Over 40% of agentic projects will be cancelled. How not to be one of them
Gartner and IBM point at the same three failure modes: cost, unclear value and weak controls. Each one has a design answer you can apply before the first line of code.
Two numbers from the past few weeks belong on every AI steering committee slide.
Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027. It also estimates that only about 130 of the thousands of vendors marketing "agentic AI" are offering the real thing. And an IBM survey of 2,000 CEOs found that only 25% of AI initiatives delivered the expected ROI over the last few years, and only 16% have scaled enterprise-wide.
Gartner names the causes plainly: escalating costs, unclear business value, inadequate risk controls. None of them are model problems. All three can be designed for.
1. Cost: price the case, not the project
Most pilots are budgeted as a project. Production is paid per case. We ask for one number early: what does a completed case cost today, in people time and systems? Then every design choice is measured against cost per case: which model, how many tool calls, when to stop and hand off. An agent that is 10% better and 5x more expensive is usually the wrong agent.
2. Value: pick a workflow with a stopwatch
"Use AI in claims" is not a use case. "Get claim status from the insurer for every open claim older than 30 days" is. Good first workflows share three traits:
- They are high-volume and repetitive.
- Today they are done by people following a written or unwritten script.
- The outcome is checkable: the status is either right or wrong.
If you cannot time the current process, you will not be able to prove the new one.
3. Controls: decide where people stay
Risk controls fail when they are bolted on after the demo. We decide up front:
- which actions the agent may take on its own,
- which require approval,
- which cases are routed straight to a person,
- what is logged, and who reviews it.
This turns "is it safe?" into a list someone can sign.
The pattern behind successful projects
IBM's survey also found 61% of CEOs are actively adopting AI agents and preparing to scale them. The organizations that get there will not be the ones with the most pilots. They will be the ones that treated the first pilot like a production system: measured, priced and governed from day one.
That is the bar we hold our own work to.
Sources
- Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, Gartner, June 25, 2025.
- IBM Study: CEOs Double Down on AI While Navigating Enterprise Hurdles, IBM, May 6, 2025.