On 12 March, we brought together 50 senior leaders in London to launch our AI Trust Ceiling research.

The room included heads of AI from global banks, Big Four partners, public sector policy leads, healthcare execs. The kind of room where nobody needs AI explained to them, but everyone’s trying to figure out how to embed trustworthy AI into their work.

The Q&A went over time and the questions covered everything from liability frameworks to whether stacking multiple AI models helps or makes things worse. The panel didn’t always agree, which made for a better conversation.What stuck with us was the reaction in the room when we showed the data. When William mentioned that 77% of leaders spend significant time verifying AI outputs, people started nodding before he’d finished the sentence. One attendee said afterwards that “significant time” was generous.

Networking continued well past the scheduled finish. People compared notes on verification workflows, swapped stories about AI failures they couldn’t talk about publicly, and asked follow-up questions about the technical approach. The consistent theme was recognition - the research had put numbers to something they’d been experiencing but hadn’t seen quantified.

Thanks to everyone who came, and to the team who made it happen.

The full report is now available for a Download here.