Only 28 per cent of AI use cases in infrastructure and operations (I&O) fully succeed, while 20 per cent fail outright, according to a survey by consultancy Gartner.

Gartner surveyed 782 I&O leaders and found that 57 per cent reported at least one failure to integrate AI into operations. Some 77 per cent said AI succeeded in at least one integration.

Melanie Freeze, director of research at Gartner, wrote in a blog post that the high failure rate is largely driven by unrealistic expectations of the technology. She said that for the 57 per cent who reported at least one AI integration failure, many of the issues came from leaders expecting AI to immediately automate complex tasks, cut costs or fix long-standing operational issues.

When these “overly ambitious” goals are not met quickly, confidence drops and projects are unable to move forward. Freeze gave auto-remediation, self-healing infrastructure and agent-led management of workflows as examples of tasks that AI was not yet equipped to deal with, despite expectations.

Other major failure points were skills gaps and poor data quality, both noted as points of failure by 38 per cent of respondents.

In contrast, Freeze said, successful projects integrated AI into existing systems with clear business cases for its adoption. The greatest successes have come from using generative AI in IT service management, where 53 per cent of reported successes occurred. If AI becomes part of day-to-day operations, Freeze said, it boosts adoption and creates visible impact within organisations.

The key to determining which AI projects to prioritise is to ensure each use case is linked to a business goal, and to track their collective impact on I&O and business outcomes.

From there, Freeze said, “I&O leaders can work alongside their chief information officers, data and analytics, security, legal and finance stakeholders to assess each use case for feasibility, risk, cost and expected business impact.”

---