EY says 74% of UK consumers use AI but trust still lags

TL;DR:

  • EY research finds 74% of UK consumers have used AI in the past six months, but only 14% are comfortable relying on fully autonomous, agent-led AI systems.
  • Trust in institutions handling AI data remains limited: 43% of UK respondents trust companies to manage AI data effectively, 41% trust governments. 73% are concerned about AI being hacked or breached.
  • Resultsense view: this is the cleanest UK trust-vs-usage data point of 2026 so far. The implication for UK businesses is direct — adoption is no longer the constraint; demonstrable governance is. As Anthropic, OpenAI and Microsoft push agentic systems hard this quarter, UK customers are signalling they want oversight, not autonomy.

The findings are based on a 15,000-person, 15-country survey including 1,000 UK respondents. Matthew Ringelheim, EY UK and Ireland AI Leader, said adoption is “rapidly advancing, but trust is not keeping pace with technological capability”.

Where UK adoption is strongest

UK respondents reported using AI for customer support, route planning, health-related information, research, content generation and decision support. Half had consciously used AI as part of a health or wellness experience in the past six months; 35% had used it in financial activities — sectors where privacy and consistency carry greater weight.

The drivers cited mirror utilitarian, results-focused expectations: 59% pointed to improved response times, 52% to better value, 35% to reliability. Novelty was not a top driver — UK consumers judge AI “by visible outcomes”, not by the fact of it being AI.

The trust gap

The 14% comfort with fully autonomous agent-led AI — set against 74% recent usage — is the most consequential number. UK consumers are comfortable with AI that assists, but not with AI that acts independently. That gap is what Anthropic and OpenAI’s agentic-AI launches this week need to close before they can reach scale in UK consumer-facing deployments.

Frequent users were not necessarily more reassured: scepticism was often highest among educated white-collar workers who use AI regularly, suggesting familiarity sharpens concerns rather than reducing them. This matters because it inverts the usual “users grow into the technology” assumption — the more these users see, the more they want governance.

The skills layer

Only 23% of UK respondents said they had received significant training or education in AI. Ringelheim argued: “Workforce confidence — built through the right skills — will be decisive in turning AI momentum into long-term growth for the UK.” Training appears linked to confidence as well as practical use; the same population that uses AI daily is also the population most uneasy about its boundaries.

UK relevance

Three takeaways. First, UK consumer-facing businesses deploying AI should foreground human oversight in their UX — that is what 86% of users are signalling they want. Second, autonomous agent products will likely need a “supervised mode” for the UK market for the next 12–24 months even if their US equivalents go fully agentic. Third, the skills gap is the layer where genuine market differentiation is now possible — UK firms that invest in customer-facing AI literacy will translate the trust gap into adoption advantage faster than peers.

Looking forward

The next round of EY data (typically biannual) is the natural follow-up to track. The leading indicator to watch is whether the autonomous-AI comfort number moves above 20% — that is a plausible inflection where agentic deployment in UK retail, banking and insurance becomes commercially viable at scale. Below that threshold, vendors should design for human-in-the-loop, even where the underlying capability allows full autonomy.

Share this article