Press
UnlikelyAI
City AM covered our AI Trust Ceiling research this month alongside a question that cuts to the heart of the UK’s AI moment: are businesses actually ready to fully integrate AI? Adoption is not the constraint, as 99% of senior decision-makers are already using AI. What is constraining value is trust and it is not a sentiment, it has a measurable cost.
The gap between policy and practice
The UK sights on leading the G7 in AI adoption, standing up to pledge £2.5bn in AI investment. The UK has the talent, the financial infrastructure, and the regulated industries that stand to gain most from trustworthy AI. The data from our AI Trust Ceiling research, cited by City AM, points to a problem that sits between the ambition and the outcome: across larger businesses in the UK, the cost of employees verifying AI outputs rather than acting on them is estimated at £29bn a year. That is not a skills gap or a funding gap, instead it is a trust gap, and adoption alone will not close it.
£29bn
Estimated annual cost of AI verification burden across UK businesses
57%
Of leaders report frustration when validating AI outputs
More adoption does not solve this
The assumption built into most AI growth strategies, including the government’s, is that adoption is the bottleneck. Get more organisations using AI faster, and productivity follows. But our research across 1,000 senior decision-makers in regulated industries shows the bottleneck is not adoption, it is reliable. Senior leaders are already using AI and they are spending almost as much time checking it as using it because they cannot trust it to be right without watching it. Accelerating adoption of systems people cannot rely on does not close the productivity gap, it widens it.
“”Rapid adoption without the right foundations carries real risks. This investment must be matched with the skills, infrastructure and guidance organisations need to deploy AI securely and effectively.” — Richard Thompson, CEO, ANS — as quoted in City AM”
What trust actually requires
The leaders in our research were clear about what would change their behaviour. Thirty per cent said they would increase AI investment immediately if accuracy and explainability could be guaranteed. They are waiting for AI they can actually rely on, which is systems that can show their reasoning, produce auditable outputs, and operate within the logical boundaries that regulated industries demand. That is not what large language models were built to do. It is exactly what neurosymbolic AI is designed for.
The UK will not lead the G7 in AI by being the fastest to adopt tools that nobody fully trusts. It will lead by being the first to deploy AI that gives organisations a reason to stop checking and start relying.
Not ready to speak to us? Start here.
See the sector by sector breakdown
