The reasons companies fall short are the same everywhere, and Forrester names them. ROI uncertainty keeps teams stuck in pilot mode. More than half report governance sprawl even after they adopted a formal AI framework. And nobody owns the agent across its life. Forrester is clear that none of this is the model's fault. It is a management gap.

The report names three moves that separate the companies getting results from the ones still chasing them.

Build the orchestration before you add more agents. Agents wired together with no shared registry and no clean hand-offs collapse into duplication and drift.
Redesign the actual work, not the tooling around it. An agent bolted onto a human-paced workflow saves a few tasks and not much else.
Give every agent a named owner. Its own credentials, least privilege, full logging, and one person accountable for it from day one.

The companies running agents at scale are not running more of them. They are running the right ones, with clear ownership, inside boundaries someone drew on purpose. Most pilots die because nobody answered those three questions before the build started.

How we build against this

We put a named owner on every agent before it ships, and the first week of any build goes to mapping the work, not the tooling. That is the whole point of an AI Audit, and it is why how the system gets governed is the first conversation we have, not a box we check at the end.

Source: Forrester, 2026 State of Agentic AI, published June 2026.