NewAI video is here. What should creatives actually use it for?
Tools

Building an AI-assisted design workflow, one step at a time

Where AI tools fit into a real design process, from research to production, which tasks to automate first, and where a human should stay in charge.

Creative Pouch · · 2 min read · Updated

A person drawing on a tablet with a stylus
Photo: Jakub Żerdzicki on Unsplash

The best AI workflows don’t replace a design process. They speed up its slowest, most repetitive parts and leave the decisions to you. The mistake most teams make is trying to automate everything at once. A better approach is to map your process, find the bottlenecks and add AI one step at a time.

Start by mapping your process

Before choosing tools, write down what actually happens on a typical project: briefing, research, exploration, design, review, production and handoff. Next to each step, note how long it takes and how much of it is repetitive. The steps that are slow and repetitive are where AI pays off first. The steps that depend on taste, context or client relationships are where it should only assist.

Research and moodboards

AI is good at gathering references quickly and summarising them:

  • summarising competitor sites, articles and user reviews;
  • clustering interview notes or survey answers into themes;
  • generating a first moodboard direction from a written brief.

You still decide what’s relevant. Treat every summary as a lead to check, not a fact. Language models are confident even when they’re wrong, so keep links to original sources in your research doc.

Exploration

Use image generators for wide, rough exploration early on: twenty directions for a hero image, ten moods for a brand, quick variations on a layout idea. Their value is breadth, not finish. Once a direction is chosen, narrow down by hand in your design tool, where you have full control over type, grid and detail.

A good rule: AI output should leave the exploration phase as a reference, not as a final asset. That also keeps you clear of most licensing questions.

Design and prototyping

Inside the design tool, AI helps with smaller tasks: renaming and organising layers, generating realistic placeholder content, suggesting auto-layout structure or turning a screenshot into editable layers. For interactive prototypes, AI coding assistants can produce a working HTML prototype from a design in minutes, which is often faster to test with users than a click-through mock-up.

Production

Production is where automation pays off right away, because the work is repetitive and the rules are clear:

  • resizing and exporting assets for every channel;
  • creating variants of banners, social posts and ads;
  • writing alt text and first drafts of microcopy;
  • checking colour contrast and spotting inconsistent styles.

These tasks rarely need creative judgement, but they eat hours every week.

Review: where humans stay in charge

Taste, accessibility and brand fit stay human. Always check the result yourself, especially:

  • accessibility — automated checks catch contrast issues, not confusing flows;
  • brand voice — AI copy tends towards the generic;
  • facts and claims — anything a client will publish must be verified;
  • rights — know where generated images come from and what the tool’s licence allows.

Roll it out gradually

Add one AI step to one project, measure the time saved and the quality, then decide whether to keep it. Document the prompts and settings that worked, so the whole team gets the same results. After a few projects you’ll have a workflow that’s genuinely faster, and that you understand well enough to trust.

  • #workflow
  • #productivity
  • #tools
  • #ai tools
  • #design process

More in Tools