A colleague showed me his AI workflow last month. He was visibly proud of it. He'd paste in a brief, get a first draft back in thirty seconds, then spend forty-five minutes fixing it. He called it "a massive time-saver."
I didn't say anything. But I've been thinking about it since.
The most persistent myth about using AI at work isn't that AI is overhyped, or that it will take your job, or any of the dramatic ones. It's quieter than that. It's the belief that AI's main value is speed — that the point is to produce something faster than you could before, and that faster means better.
It's wrong. And the reason it's wrong is also the reason it's so hard to shake.
When people first use AI for real work — not toy experiments, actual deliverables — they reach for the thing they already do and try to do it quicker. Write a report? Ask AI to write the report. Draft an email? Get AI to draft the email. The task stays the same. The clock just runs faster. And yes, it does run faster, so the belief gets confirmed. Thirty-second first draft. Forty-five minutes of fixes. Net saving: maybe ten minutes. Which feels like a win.
But the task was the wrong task to begin with.
Here's what I mean. The real gain from working with AI isn't in the execution — it's in the thinking before the execution. Using AI to interrogate your brief before you write anything. To stress-test your assumptions. To generate the ten alternatives you'd normally skip because generating them manually takes too long. To ask "what's the weakest part of this argument" when you're too close to see it yourself. That work — the upstream, messy, uncomfortable work of thinking more clearly — is where AI earns its keep. Not in the typing.
The complication I should be honest about: that upstream work is slower. It doesn't feel like a productivity win. It feels like... more work. You're doing more before you start "doing". And the output still needs editing. So if you're measuring AI's value in minutes saved per task, you will probably conclude it's not worth it, or that you're using it wrong, or that it only really works for other people's jobs.
You're not using it wrong. You're measuring the wrong thing.
This is where a piece of behavioural science is useful. What keeps the speed myth alive is something called present bias — our tendency to overweight immediate, visible rewards over slower, harder-to-see ones. A thirty-second draft feels like value. The improved thinking that comes from spending twenty minutes interrogating your strategy with AI feels like overhead. Present bias means we keep reaching for the quick draft and calling it progress, because the benefit lands right now and the cost of shallow thinking lands much later — in the meeting where your idea doesn't hold up, or the project that misses the point.
So the myth sticks not because people are lazy or unimaginative. It sticks because human brains are wired to prefer it.
The reframe I'd suggest is this: stop asking "how do I get AI to do this faster?" and start asking "what would I think about differently if generating options were basically free?" Because that's the actual shift. Generation is now cheap. Thinking — real, structured, directed thinking about hard questions — is still the scarce thing. AI is a way to have a thinking partner available at eleven at night when you're trying to figure out whether your strategy has a hole in it.
It is not a faster typist. Treating it as one is a bit like buying a really good kitchen knife and using it to open envelopes.
The colleague with the forty-five-minute edit session isn't doing it wrong because he's bad at prompting. He's doing it wrong because he handed AI the wrong job. The brief needed interrogating. The assumptions needed testing. The structure needed challenging before a single word of the draft existed. AI could have done all of that with him — and the draft that followed would have needed ten minutes of fixes, not forty-five.
That's the bet worth making. Not speed. Depth.
If this is the kind of thing you want to actually build into the way you work — not just agree with in theory and then go back to the thirty-second draft habit — that's exactly what The AI Habit is for. It's 15 minutes a day, and the first 14 days are free. Small daily practice, built around the real cognitive shifts rather than the party tricks. The link is below, and if you're a working professional who wants to think better with AI rather than just produce faster, it's worth a look.
https://harrysharman.com/projects/ai-habit/