Gartner says 40% of agent projects are going into the bin

Gartner expects more than 40% of agentic AI projects to be cancelled by end of 2027. Three
reasons: costs that ran well past the business case, value that disappeared once someone
asked for numbers instead of a demo, and risk controls that were never built.
What’s not on that list is telling. The models didn’t break. The agents couldn’t reason. The
compute fell short. None of that came up. The tech did its job. Nobody built the structure around
it to make that job matter.
On Gartner’s Hype Cycle, AI agents are sitting right at the peak of inflated expectations. The
trough of disillusionment is expected throughout 2026. We’re in it.
They’ve also named a pattern they’re calling “agent washing” — vendors slapping the word
“agentic” on chatbots, RPA bots, and basic assistants that haven’t materially changed. Gartner’s
estimate is that only about 130 out of the thousands of vendors claiming agentic capabilities are
genuine. Everyone else is selling a rebrand to buyers who can’t tell the difference.
Longer term, Gartner expects 33% of enterprise software to include real agentic AI by 2028, up
from under 1% in 2024. Agents are going to matter. But deploying them without orchestration,
governance, or any clarity on who’s accountable when an agent makes a bad call on your
company’s behalf — that’s how you end up in the 40%.
Source: Gartner — Over 40% of Agentic AI Projects Will Be Canceled by 2027