Half of enterprise AI projects stall at the pilot stage

TL;DR:

  • A Valliance survey of 1,000 senior leaders at Europe’s largest firms found 40% of AI initiatives stay pilots — rising to 48% at mature AI organisations.
  • UK firms spend an average £39.2m a year on AI, up 27%, yet strong returns remain rare.
  • Researchers blame “pilotitis” — poor metrics, low adoption and stalled scaling — rather than the technology itself.

Close to half of AI projects at large European businesses never make it past the pilot stage, a new Valliance study finds — and the firms most committed to AI are among the worst affected. Surveying 1,000 senior leaders, the consultancy found 40% of AI initiatives remain pilots by design, climbing to 48% at organisations with established AI programmes. UK firms are spending an average of £39.2m a year on AI, up 27% year on year, with IT and finance the biggest spenders, yet value stays elusive.

A pattern, not a one-off

The finding slots into a growing body of evidence. PwC, which in 2023 declared generative AI had “reached a tipping point”, now admits boardrooms meet AI returns with “silence”. Late in 2025, IBM found 60% of CEOs were still stuck experimenting, despite two-thirds expecting to scale a year earlier. A widely-cited 2025 MIT paper found fewer than one in ten firms had seen positive financial impact. Valliance diagnoses “pilotitis”: endless experiments killed by poor metrics, weak adoption and misaligned consultants. Stuck-in-pilot firms report success rates of 43% against 50% overall, and only 20% see strong ROI versus 76% among firms that scale.

This is the unglamorous counterweight to the week’s bullish AI headlines. It echoes the cautionary tale of Uber capping staff AI spend after a budget blowout, and Valliance cites a firm that accidentally spent £395m ($500m) in a month on Anthropic’s models without spend limits.

Looking forward

For UK businesses, the lesson is operational, not technological. Valliance’s advice — treat pilots as time-boxed learning cycles with clear gates to production, “measure obsessively and kill what doesn’t work quickly” — is mundane but rarely followed. With UK firms reporting uneven readiness to scale, the gap between AI spend and AI return is becoming the defining management question of 2026.

Share this article