A couple of years ago, ask a designer whether AI would take their job and you’d get a nervous laugh. Now the question has changed. It isn’t “will it?” so much as “what am I actually for?”
My honest answer is in how I divide the work. I use AI every day. ChatGPT, Gemini and Claude all live in my toolbox, and I’m not going to pretend otherwise. What I care about is the line: where its work stops and mine starts.
Where AI gets the pen
- First drafts of anything. Copy options, edge-case lists, error messages, test scripts.
- Variations. Ten ways to word a label, so I can see quickly which ones I hate.
- Grunt work. Tidying notes, reformatting, turning a transcript into a list I can argue with.
- Rough prototypes. Something clickable to put in front of a person sooner.
It’s quick, and it never gets bored of the dull version. That speed is real and I’d be foolish to turn it down.
Where I take it back
- Choosing the question. AI answers what you ask. It can’t tell you that you’re asking the wrong thing.
- Reading people. A transcript has words. A session has hesitation, a laugh, and someone quietly working around the thing you built.
- Taste and trade-offs. It will hand you something plausible. Plausible is the enemy of good, because plausible looks like everything else.
- The signature. If my name is on it, I can explain every choice in it. If I can’t, it isn’t finished.
- First drafts of anything
- Variations
- Grunt work
- Rough prototypes
- Choosing the question
- Reading people
- Taste and trade-offs
- The signature
Three questions before I accept anything AI hands me
- Who is this for, and would they recognise themselves in it?
- Would it survive somebody’s worst Tuesday? Tired, distracted, one bar of signal.
- Can I say why, in my own words, without looking at the screen?
If I can’t answer all three, I haven’t used AI. I’ve outsourced the thinking, and the thinking is the job.
Think first, then prompt
There’s an order of operations I’ve come to care about. If I open a chat to escape the blank page, I lose the muscle that makes me useful. If I do the hard thinking first, even badly, on paper, and bring AI in after, I keep the muscle and gain the time. Same tool, opposite effect.
The real trap
The danger isn’t that AI does the work badly. It does it competently and instantly, and that makes the human part feel optional. It isn’t. Empathy, judgement and accountability are what let a team trust the work. Speed without them is just faster mistakes.
So my tagline isn’t a slogan, it’s a job description. AI makes me faster. Humans keep me honest. The second sentence is the one I’d hire for.