Seven design skills AI is absorbing, and five that got more valuable

Seven things the tools now do adequately

• Producing a first-pass layout from a description
• Generating variations of an existing design
• Stock-style imagery and placeholder visuals
• Icon sets in a consistent style
• Converting a static design into working markup
• Resizing and adapting assets across formats
• Writing the kind of interface copy nobody reads twice

Adequately is the right word. Each of these produces output that is competent and unremarkable, which for most of these tasks is exactly what was needed.

Five that became more valuable

1. Deciding what to build

When producing an option costs minutes, the constraint moves to choosing which option is right. Taste and judgement stop being finishing skills and become the central ones.

2. Understanding the user

No model knows why your particular users abandon a particular flow. Research, interviews and reading behaviour remain entirely human, and the designs built on them are the ones that work.

3. Systems thinking

Generated components are locally reasonable and globally inconsistent. Someone has to hold the system — tokens, patterns, the rules that keep a hundred screens coherent.

4. Making the case

Explaining why a design is right to people who will fund or build it. That is a persuasion skill operating on humans, and it gates everything else.

5. Knowing when it is wrong

The ability to look at competent-looking output and recognise that it misses the point. This is the hardest to teach and the most valuable thing a senior designer currently has.

What this implies

The junior end of the craft — production, variation, adaptation — compresses. The judgement end expands. Which is uncomfortable, because judgement was traditionally built by doing the production work that is now automated. That is a real problem for the pipeline, and nobody has solved it yet.