UK Finance AI Adoption Rises, Autonomy Lags
A Bank of England survey finds AI adoption widespread among responding financial firms, but only a small minority report broadly autonomous AI agents.
AI is becoming common across UK financial services, but giving software independent authority to act is a different decision. New industry survey results show how far adoption has progressed—and where institutions remain cautious.
What happened
The Bank of England and Financial Conduct Authority published their 2026 survey of AI in UK financial services. Among firms responding to the survey, 86% said they use AI, compared with 75% in the 2024 survey. The institutions cautioned that differing respondent groups and question scope limit direct comparisons between years.
Of the firms completing the detailed questionnaire, 80% reported having AI agents deployed and a further 14% expected to deploy them within the next year. However, most deployments were restricted in what they could do. Eighty percent reported agents that support human decision-making; 46% reported some agents performing actions within predefined constraints, and only 2% reported broadly autonomous agents.
The survey received responses from 196 firms, of which 105 completed the full questionnaire. The agent results relate to that detailed respondent group, not a census of every UK financial institution.
Why it matters
Banks, insurers and investment firms are using AI to reduce administrative work, find information and support internal decisions. Allowing agents to approve customer outcomes, place trades or move money introduces substantially higher operational, compliance and accountability risks.
This distinction creates a commercial opportunity for companies building model governance, audit trails, secure integrations, permissions and monitoring. A successful deployment is not necessarily the one that acts most independently; it may be the one that delivers measurable efficiency while maintaining control.
The bigger picture
The industry appears to be moving from experimentation toward broader integration, but trust and risk-management systems still set the limits of autonomy. For financial-services AI startups, reliability and controlled execution may matter as much as model intelligence.