Three years into the enterprise AI era, most organisations are using AI and almost none of them fully trust it. That gap, between use and reliance, is where the value disappears. This report names it, measures it, and sets out what it will take to break through.

Based on a national study of 1,000 senior UK decision-makers across five regulated industries, the 2026 AI Trust Report is the first piece of primary research to quantify exactly what the trust deficit is costing in ROI, in productivity, and in the people doing the work.

Why this research values

The experiment phase is over. AI adoption is up across every regulated sector, but the productivity gains that were supposed to follow have not arrived, and boards are running out of patience for vague answers.

The problem is that the conversation has been shaped by vendor case studies, analyst projections, and anecdotes but not by the people actually carrying the accountability: heads of risk, compliance, data, and operations inside regulated businesses. We wanted to know what they were actually experiencing, so we asked 1,000 of them directly, across finance, energy and utilities, healthcare, insurance, and the public sector.

What they told us is more useful and more uncomfortable than most AI narratives in circulation.

What the AI trust ceiling is

The AI trust ceiling is the point at which AI adoption stops translating into value.

Most organisations have already hit it. It shows up not as a rejection of AI but as a particular kind of friction: AI is used, but never quite trusted. Meeting notes get summarised without a second look. Research outputs get checked line by line. Lending decisions stay firmly with a human. The further a task is from low stakes, the less AI gets near it.

Below the ceiling, every output gets verified. Breaking through it means AI acts autonomously, trusted enough to run without checking. That is where the real return lives, and that is what almost no organisation has reached yet.

What the report found

Here is a taste of what the data shows. The full picture is in the report.

• The trust gap is wider than most organisations admit

59% of leaders say they do not trust AI completely and 77% spend more than an hour each week verifying outputs. The reason they are verifying is not caution for its own sake but because the outputs cannot yet be relied on. That is the trust ceiling in action, and that is exactly where the ROI gap lives.

• The returns are not arriving and the reasons might surprise you

Only 22% of senior leaders report significant ROI. Board members are the least likely of any role to see it, at just 11%. The report traces exactly why the numbers look like this, and the answer is not what most AI vendors would have you believe.

• The sector splits reveal something counterintuitive

Finance leads on positive ROI but carries the heaviest compliance burden. The public sector has the least compliance friction yet the weakest returns of any sector. Understanding how the difference across sectors look is one of the more useful things in the report.

• The human cost is no longer background noise

86% of senior leaders report negative psychological impacts from working with AI. The report puts numbers to what that looks like across roles and sectors and makes the case that this is a signal, not a side effect.

• The unlock is closer than it looks

30% of leaders said they would increase their AI budget immediately if accuracy and explainability could be guaranteed. The report sets out what leaders said they would do differently, sector by sector and role by role, and what that means for AI strategy in 2026.

What this means for how AI gets built

The trust ceiling exists for a structural reason. Most AI tools in the market today are built on large language models, powerful at generating fluent output, but unable to show their reasoning, guarantee consistency, or be held to account when they are wrong. These are not fixable with better prompting. They are limitations of the architecture.

The data is in the report. So is our thinking on what comes next.