UK IT leaders report surprise AI bills and governance gaps

A poll of UK-based IT leaders by work management platform Asana found more than 80% experienced unexpected cost increases from AI over the past 12 months. Fifty-three percent reported an AI tool or agent took an action in the same period that produced financial, legal, reputational or compliance harm. Just over 60% of respondents said they feel highly or fully accountable for AI-driven business results even as use expands beyond formal IT control.

Respondents estimated that about three in ten new AI tools and agents introduced in the last year were adopted without formal IT review or approval. One in four employees said they often use AI tools that are not formally approved by their organization, and 38% regularly use personal AI accounts for work tasks. Nearly half of UK workers said they do not know how their emails, documents and tasks are used to train or improve AI systems.

IT leaders linked unexpected costs to rapid uptake of agent-style AI and heavy use of language-model compute. Some companies, including Uber, reported very large bills after internal pushes to increase AI use. Organisations have responded by capping token use, tightening approval processes and changing how they measure productivity gains tied to AI.

Survey responses also flagged execution and context problems. Forty-six percent of IT leaders said AI initiatives often or always fail or stall because systems lack complete organisational context. Thirty-seven percent of knowledge workers said they spend at least 30 minutes a day fixing or reworking AI outputs. Forty percent of workers reported needing to consult three or more tools to gather the information required for tasks, and roughly two-thirds said they must re-explain context to AI tools at least some of the time.

IT teams said costs are hard to track because adoption happens across departments outside traditional procurement and governance chains. The survey found some firms have introduced token limits and stricter approval rules to try to reduce unplanned spending and increase visibility.

Christina Francis, head of UKI and Northern Europe at Asana, commented: “Most organisations are past asking whether their people will use AI — they already are. The challenge now is turning that into measurable business value, without losing the governance and visibility needed to manage risk.” She said AI performs best when it has access to goals, decisions and workflows that surround work and when employees understand how their output feeds AI systems.