AI
Using AI in creative work responsibly
How to set boundaries, check output and stay open with clients when AI tools are part of a project.
Decide what AI is for. Be specific about where it helps: early concept options, extending a background, drafting alternative copy, cleaning up an image. Equally, decide where it does not belong. Writing these boundaries down before a project starts keeps the team and the client working from the same understanding.
Treat output as a draft. AI-generated images and text can contain odd details, wrong facts and flaws that are easy to miss at first glance. Someone on the team should inspect every output closely, at full size, before it goes into a layout. Review is part of the job, not an optional step.
Check rights and terms. Tools differ in what they allow for commercial use, and those terms change over time. Read the licence for each tool you use and keep a record of which tool produced which asset. If a client needs clear ownership, say early what can and cannot be guaranteed.
Protect confidential material. Do not paste unpublished figures, client documents or personal data into a tool without knowing how that input is stored and used. Where the terms are unclear, use anonymised or placeholder content instead. It is a simple habit that avoids a serious problem later.
Be open with clients. Tell them when and how AI is used in their project, and let them opt out of it. Most people accept the answer when they hear it early. Surprises are the problem, not the tool itself, so put the conversation in the first proposal rather than the final handover.
Keep human judgement on the final look. AI tends to produce generic, familiar results. A designer's role is to choose, edit and combine so the finished piece fits the brand and says something specific. If an image could belong to any company, it needs more work or a different approach.
Watch for bias and misrepresentation. Generated images of people, places and cultures can lean on stereotypes or show things that are simply inaccurate. Review output with this in mind, particularly for work about real communities, real products or real data. Use real photography where accuracy matters.
Revisit your approach regularly. Tools, laws and client expectations are shifting. Review your own guidelines every few months, note what worked and what caused trouble, and update them. A short written policy that people actually use is better than a long one that sits unread in a folder.