Ask a table of senior leaders who owns AI in their organisation, and you’ll get a different answer from every seat.
That’s what happened at our Beyond Headcount VIP dinner at Lucky Cat Bishopsgate. Hosted by our CEO Ollie Whiting and CGO Lucy Kemp, and joined by our MD Terry Dawson and Academy Director Claudia Cohen, the evening brought together leaders from financial services, payments, media, utilities, facilities management and technology. Around the table were CTOs, a COO, a chief of staff, heads of AI and data, and an agentic AI specialist just back from San Francisco.
Why we convened this dinner
Eighteen months ago, we kept having the same conversation, inside our own business and with the people we work with. Everyone knew they had to bring AI in. Nobody had done it before.
We surveyed UK tech workers on how they’re using AI and what they want from their leaders. The answers raised more questions than they settled. Leaders were being asked to plan workforces that might not exist in a few years, hire for capabilities they’d never seen and reskill people for skills nobody could yet define.
That’s why we convened the AI Workforce Council. The Beyond Headcount Blueprint is the result of their first meeting, and it framed the evening around three questions:
- How do you bring your workforce with you, so people grow with AI rather than apart from it?
- Where should AI capability sit, and who owns it?
- How do you govern AI without rewriting your strategy every quarter?
As Lucy said when she opened the night: “Owning AI in an organisation is really lonely. It’s hard to put your hand up and say, I’m not sure this is going to work. Tonight, you’ve got people in the same boat as you.”
What’s your biggest challenge with AI right now?
We went round the table with one question. Ollie went first.
“Whose job is it? Everyone’s asking for budgets. Everyone’s asking for accountability. Who actually owns it?”
Every answer that followed fell into one of the Blueprint’s three questions, plus a few the room added:
- Ownership. Some organisations have put AI with technology. Others have deliberately placed AI and data outside the tech function, to bring the business and technology together, with the CTO still owning stability and infrastructure. Either way, everyone agreed that each executive still has to own their own numbers.
- Data foundations. Years of mergers and legacy systems mean data has to be brought together and cleaned before anything meaningful can sit on top of it. Without that, you can’t move from fixing single problems to agentic workflows.
- Demand. For some, the biggest problem is simply demand. Everyone wants the tools now, and in regulated businesses especially, saying “not yet” can make you look like the IT killjoy.
- Proving value. The Blueprint notes that leaders are being asked to show the return before they’ve done the work. Boards want to know the size of the prize, and most can’t yet say.
- Expert time. AI projects need the people who know the work to give it real time, not fit it in around the day job.
- Noise. Terry raised a problem many recognised. “I’m seeing three or four decks a week, and they all look great. My worry is that the best ideas get lost because they’re one of ten things someone has seen that week.”
How do you bring your workforce with you?
The organisations making most progress weren’t the ones with the best tools. They were the ones whose people could see a role for themselves.
One COO described a deliberately bottom-up approach: an internal AI academy, time for people to be curious, AI written into everyone’s goals and awards for the best ideas. Their latest engagement survey showed real enthusiasm, because people had seen what AI was already doing for them.
A head of AI tackled it from three directions at once. Leaders get training on spotting use cases, measuring return and helping their teams past the fear of losing their jobs. AI champions and company-wide training weeks build confidence from the bottom up. And “art of the possible” days put graduates in a room with the people who have the problems, to build something they can use at their desk the next day.
Not every team moves at the same pace. Sales teams were often slower, because their targets don’t naturally line up with AI adoption. The answer was to show them how AI makes the job better, not how it measures them.
Claudia shared a way of thinking about it that landed with the room. “Your workforce falls into three groups. The enthusiasts driving it forward. The middle, who are using it but aren’t convinced. And the people letting it happen alongside them, hoping the old ways come back. Your change strategy has to talk to all three differently.”
What are we actually hiring for?
The Blueprint argues that workforce planning has to move from counting roles to building capability, planning for the mix of people and AI you need rather than a number of heads. The room took that further.
Our agentic AI specialist made a distinction several guests wrote down. Many organisations fused process (the steps that get work done) with methodology (the philosophy and ceremony around it). Untangle the two, and you can see which work creates value and what people need to be doing.
That changes who you hire. Team leaders used to be hired to organise and coordinate. Now an agent can track the SLAs, and you need people who solve problems and spot insights. Lucy shared an approach from the Council: give candidates a real problem and ask which prompts they’d use to solve it. It quickly shows who has the curiosity to work in this new way.
Ollie noted how much clients’ priorities have shifted. A few years ago, logical thinking and reasoning topped the list. Now it’s resilience, empathy and curiosity, which are also the hardest things to test in an interview.
What happens to the next generation?
The conversation that stayed with people most was about the next generation.
A data science leader in media worried about a generation that hasn’t built “that discernment muscle” by doing the work, but is now being asked to check the work of an agent. Claudia sees the same tension in our Academy. “Our ethos has always been techniques over tools. Teach people how to learn, and they’ll adapt to whatever comes along. Give tech-native, smart people the right tools and they can deliver the output of much more senior roles.”
She also raised something many organisations miss: the apprenticeship levy can be used to upskill the people you already have, not only to bring in new ones.
Lucy made the case plainly. “As leaders, we have a responsibility to keep bringing in the younger generation. Otherwise, we won’t have the culture. We won’t have the brilliant middle managers coming up through the ranks, because we never brought them into the game.”
Not everyone thought it was simple. One guest pointed out that every commercial business is under short-term pressure, and developing junior talent is a long-term investment. Others argued that bringing in young people is how you find out where the edge really is. “If you bring someone in and they’re young, they have to be a bit of a disruptor.”
How do you govern AI without slowing down?
The room’s answer: start with yes.
One Head of AI argued that most AI policies are old infosec policies “with an AI T-shirt on”. A better policy starts with permission. Yes, you can brainstorm with free tools. Yes, you can build something customer-facing, with an approved tool and a risk assessment. The rules tighten as the stakes rise, but the starting point is yes.
A CTO made the case for compliance by design: thinking about threats, privacy and accessibility up front, then using tooling to monitor and correct. He tracks the cost of every feature his team builds, down to individual engineers and clients. “Your cost metrics with AI can run away before you know you’re in trouble.”
The data leader from a regulated business raised a risk few had considered: permission debt. Long-serving employees build up access from every role they’ve held. Give them an AI assistant that can see everything they can see, and it can combine information it shouldn’t, without explaining where its answer came from.
Is AI actually giving us time back?
Late in the evening, one guest asked the question everyone had been circling. If AI makes a team 30% more efficient, what happens to that 30%? Do you bank it by cutting roles, or reinvest it in growth?
The honest answer from the room: most people don’t feel less busy. Meetings now produce fifteen actions instead of three. Presentations run to thirty pages instead of two. One guest called it “an illusion of productivity”.
Real gains come from redesigning a whole workflow, not speeding up a single task. And what you do with those gains is a business decision. As several guests put it, there’s no such thing as an AI strategy. There’s a business strategy, and AI is how you deliver it.
That links back to the Blueprint’s point about planning. Most leadership teams still plan a year ahead, while AI changes quarter to quarter. The answer isn’t to keep rewriting the plan. It’s to anchor AI to the business problems that matter most, and build from there.
What leaders are taking back to work
We closed by asking everyone what they’ll do differently. Their answers make a practical starting point for any leadership team.
- Rewrite how you hire. Build job descriptions, interviews and assessments around curiosity, resilience and problem-solving. Give candidates a real problem and watch how they use AI to solve it.
- Grow the fearless people you already have. Don’t only look outside for disruptors. Give your existing people the space and permission to challenge how things are done.
- Use AI to buy back time for people. Swap the AI-generated report for a walk and a coffee. Terry put it simply: “For me, it’s doubling down on human connection, with clients and with my own people. Use the tools when we need them to speed us up, then spend that time building relationships.”
- Keep the door open to the next generation. One leader admitted that, a year into a role focused on transformation, they’d forgotten about the talent coming next. Make early careers part of your AI plan, not an afterthought.
- Start with one valuable problem. Pick the business problem AI is best placed to solve, get the data behind it right, and set your guardrails before you scale.
- Keep talking to your peers. “Everyone thinks they’re behind. Everyone thinks everyone else knows what they’re doing.” Conversations across companies and sectors are how leaders get through something nobody has done before.
Ollie closed with a personal reflection. “I wouldn’t be in the role I’m in now if someone hadn’t given me a shot at 20. If we don’t get entry-level talent right, we don’t have a future. My takeaway is to listen more on this, and talk less.”
Hear from the AI Workforce Council
The Beyond Headcount Blueprint answers the workforce questions UK leaders are asking about AI, from reskilling to ownership to governance.
Join us at Beyond Headcount Live on Wednesday 4th November, where members of the AI Workforce Council will take your questions. Everyone who registers gets the Blueprint first.
Register for Beyond Headcount Live and find out more about Beyond Headcount: