AI is now the default in ad creative production. Kim Hye-jin, a 14-year designer and deputy team lead at creative.Lab inside Korean ad-tech firm Mobidays, told Mobiinside that “there are almost no cases where we don’t use AI in day-to-day work.” Yet deciding what makes a good result, she says, is still a human job.
When she joined in 2016, everyone was a graphic designer and there were no sub-teams. Today the group is split into graphics, video and copy. As ad creative expanded from single images to video and copy, the expertise split too.
AI has sharply cut the time spent on tasks that used to drag: cutting out subjects, generating missing backgrounds for cropped images, creating objects that once lived only in the imagination, and producing composite shots that look professionally staged.
Banners generated end to end by AI, including text, still look obviously AI-made. So her approach is to use AI for source material and handle text and anything that needs clean finishing by hand. That, she says, is how to produce high-quality ads.
Using AI doesn’t guarantee good work. The ability to judge whether a generated image works, whether it looks off, and what still needs fixing has become more important.
Good creative still connects to the brand and moves people. Not a hook disconnected from the brand, but an ad that is unmistakably the brand while being funny, surprising, cute, relatable or desirable.
She also stresses collaboration. When planners, marketers or clients describe what they want vaguely, designers should point out what’s unclear first. That speeds up direction and reduces revisions.
Company AI workshops and study groups helped her team test tools and find the ones that fit real work. In five to ten years, she expects AI skills to be a baseline, with designers managing AI like teammates whose strengths and weaknesses they know.
Use AI to generate backgrounds, objects and composites in volume, and keep copy layout and brand elements with people. For an agency that cut workflow time with in-house AI, see Building an In-House AI Platform at an Ad Agency.
Define what to check in AI output, such as brand colors, tone and telltale artifacts like distorted hands or text, so quality decisions don’t vary by reviewer.
When AI makes sources fast, the bottleneck shifts to knowing what you want. Clear references and emotional goals cut revision rounds. For a lean, AI-native marketing team model, see How Liquid Death Is Building an AI-Native Marketing Organization.
Time-consuming source work such as cutting out subjects, generating missing backgrounds, creating objects and producing composite shots.
When text and layout are also automated, the result still looks noticeably AI-made. Finishing text and clean details by hand raises quality.
Judging whether AI output works and what still needs fixing, plus communication skills to pin down what collaborators actually want.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
Buzzvil's BD strategy team has zero developers but runs 32 AI workflow automations — from 7-minute revenue drop analysis to 90% fewer manual steps. How they scaled from personal hacks to team systems.
Meta’s Muse Wonder and Muse Wonder Full add Active Memory and Active Focus so its AI agent can handle tasks that take days. But a Toronto user’s Marketplace mishap shows the risks of agentic AI.
Pew cloned thousands of panelists as AI digital twins and found a 12.4-point average error. Five ways synthetic respondents fail — and where marketers can still use them in research.