On Thursday 8 October, 15 CEOs, business owners, and technology leaders from UK businesses sat down for lunch at The Cocochine in Mayfair.
They came from sectors including professional services, financial services, recruitment, healthcare, hospitality, property, residential services, and facilities management.
The brief was simple: an off-the-record conversation about what AI means for the way leaders lead, with every contribution kept anonymous. Three hours later, few people were in a hurry to leave.
“The best three hours I've spent in years.” - CEO
“An exceptional use of my time.” - CEO
Our host, Ben Legg, former COO of Google Europe, opened not with a slide deck but with his own working day. His argument was simple: understanding what AI can do isn't something a leader can delegate. Every CEO needs a personal tech stack and a plan to keep learning. Not everyone needs to build agents, he said, but leaders who aren't using them are spending more time than they need to.
He began his own agentic AI journey by asking AI to build him a 30-day training plan, one hour a day, with the goal of building his own agents. Today he runs around 60 agents at once and uses AI to schedule three recurring learning sessions in his calendar.
His inbox: one agent clears out unwanted email, another checks spam, and a third drafts replies in his style. Within two weeks, the drafts were 70–95% accurate. He still reviews each one before it goes out, refining it to better match his voice and style.
His diary: AI briefs him before each of his 5–12 daily meetings, and an agent handles new meeting requests.
His children's news: two AI tools work as a team to find suitable stories, write a script, produce a daily news video for his young children, and review the final output.
The children's news bulletin prompted the biggest reaction of the afternoon. One CEO admitted they'd be trying it at home that weekend.
The habit, he said, starts in everyday life. He keeps the final 5% of every task for himself because he values the last human interaction.
For Ben, AI makes everyone a B-grade performer; it's the human who turns the work into an A.
As the discussion widened, the challenges were remarkably similar: AI moving faster than investment cycles, fragmented data and legacy platforms, M&A integration, scarce practical skills, and slower-than-expected returns.
Leaders also shared what they were already trying:
One leader had appointed an AI champion in every part of the business and brought them together as a committee.
Another believes AI can remove “the work to do the work”, allowing people to focus on more meaningful contributions. For now, however, many employees still view it as a threat.
A leader explained why they deliberately keep AI out of their creative pursuits to protect its authenticity and use it for planning instead.
Several were already encountering the limits of AI investments and platform decisions made just two years earlier. Others were paying heavily every month for systems they were reluctant to change, worried that a replacement could prove even more costly.
Not every topic drew the same level of agreement. Reliance on overseas providers was a recurring theme, with guests questioning the reliability of Chinese and US models and discussing the wider AI landscape in Europe and Russia.
On AI going rogue, the room broadly agreed that a kill switch was necessary. Equally, one system failing wasn't seen as a reason to shut everything down. There was a broad consensus around the importance of regulation.
Cost divided opinion too. Some compared the cost of agents, including tokens, directly with the cost of people. Ben treats people and token spend as a single cost, the cost of executing a task, and reflects both in his P&L. Others pointed out that once subscriptions, connections, and scaling are counted, agents aren't automatically cheaper than people.
If one idea united the room, it was this: leaders no longer manage people and technology separately; they lead a single system of people and agents working towards the same goal.
Ben directs his agents as he would a team: they draft, brief each other, and check each other's work, while he owns the outcome. The discussion quickly turned to what that new leadership role requires:
Set the goal and the brief: agents, like people, need a clear objective, a defined scope, and a check before work goes out.
Give every agent an owner: someone is accountable for each agent's work, including what happens to a leaver's agents.
Redesign the work: decide which steps suit people, which suit AI, and where AI doesn't belong, and explain your intent before your team fills the silence.
Make it everyone's job: as agents take on the groundwork, people move towards customers, storytelling, and directing agents, so managers need to lead mixed teams with confidence. As one leader observed, the frontline people who see gaps in processes every day are the ones to give the tools to.
Own it at the top: the strongest view was that AI needs a board-level owner, and Ben recommended “one geek per team” to spread what works.
Inspired by Ben's approach, each attendee left with practical actions to test over the next 90 days. These included:
Ask AI to build a personal 30-day learning plan and set calendar blockers to learn something new.
Automate one recurring task, starting with meeting prep.
List the agents already in the business and give each a named owner.
Name a board-level owner for AI.
Our Technology Performance Index, conducted by Opinium, helps explain why leadership is such a strong differentiator. 84% of Tech Vanguard businesses report strong leadership and governance to scale technology-led change, against 3% of Tech Bystanders. On AI, 57% of the Vanguard say they've adopted it well, against 21% of Bystanders.
Thank you to everyone who joined us. It was the candour in the room that made the conversation so valuable.
This was the first in a series of TPI Executive Lunches. More dates will be confirmed soon.
To see where your business sits today, take the Technology Performance Index assessment.
Source: Babble Technology Performance Index, conducted by Opinium with 1,000 UK SMB owners and C-suite executives, February 2026.