannot control an autonomous system that invokes tools, escalates privileges and operates beyond real-time human oversight. The result is uncoordinated deployments, duplicated effort and drift.
Platform confusion. Teams are frozen between options: off-the-shelf SaaS agents, systems integrator builds, or custom in-house development. Each path has different trade-offs in cost, control, time to value and lock-in, and most organisations lack the internal expertise to evaluate them.
Trust and risk management. In Forrester's 2026 Security Survey, 49% of security decision-makers named agentic AI as a concern. The threats are structurally different from traditional software risks. Agents can impersonate each other. Their populations grow faster than identity systems can track. When coordination breaks down, a small misjudgment can become an outage.
The knowledge readiness problem
Forrester's findings are echoed by a separate Sinequa survey of 740 senior executives at companies generating between $1 billion and $20 billion in annual revenue. Only 10% of respondents have deployed true multi-agent systems with collaborative, autonomous capabilities. The majority, 70.7%, are operating assistive AI or below: sophisticated knowledge-retrieval tools that cannot independently pursue goals or take actions.
The Sinequa data also reveals an "agent-washing" crisis. Eighty-four per cent of enterprise leaders encounter products rebranded as agentic AI during evaluations, and 87.5% say this has damaged trust in AI broadly. Nearly a third report it has made it harder to secure budget for legitimate projects. Trust, not technology, is the defining barrier.
Behind the trust problem sits a knowledge readiness gap. An agent is only as good as the information it can access. Thirty-eight per cent of leaders struggle with data that does not update, and 31.4% are hampered by data silos. Organisations with true agentic deployments are nearly twice as likely to have sophisticated, enterprise-scale knowledge architectures in place.
What the leaders are doing differently
Forrester highlights Bank of New York as an example of a regulated enterprise that is further ahead. BNY has not captured the full value of agentic promises yet, but it does have something most organisations lack: a workforce ready to manage highly autonomous agents inside a tightly regulated business. That readiness is the differentiator.
The companies pulling ahead are not the ones with the most agents. They are the ones building the infrastructure those agents need. Forrester recommends three moves for organisations that want to close the gap.
Invest in orchestration before adding agents. Shared registries, hand-off patterns and context discipline are essential. Long-running agents behave like distributed systems, and distributed systems demand orchestration. Stitch ten agents together without shared context and coordination collapses into duplication and drift.
Redesign the work, not just the tooling. Agents bolted onto human-paced legacy workflows produce task savings, not step-change value. The organisations seeing results pick a few high-friction workflows and rebuild the roles, approvals and hand-offs around autonomy.
Treat every agent as a governed identity. Each agent needs unique credentials, least-privilege access, full logging and a named owner who manages its lifecycle. Unowned autonomy is a risk no regulated enterprise should accept.
What this means for enterprise buyers
The core message is that agentic AI adoption is not a procurement decision. It is a transformation decision. Buying an agent platform without the orchestration layer, knowledge infrastructure, governance framework and workforce readiness to support it will leave most organisations with expensive chatbots.
Buyers should evaluate agentic solutions with the same rigour they apply to any enterprise system: clear use cases, measurable outcomes, defined ownership, audit trails and rollback paths. The vendors that can demonstrate production deployments in regulated environments, not just demos, will separate themselves from the agent-washing crowd.
What this means for suppliers
For suppliers building and selling agentic solutions, the challenge is shifting from "can our agent do X?" to "can our agent do X reliably, governably and at scale inside a regulated enterprise?" Procurement teams are beginning to ask harder questions about runtime controls, identity management, observability, cost predictability and evidence of real deployments.
Suppliers that can show production references, explain their governance model, offer transparent pricing and demonstrate integration into existing enterprise stacks will have a significant advantage. The window for selling on capability alone is closing.
The Agentic Expo angle
Agentic Expo is built around market-ready AI agents. Forrester's research shows that market-ready means something different in mid-2026 than it did a year ago. It is no longer enough for an agent to demonstrate capability. It must also demonstrate reliability, governance, cost predictability and evidence of production deployment in real enterprise environments.
The gap between the three-quarters of enterprises chasing agentic AI and the minority actually catching it is where the opportunity lives. Buyers need better ways to evaluate what is real versus what is agent-washed. Suppliers need platforms where production readiness, not just marketing claims, is the standard. Agentic Expo exists to close that gap.