AI is not good or bad. It is good at some things and bad at others. Knowing the difference saves you a lot of wasted time.
What it is genuinely good at
- Reading huge amounts of text without getting tired.
- Applying the same tagging rules across thousands of items.
- Summarising long documents that you give it.
- Spotting surface patterns in large piles of data.
- Writing first drafts, so you never face a blank page.
- Translating and adjusting tone, patiently, as many times as you like.
What it is genuinely bad at
- Deciding which findings actually matter to the business.
- Understanding your company, your history and your politics.
- Checking its own facts without outside tools.
- Pushing back, or defending an unpopular truth.
- Coming up with truly new ideas outside existing patterns.
- Knowing whether it actually finished what you asked.
Look at the two lists side by side. The first is about speed and volume. The second is about judgement, context and responsibility.
The fast intern rule
When you are unsure whether to hand a task to AI, ask yourself one question:
“Would I give this to a very fast, very capable intern who knows nothing about our company, our history or our strategy?”
If yes, hand it over. Reading, tagging, organising, drafting and summarising all pass this test.
If no, keep it. Strategy, sensitive negotiations and high-stakes trade-offs do not. There is a practical reason too. To check the machine’s answer on these, you would have to do all the hard thinking anyway. So you saved nothing.
How to brief your intern
A good intern still needs a good brief. The same goes for AI.
- Give it the material. Do not ask it to guess. Paste the notes, tickets or documents.
- Say what good looks like. “Group these by underlying problem, with ticket IDs” beats “summarise this”.
- Set the limits. Tell it what to ignore and what is out of scope.
- Plan the check. Before you start, decide how you will check the result.
A small exercise
List the tasks you did last week. Next to each one, write yes or no for the fast intern question.
Most designers find the yes list is longer than they expected, and the no list is more valuable than they realised. Both of those are good news.
The one line to remember
Treat AI like an intern with a photographic memory and no stake in how things turn out. Brief it accordingly.
The same thinking applies when you point AI at the design itself. That is next: AI can draw the screen, you still own the design.