Insights4 min read

What will survive AI in fashion

Anton Viborniy

Co-founder & CEO of Apiway

When I worked in 3D rendering before Apiway, every wave of automation followed the same pattern. Auto-rigging ate the bottom of the layer cake first — grunt rigging work disappeared overnight. Render farms killed local render-workstation specialists. AI denoising replaced denoiser experts. But the top of the cake — lookdev, creative direction, art direction — kept getting more valuable. The matte painters who painted backgrounds for Hollywood through the 1980s lost their craft. The animation directors at Pixar got more powerful. AI in fashion is replaying the same pattern. Most of fashion survives. Some of it doesn’t. Here’s the layer-by-layer map of what makes it through.

Why most of fashion survives the AI transition

Most of fashion survives the AI transition because AI catalog production only operates on one layer — the volume imagery layer of ecommerce. Everything above and adjacent to that layer continues. Garment design isn’t AI work. Pattern-making isn’t AI work. Sample development, sourcing, production engineering — all unchanged. Editorial photography (campaign hero shoots, magazine covers, runway-adjacent narrative imagery) is structurally AI-resistant because it depends on named creative authorship, cultural moment, and the unrepeatable conditions of a specific shoot. None of those are properties AI rendering produces. The cleanest framing I’ve come across: think of fashion as a layer cake. AI affects one specific layer in the middle. The layers above and below survive in shifted form but remain — not unaffected, but recognizably themselves.

What AI cannot do in fashion catalog production

What AI cannot do in fashion catalog production is generate the taste judgement that decides what to shoot in the first place. AI executes a brief. It does not generate the brief. Someone still has to decide which SKUs deserve catalog coverage, what styling makes sense, what model identity reads right for the brand voice, what cultural moment the campaign should ride. These are creative-direction decisions; the AI tool runs them but does not author them. Same for merchandising — the catalog cadence acceleration that AI enables (50 ad creative variants per week instead of 5) requires more taste-led calls per unit of time, not fewer. Merchandising and buying have actually become more important post-AI, not less. The bandwidth constraint replacing the photography constraint is merchandising bandwidth. Brands that confuse “AI fashion” with “AI brand” are misunderstanding the technology. AI fashion catalog production sits downstream of every interesting fashion decision, not upstream of it.

Why UGC and physical retail get more valuable in the AI era

UGC and physical retail get more valuable in the AI era because they’re the credibility layer AI cannot replicate. As AI catalog imagery saturates the ecommerce surface, the relative value of authentic credibility imagery rises. This is a scarcity dynamic. The more AI imagery proliferates, the more a real photo of a real customer wearing a real garment is worth in the conversion funnel. Physical retail follows the same logic — the fitting room, the salesperson interaction, the in-store curation all sit completely outside AI catalog scope and become the conversion-anchor for fits and constructions where commercial imagery (AI or traditional) doesn’t substitute for try-on. Fashion week, the editorial calendar, designer-led narrative arcs — these survive entirely unaffected. They operate on cultural logic, not commercial logic.

How fashion brands should split AI from human work

Fashion brands should split AI from human work along the layer cake — AI for the volume catalog layer, humans for everything else. Run the volume catalog (white-background flat-lay, on-body lifestyle, model variants, marketplace SKU coverage) on AI. We built White Studio at Apiway specifically for that layer — it’s where AI economics dominate 30:1 and where brand-voice templates carry the human creative direction without per-shot human labour. Keep editorial, campaign hero, and brand-voice template authorship as named-creative work. Invest more in merchandising and UGC programs — they grow in importance as the volume layer gets cheap. The brands I see operating this split cleanly outperform brands pushing AI into editorial or holding the volume layer on traditional photography. I went deeper into the volume side in my essay on the end of fashion stock photography.

When 3D rendering took over animation in the 1990s, the prediction was that animators would disappear. What actually happened: certain animator work disappeared (cel painters, in-betweeners), and lead animators became more powerful — they became Pixar story directors. AI in fashion is replaying that pattern. The work doesn’t go away. It moves up the layer cake. I made the same argument about photographers in my essay on how photographers earn in the AI era. If you operate a fashion brand and want to think out loud about which layer your team should hold, come into the Apiway creator marketplace or DM me on Instagram.

— Anton

P.S. ActiveCampaign automated the email layer of marketing back when I was reselling it in the early 2010s. Email marketers didn’t disappear — they moved up the layer cake into lifecycle strategy and segmentation. Same mechanism, different decade. 🚀