NEWS

Chunk of agentic AI projects doomed to fail

Gartner forecast more than 40 per cent of agentic AI projects will be scrapped by the end of 2027, citing high costs, limited business value and insufficient risk controls.

The research company noted that most current initiatives are early stage pilots or proofs of concept, with many organisations underestimating the complexity of scaling agentic AI systems.

Anushree Verma, senior director analyst at Gartner, noted that projects are often “driven by hype and are often misapplied”. “This can blind organizations to the real cost and complexity of deploying AI agents at scale, stalling projects from moving into production.”

The company also also highlighted the rise of “agent washing”, where vendors rebrand existing technologies such as robotic process automation (RPA) tools, chatbots and AI assistants, as agentic AI without delivering true agentic capabilities. It estimated that only around 130 vendors in the space populated by thousands of agentic AI providers offer genuine solutions.

Indeed, recent Gartner poll of webinar attendees found that only 19 per cent of respondents claimed their company had made significant investments in agentic AI, while 42 per cent reported conservative investments. Meanwhile, more than 30 per cent were either unsure or had adopted a “wait and see” approach.

However, despite early challenges, Gartner maintains that the technology represents a major shift in market opportunity, particularly for enterprise automation. The company predicted that 15 percent of daily work decisions will be handled autonomously by AI agents by 2028, marking a rise from zero in 2024. Additionally, 33 per cent of enterprise applications are expected to embed agentic AI by 2028, up from less than 1 per cent last year.

Looking ahead, Gartner advised organisations to prioritise use cases with clear ROI and to avoid the complexities of adding AI agents into legacy systems. “To get real value from agentic AI, organisations must focus on enterprise productivity, rather than just individual task augmentation,” Verma concluded.

Back to Homepage