Everyone seems obsessed with prompts. The perfect prompt. The magic formula. The seven-word question that unlocks everything. I get it — but I think it misses the point.
For me, using AI well has never really been about writing the perfect prompt. Good prompts help, of course. But the bigger shift in how I work came when I stopped trying to get AI to do everything, and started thinking more carefully about what I am good at — and what AI is genuinely better at than me.
Play to its strengths, not yours
AI is excellent at certain things. Structuring and summarising research findings. Pulling together notes from a long meeting and surfacing the details you half-heard and nearly missed. Processing large amounts of information quickly and consistently. These are tasks where it genuinely earns its place.
Research backs this up: AI summarisation tools free up time for the parts of work that most require human insight — critical evaluation, synthesis, and original thinking. Let AI do the legwork so you can focus on the judgment.
When I use AI to handle the structural, administrative, or research-heavy parts of a task, I find I have more mental space for the bits that actually matter.
As Ginni Rometty, former CEO of IBM, put it: “AI is a tool, not a destination. The real value will come from using AI to enhance human creativity and innovation.”
But don't ask it to do the creative bit
What nobody really talks about in all the AI excitement is this — it’s not very good at imagination. And it’s not great at coming up with something genuinely new. At least not yet.
Yes, it might spark an idea you hadn’t thought of. But that’s mostly because it’s read more than you have, not because it’s being creative. It’s joining dots that already exist, not drawing new ones.
Dr. Fei-Fei Li, Professor of Computer Science at Stanford and one of the most respected AI researchers in the world, makes this point clearly. For all the data and computing power behind these models, she argues they still can’t make the kind of leap that allowed Newton to discover classical mechanics or Einstein to rewrite physics. That sort of thinking — where you question everything you already know and arrive somewhere nobody has been before — isn’t something AI can do.
And when you look at how these systems actually work, that makes sense. They predict the next likely word or idea based on patterns in their training data. They’re reorganising and rephrasing what already exists. So if something hasn’t been done before, AI has nothing to draw on.
What’s worth paying attention to is the knock-on effect. Research has found that people who rely heavily on AI start to experience what’s been called “design fixation” — they think less divergently, take fewer creative risks, and gradually stop pushing into new territory. The more you hand over the creative work, the less you do it yourself.
The prompt obsession gets it backwards
When people focus on perfecting their prompts, they’re often trying to get AI to do the hardest, most valuable part of their work — the original thinking. But that’s precisely where AI is weakest.
One telling piece of research found that AI improved the output of less creative people significantly, but made almost no difference to those who were already strong creative thinkers. The implication is uncomfortable: if AI is doing a lot of your creative heavy lifting, it might be filling a gap — and that gap is worth closing yourself.
How I have done this in practice
I use AI to structure, summarise, find things I might have missed, and draft at pace. I built a plugin recently where AI handled the code — but the idea came from me, knowing the problem existed and knowing my clients well enough to know they’d want it. Same with reports and briefs: my understanding of what the client actually needs is what makes the output useful. Without that, it produces something generic.
The creative bit — the original angle, the judgment call, the thing that hasn’t been done before — that stays with me. Not because I’m precious about it, but because AI genuinely can’t do it. As I explored in Creativity needs friction, the machine can generate the response. It can’t decide whether it’s worth making.
The goal isn’t the perfect prompt. The goal is to know which parts of your work are worth protecting.
Further reading
- AI can make you more creative — but it has limits — MIT Technology Review
- Exploring creativity in human–AI co-creation — Frontiers in Computer Science
- Deepening ideas vs. exploring new ones: AI strategy effects in human-AI creative collaboration — PLOS ONE
- The paradox of creativity in generative AI — Frontiers in Psychology
- How AI ideas affect the creativity, diversity, and evolution of human ideas — arXiv
- Fei-Fei Li on AI, creativity and what machines still can’t do — Issues in Science and Technology
