Over 40 per cent of current agentic AI initiatives are projected to be halted by the close of 2027, Gartner warns, owing to spiralling costs, opaque business value and insufficient risk frameworks. Despite overwhelming industry enthusiasm, most ventures remain confined to pilot stages or proofs of concept and are driven more by hype than by tangible returns.
Gartner’s senior director analyst, Anushree Verma, cautions that many vendors are indulging in “agent washing”—rebranding chatbots or robotic process automation tools without genuine autonomous capabilities. According to the firm, only about 130 out of thousands of purported agentic AI providers deliver authentic agentic functionalities.
In a January 2025 survey of over 3,400 business leaders, Gartner uncovered that while 19 per cent have made substantial investments, a further 42 per cent have taken a cautious tack, and 8 per cent have yet to invest. Around one-third either remain undecided or are merely observing the technology’s progression.
Verma emphasises that current agentic AI lacks maturity—it struggles with executing complex, nuanced instructions or autonomously achieving strategic objectives—which hinders firms from realising meaningful ROI. The absence of robust risk oversight and clear business targets is also holding back widespread deployment.
Despite the setbacks, Gartner projects that by 2028 roughly 15 per cent of routine business decisions will be autonomously made by agentic AI, and approximately one-third of enterprise software will embed such systems—up dramatically from under 1 per cent today.
A parallel KPMG report reveals that full-scale deployments of agentic AI have tripled in recent months, with nearly one-third of organisations now live in production. However, challenges persist, including a shortage of specialised skills, resistance from workforces and the technical complexity of integrating autonomous systems into existing frameworks.
Steve Chase, KPMG vice‑chair for AI and digital innovation, asserts that corporate leaders recognise agentic AI less as a cost‑cutting tool and more as a gateway to fresh value and expansion. Yet a robust 46 per cent of firms cite efficiency and revenue enhancement as top priorities, while concern remains high over data privacy and regulatory compliance.
Academic scrutiny is also intensifying. A recent paper in arXiv highlights a critical misalignment between lab-based metrics and real-world outcomes. It notes an overreliance on technical benchmarks, with minimal attention to human-centric safety and economic impact; only 15 per cent of studies encompass both technical and human dimensions. This imbalance undermines productivity claims and may contribute to project terminations.
Industry leaders are calling for more calibrated evaluation frameworks, including blending technical performance with impact on operations, compliance, safety and workforce dynamics. Gartner echoes this recommendation, advising organisations to steer clear of retrofitting agents into outdated systems and, instead, redesign workflows with autonomy in mind to ensure scalability and measurable benefit.
On social platforms like LinkedIn, voices from IBM and Microsoft characterise the forecasted 40 per cent drop-out rate as a healthy indicator for emerging technologies. IBM’s Armand Ruiz argues that such rate is “not unusual in tech”, reflecting a natural weeding-out process. Microsoft’s Nitin Aggarwal concurs, describing it as “a HUGE success” relative to expectations.
The looming collapse rate underscores a broader cautionary signal for businesses amidst the AI fervour: agentic systems offer transformative potential—but only when supported by mature, goal‑oriented deployments with clear ROI metrics and comprehensive oversight. Only those that pass this rigor may reshape decision-making and automation in enterprise environments through the coming decade.