I've been working with students at a further education college. Seventeen to twenty-one-year-olds. Bright. Engaged. Trying to figure out what the working world wants from them.
It struck me, during the last session, that something was off.
These are young people who've come of age surrounded by AI. The generation that should be running rings around the rest of us with these tools. And what I'm actually seeing is ChatGPT being used like a vending machine. Put the question in. Get the answer out. Move on. The idea that you might push back on what it gives you, come at the same problem sideways, use it to challenge your own thinking rather than replace it. Even try another AI tool. None of that is happening.
It’s functional use. Not exploratory
Nobody taught them to be curious about it.
I also mentor at Sheffield Business School, students aged 20 to 25. Same thing with AI. Slightly more polished, good at getting round AI detection software, but beyond that, still fundamentally surface-level in how they're using it.
Not because they're lazy or thick. Because the education system has been woeful at teaching AI use with any imagination whatsoever.
This has been rattling around my head for weeks. Then, within three days, three things landed that pulled it into focus.
First, Zoe Scaman published The Pipeline Problem. If you haven't read it, stop reading this and Go. Read. That. First. She's forensic and furious in equal measure, arguing that the creative industry's current love affair with "senior talent plus AI equals superpower" is a half-truth with a shelf life. Those people making it work built their judgment through years of friction, failure, and being told their briefs were bollocks. The industry is now systematically dismantling the conditions that produced them, and calling it progress.
Two days later, Ethan Mollick dropped The Shape of the Thing, charting where AI capability has actually got to.
We're not prompting any more.
We're managing.
You hand AI agents hours of work and walk away. It's Tuesday afternoon and the future is already here, just unevenly distributed, as someone once said.
Sitting underneath both was the 2026 State of the Workplace report. 443 million hours of behavioural data. Over 1,100 organisations. The headlines look brilliant: 80% AI adoption, productive hours up 5%, burnout falling. Dig in and it gets uncomfortable. Focus efficiency has hit a three-year low of 60%. The average focused work session lasts thirteen minutes. Thirteen. And nearly one in four employees is now chronically disengaged. Not burned out. Under-challenged. Sat there bored and no one is noticing.
Three articles. Three days. All pointing at the same problem from different angles.
We are hoarding the wrong assets.
The creative industries are obsessively stacking up AI capability. The tools, the platforms, the integrations, the adoption numbers to wave at the board. And while we're busy collecting all of that, we are quietly starving the one human quality that makes any of it actually work.
Curiosity of youth.
Not the corporate buzzword version.
The real thing.
The ability to sit with a problem that hasn't resolved. To hold the "what if" open for longer than is comfortable. To push past the first correct answer, which AI will always give you very efficiently, into the genuinely interesting one, which it won't without a burning curiosity for more.
That workplace data has a number that should worry anyone running a creative business. Of all the AI users they studied, only 3% operate in what the report calls the "sweet spot" of productivity: roughly 7 to 10% of their total work time spent in AI tools. The majority, 57%, spend less than 1% of their time in AI tools at all. People have access to extraordinary instruments and they're barely touching them.
That's not a training gap.
That's a curiosity gap.
A massive one.
AI is a very powerful set of instruments. You still need to know what music you're trying to make.
Zoe puts it better than I will: the industry is "using the last generation of properly developed talent to argue that properly developing talent is no longer necessary."
The people celebrating the senior-plus-AI model got where they are through years of being wrong in public, having their thinking torn apart, sitting in rooms absorbing things they couldn't yet name. That kind of judgment doesn't arrive with a job title. It's built. Slowly. Through the exact conditions we're now stripping out of early careers because they look like inefficiency.
If Mollick is right, and I think he is. We've entered an era of managing AI rather than working alongside it, then we need a generation who know what to manage toward.
That takes judgment.
Judgment takes curiosity, time, and productive struggle. Three things the modern workplace is systematically removing from the first jobbers experience.
The companies rushing hardest to implement AI while skipping the cognitive development of their people are building something really self-defeating. You can't use AI curiously if you've never learned to be curious. Skip the human development and your AI investment gets dumber, not smarter. You end up with a workforce of operators. Clean executors. People who follow the rules brilliantly and never once ask whether the rules are matter.
Operators are fine. Necessary, even. But nobody ever built anything remarkable by following the rules. Remarkable comes from the person in the room who says "what if we tried something else?" and has the cognitive stamina to sit in the silence that follows.
So what do we actually do about this? Because I do think there's something to be done, and for fecks sake its not another lunchtime workshop on prompting!
It's structural.
The first six months of any junior role in any business should be designed around the habit of questioning. Not executing. Not producing deliverables. Questioning. What does this problem actually want from us? What assumptions haven't we tested? What would a completely different answer look like? Build the muscle first. The tools will follow, and they'll be used properly, because the person holding them will know what they're looking for.
Protect the cognitive space where ideas actually form. Thirteen-minute focus sessions and a 23% disengagement rate tell you something important about modern workplaces: we've designed environments that interrupt thinking before it gets anywhere interesting. The "what if" that produces something genuinely original needs more than thirteen minutes to show up. That's a fixable problem, if anyone decides to care about it.
And redesign the junior role.
Zoe draws the comparison with medical residency and she's right. You don't hand a trainee surgeon a scalpel on day one. You teach them to observe, to evaluate, to develop judgment through proximity to people who've done it properly. The junior who learns to look at AI output and say "that's technically correct and also completely bland" is building something that compounds over a career. The junior being used as a cheap pair of hands is not.
The window for this is narrow.
Mollick says we're at a moment where individual and organisational choices are setting precedents. Zoe says the damage, left to compound, becomes irreversible. The Workplace data says the real challenge of 2026 isn't AI adoption, it's whether the humans alongside AI are being developed or merely deployed.
Take it from someone who's spent a career zigzagging between creative director, strategist, agency founder, and occasional spectacular failure: the most valuable thing I ever learned wasn't a tool or a framework. It was how to sit with not knowing. How to stay curious when every pressure around me was screaming for the quick answer. That's the asset that appreciates.
The AI tools you're stockpiling will depreciate.
Faster than you think.
The curiosity you're not developing in the next generation of talent? That was the investment that mattered.
And right now, almost nobody is making it.
Written by a human, made readable by Claude.ai.
Spark was reading Ethan Mollick's 'The Shape of the Thing' (March 12, 2026), Zoe Scaman's 'The Pipeline Problem' (March 10, 2026), and the ActivTrak 2026 State of the Workplace report. I am Philip Slade.
