From the Jack & Jill breakfast panel in partnership with Ashby, featuring Matt Wilson (CEO, Jack & Jill), Sabrina Castiglione (CFO, Omnia) and Don Fogarty (Head of Talent, Attio).
Most teams now have access to AI tools. Very few have the behaviour change to match.
The gap between the two isn't a technology problem. It's a people problem. And it shows up in a specific way: leadership buys in, a few individuals experiment, a handful of use cases get celebrated in an all-hands, and then it quietly stalls. The tools sit open in browser tabs while the team gets on with the actual work.
Getting past that stall, and building a team that genuinely compounds on AI over time, was the subject of a breakfast conversation we hosted in London with Ashby. We brought together founders and talent leaders who are working through exactly this, and what came out was more useful than any framework: honest, specific thinking from people doing it in real time.
Here's what we learned.
Getting your team using AI: it starts before the tools
The instinct is to start with tooling. Which platform? Which workflow to automate? Which team to pilot with first?
The leaders in the room were consistent on this: that's the wrong starting point.
"The most important trait we're trying to build is not being super AI-skilled for the sake of it, and not being cynical either. It's openness. Thoughtfulness about what is the right tool for the right problem."
Sabrina Castiglione, CFO, Omnia
The distinction matters. Maximising AI usage is not the goal. Building a genuine orientation toward it is. Curious, thoughtful, willing to try things and share what worked. You can teach someone to use a specific tool. Teaching the disposition to keep seeking better ways of working is much harder.
Don Fogarty at Attio set the bar plainly:
"If you're not up for it, not trying it, and not brave enough to experiment, it's probably not going to work."
Don Fogarty, Head of Talent, Attio
That's not a warning. It's a recognition that culture change requires genuine intent from people, not just access to software.
Getting ROI: manufacture the first win
Access and mandate aren't enough. The gap between having the tool and getting real value from it is where most adoption efforts break down.
The most reliable bridge is the shared win. When one person on a team does something genuinely useful with AI and shows it to everyone else, it unlocks thinking across the whole group. The problem is that shared wins require individual wins first, and individual wins require time and a low-stakes entry point, neither of which appear by accident in a fast-moving startup.
"The biggest impact is when somebody shares what they've done with the rest of the team and it unlocks everyone else's thinking."
Sabrina's team solved this with something deliberately simple. One of their engineers wrote a single prompt: it asks the person what their job is, what they do day to day, and then suggests specific ways AI could help. No training programme. No mandatory rollout. Just a prompt, and one protected hour to try it.
"You have to create time for it. Here is an hour, go and spend this hour doing this. Because I think once people realise they can do one thing, it will naturally start the flywheel."
It worked particularly well for the people who needed it most: her talent team, mostly from agency backgrounds, who had never had a reason to engage with the tools before. The first win does not need to be impressive. It needs to be real enough that the person who had it wants another one.
Compounding on AI: systems thinking is the real skill
Once people are using AI, a different problem emerges. Output goes up. Activity accelerates. But the work isn't necessarily better. Teams run faster, sometimes in slightly different directions, and the net effect is noise rather than progress.
"AI fluency is something that is really rewarded by systems thinking. It's not about movement. It's about philosophy and direction."
The organisations getting the most from AI are the ones who have thought clearly about what they are optimising for and channelled AI toward that. Not the ones with the most usage.
Matt Wilson framed the broader shift this way. For most of the history of hiring, organisations faced a binary: human nuance or software scale. A great recruiter could understand context and complexity in a way no filtering algorithm could. But they could only handle so much volume. Software could scan hundreds of thousands of profiles, but couldn't understand the person behind them.
"We're moving to a world where AI is able to deal with the nuance as well as the best humans, and do that at scale. That really is going to change the game."
Matt Wilson, CEO, Jack & Jill
Capturing that shift means being deliberate about what humans in your organisation own and what AI handles, rather than layering AI across everything and hoping the value follows.
Accountability: the one thing AI doesn't take on
When the conversation turned to trust and AI errors, one answer in the room cut through.
"The one thing AI doesn't do is take away accountability from the person whose job it is. Whether it's a finance person or an engineer, they still own it and they are responsible for it."
The practical implication is that teams need to build literacy around risk, not just capability. Not all outputs carry the same stakes. A candidate outreach email with imperfect phrasing is a different category of problem to a contract or a regulatory filing.
The speed and scale belong to AI. The decision is still human.
Hiring for AI-native: what you're actually assessing
The question of what AI-native means as a hiring criterion came up early, and the answer was consistent across the room: it is much less about what someone has already built and much more about how they think.
"We ask people indirectly what they're building and how they think AI will evolve their role. If they haven't demonstrated that they've at least thought about their entire workload changing, then it's very hard to have a conversation about it."
The signal you are looking for is not technical proficiency. It is evidence that someone has genuinely reckoned with what AI means for their work: which parts of their role are most exposed, which require judgement that won't be easily replicated, and what they would do differently if the tools available to them changed significantly in the next six months. Someone who has had that conversation with themselves will keep up. Someone who hasn't probably won't.
On the hiring question itself, Sabrina brought it back to first principles:
"What is the problem that you're trying to solve with this specific hire? There is something on the other side of this hire that will be true that isn't true today. The answer to that question will determine who the right person is."
In an environment where the tools available to a single person keep expanding, the shape of what you need from a hire is shifting. The question is less often "who can cover this function" and more often "who can own this outcome end to end and figure out how to reach it with whatever is available."
Matt put the internal version of this directly:
"This is the job to be done. Go and figure out how to do it. I can't be architecting all of those things across the whole business. It needs to be owned, end to end."
At Jack & Jill, one person is responsible for everything that happens post-signing for 200,000 Jack users: support, product input, the full cycle. AI is what makes that viable for one person. It is not a reason to add headcount around it.
What this means for org design
The structural shift is already showing up in how companies plan headcount.
"We're starting to see headcount plans with specific reference to AI agents. Our trigger for hiring support people is slowing down because AI is getting better at handling some of those issues. That could happen in sales, in recruiting, across all functions."
If AI can handle volume work with reasonable quality, the marginal value of the next hire changes. The question stops being "who can do this task" and starts being "who can build and oversee the system that does this task." People who can identify which parts of a workflow should be systematised, which need genuine judgement, and where the next automation opportunity is before anyone tells them to find it.
That is a different hire to the one most TA teams have been making. And developing that capability in existing teams, rather than just waiting to hire it in, is one of the more underrated challenges in building AI-native organisations.
The honest summary
The teams making real progress share a few things in common: they make it safe to experiment, carve out time rather than hoping people find it, celebrate small wins publicly, and are clear that accountability does not transfer to the machine.
None of that is novel. It is change management, applied to a change that is moving faster than most organisations have dealt with before.
The tools will keep changing. What you are actually building when you do this well is not an AI-native team in any fixed sense. You are building a team with the disposition and habits to keep adapting as the landscape shifts. That is harder to build than any specific AI capability. It is also more durable.
This article draws on conversations from the Jack & Jill breakfast panel, hosted in partnership with Ashby in London.
