What makes someone valuable in an AI world

On my way to a job in London recently, I was half-listening to the Diary of a CEO podcast — Stephen Bartlett with Karen Hao, one of those long ones that drifts into hiring, business models, the future of work — and then he said something that made me almost stop the car to write down some notes (I did a voice note for myself instead, hands-free).

He was talking about how he’s been using AI agents instead of hiring people for certain roles. And then the harder question: once you’ve done that, what kind of person do you actually still need?

Three things. Deep expertise. Genuine curiosity about the technology. And being properly good with people.

I didn’t write those down because they were clever. I wrote them down because I recognised myself in all three — and then immediately thought about the places where I’m falling short on showing it.

Depth is the thing AI can’t fake

The example he used was a CFO. Someone with years of financial experience who now sits above a layer of AI agents, orchestrating them. Not replaced by them. Directing them.

That’s a specific kind of value. It’s not “I can use the tool.” It’s “I know when the tool is wrong.”

I think about this constantly in delivery and operations. I’ve sat inside enough agencies and scaling teams to know where the same problems come back, which processes look fine on paper but eat hours in practice, why certain handoffs break every time. That pattern recognition didn’t come from reading about ops. It came from years of being in the room when things went sideways.

Here’s where it gets uncomfortable. AI output looks plausible. Almost always. A plan, a report, a process — it comes back structured, confident, well-formatted. If you don’t have the depth to interrogate it, you’ll accept it. You’ll build on it. And you won’t know it’s wrong until someone further down the line — usually a client — spots the gap.

I watched this happen to an agency last year. They replaced a specialist supplier with AI-generated output because it looked comparable. Took months before anyone noticed the work had gone flat. Same structure, none of the nuance. By the time the client flagged it, the old relationship was cold and rebuilding cost more than the original contract.

Depth isn’t just knowledge. It’s the immune system that stops bad output from getting through. Without it, you’re just a person pressing buttons and hoping for the best.

The gap between doing and showing

The second thing was genuine curiosity about the technology. Not reading about it. Not attending the webinar. Actually building things.

This one stung a bit, because I do this all the time and barely talk about it.

I build systems with large language models that I use every day. Not experiments — things that work. Automations that handle the admin that used to eat Monday mornings. Briefing pipelines that produce consistent output in a quarter of the time. Monitoring tools that surface problems before anyone has to ask “where are we on this?”

I’ve wired together workflows that would’ve taken a junior ops person a full day, and they run in minutes now. Not because the AI is magic, but because I know which problems are worth automating and which ones need a human conversation.

But if you looked at my content from the last year, you’d think I just talked about calm delivery and good processes. Which I do. But the building underneath — the actual hands-on work with the tools — barely shows up. That’s a gap I need to close. Because the insight comes from the building. The opinions I have about what AI can and can’t do in real delivery environments come from having tried it, watched it break, fixed it, and tried again.

There’s a meaningful difference between someone who’s read about what these tools can do and someone who’s been up at midnight trying to work out why their automation just sent the wrong brief to three clients. That second person has something the first person doesn’t: earned judgment. And earned judgment is getting rarer because most people are still at the reading-about-it stage.

People skills aren’t a soft category anymore

The third one sounds obvious. Being good with people. Having real conversations. Building trust. Getting a room to move.

It sounds obvious because we’ve been saying it for decades. But here’s why it’s different now.

AI is compressing the middle of most organisations. The work that was mostly about executing instructions, following a template, producing a deliverable — that work is shrinking. Fast. Which means the floor is rising. Everyone’s doing less grunt work, and the thing that separates useful from average is shifting up the stack.

Up the stack, everything runs on relationships.

Can you sit in a room with a founder who’s frustrated and help them think clearly about what’s actually going wrong? Can you say the awkward thing that everyone’s avoiding and have it land without blowing up the meeting? Can you build enough trust with a team in two weeks that they’ll actually change how they work?

None of that is promptable. You can’t engineer your way into reading a room. And the people who’ve always been good at this — not performed it, actually done it — are about to become much more visible. Because when the tasks around them get automated, the human skill isn’t buried under admin anymore. It’s the main thing you’re paying for.

I’m not going to write a self-congratulatory paragraph here. But I will say: this is the part of the work I care about most, and it’s the part I try to do rather than talk about doing.

Why now and not five years ago

These three things aren’t new. Depth, curiosity, people skills — they’ve always mattered. What’s changed is that AI has made the gap between people who have them and people who don’t much harder to hide.

Five years ago, someone without deep expertise could still be useful because there was enough routine work to fill the day. Someone without curiosity about tools could get by because the old tools still worked. Someone who wasn’t great with people could hide behind deliverables and process.

All of those hiding places are getting smaller.

If I came into a business tomorrow as head of delivery, these three things are what I’d be starting from. Not because I read them in a podcast summary. Because they’re what I’ve spent years building — and what I’ve watched matter more every month as the tools get better and the middle gets thinner.

The honest admission is that I haven’t been showing the full picture. I talk about the calm. I talk about the systems. I don’t show enough of the building, the experimenting, the curiosity part. And that’s worth fixing, because it’s where a lot of the real value sits.

The tools will keep changing. What sits on top of the tools — depth, curiosity, and the ability to work with people — that’s the bit that compounds. And it’s the bit that’s hardest to hire for, hardest to train for, and most obvious when it’s missing.

With that in mind, I’ll be running a workshop where I’ll take you through using Openclaw! We’ll cover:

• What it is and what it can do
• How I’m using it
• Why it’s the sort of thing you need to be playing with in your business

If you are interested, sign up here: Openclaw Workshop

(date TBC)