Enterprises Adopt Agentic AI, Most Projects Stuck in Pilots

Forrester reports that 75% of enterprise leaders are adopting agentic AI, but most initiatives remain in pilot stage and few organisations use agents beyond basic chatbot-style implementations.

The research found agentic AI is technically viable in 2026, but many companies lack the infrastructure and governance to scale projects. Analysts noted firms often treat agents as chat features rather than components of a distributed system that require shared data, registries and routing to operate inside existing IT environments.

“Agentic AI is no longer theoretical – it’s technically real in 2026 – but most enterprises are still unprepared to operationalize it,” wrote Brian Hopkins, vice president and analyst at Forrester.

Forrester identified three main barriers: unclear returns on investment, orchestration challenges and trust. The consultancy said ROI uncertainty keeps projects in pilot mode and that the extra cost of auditing agent actions is a significant burden, which it described as a “trust tax.” Regulated sectors such as finance and healthcare face strict requirements to log and justify autonomous decisions.

Platform uncertainty also slows adoption. Teams are weighing whether to use a SaaS agent platform, hire a systems integrator to build a system, or develop an in-house solution. That indecision, together with auditing and governance costs, often limits deployments to narrow efficiency gains.

The report highlighted security and risk concerns and recommended governance and identity policies be enforced as code. Forrester advised treating each agent as a nonhuman identity with unique credentials, least-privilege permissions, full logging and a named owner responsible for oversight.

To move beyond pilots, the consultancy advised investing in data infrastructure, orchestration layers and shared registries so agents can interact reliably with business systems. It also recommended redesigning high-friction workflows, reorganising roles and approvals around autonomy, and scaling agents in stages after controls are in place.

A separate survey earlier this year found roughly half of organisations remain at the pilot stage, though many plan to increase investment. Forrester concluded that organisations building underlying systems for auditability and automated governance are likeliest to scale agent deployments.