AI is a research assistant, not an art director
Where generative tools genuinely earn their place in a studio workflow, and where they quietly cost you.
Category,
AI
Mara Ellison
,
Generative tools are now genuinely useful in a studio, but almost never at the point people expect. They are weakest exactly where the work is most valuable: deciding what the thing should be.
Where it pays
The reliable wins are upstream and downstream of design. Synthesising forty stakeholder interviews into themes, drafting a hundred naming candidates to react against, generating alt text, catching inconsistencies across a large set of guidelines, writing the boring first pass of documentation nobody wants to write.
In each of these, the model is doing volume work against a standard you set. You remain the one who decides what good looks like, which is the part that was never the bottleneck to begin with.
Where it quietly costs you
The cost is homogenisation. A model trained on everything returns the centre of everything, and the centre is where every competitor already is. Used early in concepting, it narrows the range of things you consider before you have noticed it happening.
