Forrester's new framing argues enterprises should only buy agentic AI from vendors that have already deployed it internally, a procurement heuristic that quietly indicts the gap between vendor promises and production reality.
The standard for buying enterprise AI has quietly shifted. Instead of asking for roadmap slides and reference logos, procurement teams are starting to ask a sharper question: did the vendor actually use this product on itself first?
A Forrester analyst has named this emerging heuristic "Customer Zero." The principle is straightforward. Enterprises should only buy agentic AI from vendors that have already deployed the technology internally. The framing echoes the metaphor the analyst uses to make the case, "buy parachutes from the person who just landed safely, not the one still sketching designs behind a desk" (Forrester).
The reason this matters now is that the gap between what AI vendors claim to ship and what they actually run internally has become hard to ignore. In late 2025, after Forrester published The Forrester Wave: Digital Experience Platforms, Q4 2025, every vendor in the evaluation claimed an "agent orchestration" strategy. Across the analyst's customer interviews, only one had actually used a vendor's AI agents in production. Three years after ChatGPT went mainstream.
That gap is what Customer Zero is designed to close. The principle reframes the adoption problem as a procurement problem. Rather than waiting for enterprise case studies to accumulate, buyers should require vendors to demonstrate that they trust their own product enough to run it themselves.
What that would surface in practice is sobering. The vendor proof points that converged across those late-2025 conversations clustered around a single use case, generating marketing copy at scale. Marketing copy generation is not autonomous execution, the systems that take actions across a business, that the broader agentic AI narrative sells. It is content production at volume, dressed in the language of agents.
This is where the Customer Zero framing doubles as a market signal. If a vendor cannot show that its AI agents are doing real work inside its own business, the argument goes, an enterprise buyer should not assume those agents are ready to do real work inside theirs. That is a rational purchasing standard. It is also an implicit acknowledgment that the agentic AI category, three years after ChatGPT, is still mostly a slide deck.
Forrester's framing sits alongside its adjacent work on enterprise guardrails for agentic AI. The firm's AEGIS framework addresses how organizations that deploy autonomous systems should constrain them. Read together, the two pieces describe the same gap from opposite ends. AEGIS is the playbook for companies that have decided to trust agentic AI enough to operate it. Customer Zero is the upstream test of whether that trust is warranted in the first place.
Independent industry coverage supports the directional read. Reporting on agentic AI increasingly frames the category as moving from assistive tools into autonomous decision-making roles inside enterprises, a transition that makes the question of whether vendors use their own software more pointed, not less.
The Customer Zero heuristic is not, on its own, a measurement of enterprise AI adoption. It is a procurement disposition. But it is a telling one. When buyers start demanding that vendors eat their own cooking, the most informative thing is not how many vendors can pass the test. It is that the test has become necessary.
What to watch next: whether major enterprises publicly cite Customer Zero in their agentic AI procurement criteria, and whether any major vendor, under that pressure, publishes a candid account of what its own internal AI agents actually do. Either development would materially update the gap between agentic AI's marketing and its deployed footprint.