Gartner Predicts 40% AI Agent Projects Will Fail Due to Poor Returns

More than 40% of agentic artificial intelligence projects will be cancelled by the end of 2027 due to rising costs and unclear business value, according to a Gartner report published on June 27, 2025.

Hype, High Costs, and the Reality Check for AI Agents

Gartner analysts warn that many early-stage agentic AI experiments will be abandoned before they deliver meaningful returns. These systems are designed to autonomously set and achieve complex business objectives without human intervention.

The research finds that hype and “agent washing” have driven thousands of vendor offerings. Agent washing refers to companies relabeling basic AI assistants or chatbots as fully autonomous agents. However, only about 130 products can genuinely execute tasks without human help.

Tech companies, including Salesforce and Oracle, invested billions into AI agents to streamline operations and cut costs. Many of these projects remain at the proof-of-concept stage and fail to scale in live environments. Gartner predicts that unclear return on investment (ROI) and mounting implementation expenses will force organizations to cancel over four in ten agentic AI efforts by 2027.

Related: Elon Musk’s xAI Burns Over $1B Monthly Amid Soaring AI Development Costs

Anushree Verma, senior director analyst at Gartner, said most initiatives today are driven more by marketing pressure than by demonstrated outcomes. 

Most agentic AI propositions lack significant value or return on investment, as current models do not have the maturity and agency to autonomously achieve complex business goals or follow nuanced instructions over time,

she notes.

Future Outlook for Agentic AI

Despite the expected cancellations, Gartner also forecasts rapid growth in genuine agentic AI adoption. By 2028, at least 15% of routine business decisions will be made autonomously by these systems, up from virtually zero in 2024. They also predict that 33% of enterprise software applications would use agentic AI in three years, up from less than 1% today. As the tech improves, businesses are expected to move from small experiments to real-world uses that actually deliver results.

Read on: Tether CEO Says Global AI Agents Will Use Bitcoin and USDT by 2040

Agentic AI refers to software agents capable of identifying objectives, planning actions, and executing tasks with minimal human oversight. Current technologies often fall short of handling evolving, multifaceted business processes without extensive guidance. The concept underpins aspirations for fully automated workflows in areas from customer service to supply-chain management.