How we use AI in the studio, and where we don’t
Engineering
By Ali
Clients ask us about AI every week, so here is exactly how we use it today.
Where it helps
- Summarising interview transcripts so researchers can spend more time on patterns
- Drafting test data, boilerplate code and first versions of documentation
- Generating many layout or copy variations to react to, never to ship as-is
- Checking code for accessibility and security issues alongside human review
Where it does not
We do not use AI to replace customer research, to make final design decisions or to write words that customers will read without a human editing them.
How we protect your data
Client data never goes into tools that train on it. We use business accounts with data retention switched off, and we tell you which tools touch your project.
What it means for budgets
Some tasks are faster, so sprints include more testing and polish. The expensive part of a product was never typing; it is deciding what to build.
How we review AI-assisted work
Anything a tool helped produce is reviewed by the person who owns that part of the project, exactly like work from a new team member. Code goes through the same tests and peer review; copy is edited for accuracy, tone and accessibility; research summaries are checked against the original recordings.
What we tell clients
Our proposals list the tools we plan to use and why. If your organisation has its own AI policy, we follow it, and if you would rather we use none on your project, we can do that too.
AI is a good assistant and a poor art director. We plan to keep it that way.