Figma asked 2,500 designers and developers how AI was changing their work. The answer that stuck with us was not the headline adoption number. It was the gap between two others: 78% of designers said AI makes them more efficient, but only 47% said it makes them better at their role, and just 32% said they trust the output (Figma, 2025).
That gap is the whole story of AI in marketing design. Speed is easy to buy now. Quality is still earned.
The two layers of every design job
Look at any piece of marketing design, a webinar banner, a sales one-pager or a quarterly deck, and it splits into two layers.
The judgment layer is the thinking: what the piece is for, what the viewer should notice first, which message leads, and whether it feels like your brand or like everybody else’s. This is where a senior designer earns their fee.
The production layer is everything that follows once those decisions are made: the fifteen ad sizes, the dark-mode version, the LinkedIn crop, the alt text, the export settings, the file naming. It is necessary work, and it is where most of the hours disappear.
AI is excellent at the production layer and unreliable at the judgment layer. Build your process around that and you get the efficiency without the sameness.
What to hand to AI today
These tasks are low-risk, high-volume and easy to check:
- Resizing and reformatting. Turning one approved master into a set of placements is exactly the kind of rule-following that AI tools now handle well.
- Variants for testing. Swapping headlines, images or calls to action across an approved layout so you can test more than two versions.
- Image clean-up. Background removal, extending a photo to fit a wider crop, fixing lighting on product shots.
- First-draft copy and alt text. Useful as a starting point, never as the final word.
- Asset tagging. Describing and tagging a library so people can actually find the right photo next quarter.
Nielsen Norman Group’s analysis of three studies found an average 66% improvement in throughput when people used generative AI for realistic work, with the best results where AI drafted and humans edited (NN/g). That “draft, then edit” pattern is the one to copy.
What to protect
Some decisions should stay firmly with an experienced person:
- The idea. AI remixes what already exists. If your campaign looks like the average of the internet, it will perform like the average of the internet.
- Hierarchy. Knowing that the proof point should outrank the product name on this particular slide is a judgment, not a rule.
- Brand voice. Tone drifts quietly. One off-brand headline is harmless; forty of them, generated in a week, reshape how people see you.
- The final check. Someone has to be accountable for what ships, including accessibility, legal claims and whether the numbers are right.
Write your AI stance down
Google’s 2025 DORA research on software teams found that a clear, communicated AI stance was one of seven capabilities that separated teams who turned AI into real gains from those who just got busier (DORA 2025). The same applies to design teams.
A one-page stance is enough. It should say:
- which tools are approved, and for which tasks;
- what can never be generated (customer logos, real people’s faces, testimonials, data);
- who reviews AI-assisted work before it goes out;
- how you label or track AI-assisted assets internally.
It sounds bureaucratic. In practice it removes the anxiety, because people know where the lines are.
A realistic workflow for a campaign
Here is how a typical campaign now moves through a small senior team:
- Brief and concept (human). One designer and one marketer agree on the idea, the key message and the hero visual.
- Master layouts (human, AI-assisted). The designer builds the key pieces. AI helps with image exploration and copy options, but the designer chooses.
- Production (AI-heavy). Approved masters are expanded into every size and variant needed. This is where hours become minutes.
- Review (human). A second person checks every output against the brief and brand rules, with an accessibility pass.
- Learn (shared). Performance data comes back, and the winning variants inform the next brief.
The time you save in step three is only valuable if you reinvest it in steps one and four. Otherwise you simply produce more mediocre work, faster.
The bottom line
AI has made marketing design cheaper to produce. It has not made good ideas cheaper to have. The teams that pull ahead will be the ones who use AI to clear the production backlog, then spend the time they get back on thinking, reviewing and raising the bar.
If you want a team that works this way, see how we approach marketing design, or read our guide to briefing designers and AI tools.
