AI with Agency Faces Disillusionment: 40% of Projects Abandoned by 2027

Gartner has just published a forecast that hits the agentic AI market like a cold shower: more than 40% of development projects will be abandoned before the end of 2027.

While it may seem pessimistic, it fairly reflects the feedback from CIOs on the ground (see our box on French CIOs). Gartner identifies three main factors to explain the situation: rising development and deployment costs, the difficulty in demonstrating clear commercial value, and the inadequacy of risk controls implemented by companies.

A Phenomenon of “Agent Washing”

Most agentic AI projects remain at the experimental or PoC stage, revealing the real complexity and costs associated with deploying these agents at scale.

According to a Gartner survey conducted in January 2025 among 3,412 webinar participants, only 19% of organizations have made meaningful investments in agentic AI, while 42% have opted for prudent investments. 8% have made no investment and 31% remain watchful or undecided.

The market is also marked by the phenomenon of “agent washing,” i.e. the rebranding of existing solutions without genuine agentic capability. Gartner estimates that barely 130 of the thousands of vendors claiming agentic AI are truly legitimate. And most current agentic AI proposals lack added value or notable ROI. Many use cases touted as agentic could be served by less sophisticated solutions.

Medium-Term Transformation Prospects

Despite these initial obstacles, Gartner believes that agentic AI represents a major advance, capable of boosting resource efficiency, automating complex tasks, and driving innovation beyond what traditional bots can achieve. By 2028, the firm expects that 15% of daily business decisions will be made autonomously thanks to agentic AI (versus 0% in 2024), and that 33% of enterprise software will incorporate these agents (versus less than 1% in 2024).