AI usually gets handed to whoever has tech in their job title. Our first Beyond Headcount roundtable started from a different idea. AI is a people tool, so people leaders should be the ones leading it.
On 7th October, we brought Chief People Officers, HR directors and senior people leaders together at The Ivy Victoria. They were the first group to discuss the findings of the Beyond Headcount Blueprint, our guide to AI workforce planning, built with our AI Workforce Council. Daniel Ferguson, Associate Partner at JMAN Group, joined us from the AI Workforce Council to answer questions and share what he’s seeing across the organisations he works with.
The conversation ran under Chatham House Rule, so we haven’t attributed anything said by our guests. Here’s what we heard.
Why AI workforce planning starts with people
Lucy Kemp, our Chief Marketing Officer, opened with the reason this room came first.
“AI is a people tool. It changes the way you hire. It changes who you hire. It changes the jobs. I was talking to someone who gives their agents OKRs, and if they don’t hit them, they have a disciplinary. That is the world we’re stepping into, and it’s not a CTO who is going to do that.”
She also named the problem that sits underneath it all. Leaders are being asked to hire and plan for AI, without a clear picture of how it’s already being used inside their own organisation.
Plan for capacity, not headcount
Most workforce plans run on a 12 to 18 month cycle. AI capability shifts every quarter. That’s why the Blueprint calls for capacity, not headcount: treating employees, contractors, AI agents and automation as levers toward the same outcome, planned and reviewed together.
Several guests are already working this way. One people leader described an organisation that had long treated headcount as shorthand for culture. Its real workforce now includes consultants, contractors and agents, and at one point it had more agents than people. Most of them weren’t any good, so governance came back in. Hiring has become more deliberate, with AI engineering and product still growing while other recruitment slows.
Another guest, deep in budget season, is using AI to change the conversation with their CFO. Instead of fielding requests for new roles, they’re showing how the ratio of support staff to the wider workforce can keep improving as the business grows.
The warning that landed hardest was about process. Put an AI tool over a broken process and you lock that process in for good. The harder, more valuable work is reimagining the workflow first.
Daniel was clear that waiting for certainty isn’t an option.
“There’s no playbook. There’s no book you can open that tells you exactly how it is, and it’s changing at a rapid pace. You just have to start.”
In practice, that means bringing headcount and AI spend into the same conversation, reviewed every quarter rather than once a year. And before automating anything, asking whether the process should exist in its current form at all.
Governance that speeds things up
Most organisations started with a wide rollout of AI tools. That created a burst of activity, some good, some not. The guests who’ve moved furthest described governance as the thing that unlocked progress, not the thing that slowed it down.
One organisation runs an internal model where ideas go in, agents are built and configured, and every agent is reassessed against key results. If it isn’t working, it’s switched off. If costs rise, it moves to a cheaper model.
Another guest shared what happened when accountability was split across risk, legal, people and technology. Everyone agreed in principle, one party took the lead, and the rest faded into the background.
Daniel’s advice was to separate the two jobs.
“Proving a concept is easier than it’s ever been. Productionising it is what we have to be mindful of. We set up a governance committee, found it was taking on too much strategy, and created a separate group for AI strategy, so we weren’t stifling creativity.”
On ownership, he echoed one of the Blueprint’s four shifts, accountability, not architecture.
“It can’t be technology that owns the outcomes. It has to be the area that needs the agent to do the work.”
One people leader put it in terms every HR team will recognise. HR builds the infrastructure, but managers are accountable for their own teams. Agents should work the same way. Some guests are already writing job specs for agents. The open question is whether managers even know which agents their teams have built.
The clearest starting point from the room was to give every agent a named owner in the function it serves, and to keep the group that sets AI strategy separate from the one that governs it.
Hire for curiosity
The word that came up most around the table was curiosity.
One leader shared what changed when they redesigned our own people team. “We went from five in our people team down to two, but using agents with those two. What they’ve achieved in a year versus when we had a bigger team is drastically different. We’ve jumped 16% in engagement from the activity of those two people.”
Another guest described a member of their talent acquisition team who built his own recruitment platform in his own time, checking with the data protection team along the way. A demo to the CEO turned into one of the most energising meetings the business had seen. Another talked about a recruiter who automated her routine work, moved up the value chain and into a promoted role.
Those stories spread. Visible role models do more for adoption than any policy.
There was also a strong view that people leaders need to be among the most AI-literate people on their executive team. Several guests had spent months upskilling themselves to make sure they could hold that conversation. And there was a clear obligation to the people who aren’t on the journey yet. They’re the ones most at risk if nobody brings them along.
For people teams, the lesson is to find the curious people in the business, give them room to experiment, and make sure everyone can see what they’re building.
Protect the entry point
The Blueprint warns that if AI absorbs entry-level work, the pipeline that produces your next senior leaders dies, and nobody owns that problem because it lands in five years. Lucy put it plainly.
“That’s where we all learnt how to act in meetings, how to talk to senior leaders, even how to write emails. That is being removed.”
The room split between concern and optimism. One guest, who sees skills data across a whole sector, described graduate roles shrinking and a mismatch between what employers want (AI-literate, confident communicators) and what the education system produces. Their response is more work placements, offered for societal benefit rather than commercial return.
Others see opportunity. Claudia Cohen, Director of La Fosse Academy, sees it in the junior talent the Academy trains and places every day.
“The people experimenting the most are often the junior level. They’re tech native and they’ve grown up with the technology around them. If you find people with critical thinking and curiosity, and then train them in the new tools, they can go in and disrupt the organisation.”
One organisation redesigned its entry-level roles away from manual reporting, added a ten-week digital bootcamp before day one, and now gets new starters in front of clients sooner. It hires for critical thinking, communication and relationship-building.
Daniel’s firm invested further in its graduate programme, with promotion cycles every three months.
“If you have the right mentality and the right skill set, you can jump over people that do not.”
For anyone taking the case to a CFO, the advice from the room was to frame early careers roles as a growth investment, and to point out how expensive it is to hire experienced managers in from outside. Hannah summed it up in one question.
“Where does your CFO think your next managers come from?”
Put that way, early careers roles stop looking like a cost to trim. They’re the business’s future leadership pipeline.
Reskilling inside the workflow
One line from the Blueprint stopped the room. Tia Cheang, an AI Workforce Council member, told us the time lag between a business deciding it needs new skills and actually delivering them is around nine years.
Guests agreed that spend isn’t the right measure. Most learning happens on the job, at the point of need, with people learning from the person next to them. The challenge is capturing that so the whole business benefits.
Daniel made the case for teaching the basics properly, from how models generate answers to what they cost to run. He shared an example of a report someone built that cost ten times more than it needed to. Paired with a technologist, it came down by ten times and became more reliable. In his firm, anyone who wanted a licence had to complete a short course first, with time set aside to do it.
That raised a real tension. We’ve told people to be curious and experiment. Now cost, sustainability and responsible use mean some of that needs reining in, without killing the curiosity that got us here.
For teams who don’t sit at a desk all day, structured programmes matter even more. Many people won’t seek out a free online course, and won’t admit they’re not confident with technology.
The approach the room kept coming back to was learning built into the moment someone uses the tool, alongside a proper grounding in how models work and what they cost.
Leaders set the tone
The last theme was leadership. The room agreed leaders don’t need deep technical knowledge. They need a baseline, a shared vision, and the confidence to make it safe for people to try things.
One guest described finding two of their team in a hotel lobby late at night, anxious about what they were and weren’t allowed to put into an AI tool. That’s the gap leaders need to close.
Hannah was direct about the stakes.
“If the leadership team are anti it, that sets the tone for everyone else to clam up. It’s leadership painting the vision of what life could look like if we go on this journey, versus everyone fighting for their job.”
Daniel’s firm won’t take on AI work unless it’s coming from the top.
“If it’s not coming from the board, with an agenda and outcomes they want to achieve, it becomes really hard to drive value.”
Guests also warned against the one-off offsite. Three months later it feels like an eternity ago. Leadership learning needs to show up in weekly rhythms, not a single event.
Key takeaways for people leaders
- AI workforce planning is a people job. It changes how you hire, who you hire and what the jobs are.
- Plan capacity, not headcount. Review people, contractors, agents and automation together, every quarter, using a mix of hire, build and borrow.
- Governance should speed things up. Separate AI strategy from AI governance, and give every agent an owner in the business.
- Hire and reward curiosity. Make your role models visible.
- Protect the entry point. Today’s graduates are your managers in five years.
- Leaders go first. A shared vision and psychological safety matter more than technical depth.
What’s next?
These conversations don’t stop here.
On 4th November, we’re bringing the AI Workforce Council to Beyond Headcount Live at Tower Suites by Blue Orchid. It’s not a panel. It’s a live Q&A built around your questions.
Register for Beyond Headcount Live
Take the AI Sense Check to see where your organisation stands today.