Organisations are continuing to invest in artificial intelligence under pressure to improve productivity and reduce costs, but new Gartner research suggests many initiatives are still falling short of meaningful returns. The findings, highlighted by Basware, show that only 28% of AI use cases in infrastructure and operations fully meet ROI expectations, while around 20% fail outright.
The survey points to a wider execution problem rather than a purely technical one. More than 57% of organisations reported at least one failed AI initiative, indicating that many programmes are being pushed forward without clear use cases, realistic success criteria, or the operational foundations needed to support them. Gartner’s findings suggest that weak governance, poor data quality, and misplaced expectations around automation remain central causes of failure.
That matters because AI is often being deployed into functions where the business case is still vague. The research suggests organisations are still trying to force AI into problems it is not designed to solve, often in pursuit of fast cost savings or broad transformation narratives. In practice, the better results tend to come from narrower, more mature environments where workflows are already well defined and data is easier to manage.
Gartner found that 33% of successful leaders embed AI directly into existing systems and workflows, rather than treating it as a standalone initiative. More than 53% of successful deployments are concentrated in established areas such as IT service management and cloud operations, where processes are repeatable and the operational value is already understood. The data also shows that 26% of successful leaders report full executive backing, while 25% cite strong cross-functional collaboration as a key factor.
Jason Kurtz, CEO of Basware, said: “Enterprises are right to rethink how and where they deploy AI as organisations often try to force AI into processes where it simply doesn’t add value, leading to overly ambitious projects, unclear outcomes, and disappointing returns. Recent research shows that many initiatives fail not because of the technology itself, but because they are poorly scoped and driven by expectation rather than clear use cases.”
For finance leaders, that argument points towards more controlled starting points. High-volume, standardised, data-rich workflows such as invoice processing and payment operations are easier to automate, easier to measure, and less dependent on speculative assumptions about future value. The underlying message from both Gartner’s data and Basware’s response is that ROI is more likely to emerge where governance is already strong and outcomes can be tested quickly.
Kurtz added: “Ultimately, organisations that prioritise targeted, ROI-driven adoption starting small, scaling what works, and aligning technology decisions with real business outcomes will be best positioned to move beyond the hype and unlock sustainable value from AI.”
The survey reinforces a familiar pattern in enterprise technology investment. AI projects are most likely to progress when companies begin with tightly scoped problems, defined ownership, clean underlying data, and leadership support. Where those conditions are missing, the technology may still work, but the business case often does not.
