Validating agentic AI in the plant: autonomy level

An assistant that recommends and an agent that acts carry different risk, involve different approvers and clear different bars. Validate the level of autonomy, not only the use case.

By Harinderpal Hanspal on June 2026. Updated October 2026

When a system recommends, a person absorbs the risk of a bad call. When it acts on a production schedule, a maintenance plan or a supplier order, the buyer trusts it with outcomes they answer for personally: throughput, safety, uptime. The validation question changes from whether the buyer wants the outcome to how much autonomy the buyer will approve.

The question shifts from wanting the outcome to how much autonomy the buyer will approve Three steps of rising autonomy: a system that recommends, one that acts with a person approving, and one that acts on its own. A red arrow beneath marks that each step pulls in more of the buying committee and often a different budget. The question moves from does the buyer want the outcome to how much autonomy will the buyer approve Recommends A person absorbs the risk of a bad call Acts, a person approves Semiautonomous: humans keep final approval Acts on its own The buyer answers personally for throughput, safety, uptime More of the buying committee, and often a different budget
The question shifts from wanting the outcome to how much autonomy the buyer will approve

That trust does not grow with model accuracy. A buyer grants it at a specific level of autonomy, for a specific use case, at a specific site. Two vendors with one use case can face very different sales cycles, depending on whether they ask approval for advice or action. The second request pulls in more of the buying committee and often comes from a different budget.

Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027, citing escalating costs, unclear business value or inadequate risk controls. Each cause is a test nobody ran: the integration and autonomy assumptions, the operational outcome the buyer confirmed, the trust threshold and the liability model.

Gartner also expects semiautonomous agents to orchestrate 10% of key production, quality and maintenance use cases by 2030, up from 2%, with humans keeping final approval. Buyers adopt autonomy in steps. Price for the step the buyer is on.

Go deeper: Governing agents in production: what to ask before an agent acts, AI agent autonomy levels: no standard scale exists, so ask for the vendor's, Where to put the human in an AI agent's work: approve the write, not the draft, Human in the loop for AI outbound agents: review the send, and every route to it and how we score agentic AI sets out the four questions buyers ask.