AI

Every studio calls itself AI-native now. Here's what actually changed in ours.

Every studio calls itself AI-native now. Ours too.

So my co-founder and I mapped the whole process - what changed, what didn't. The map is in the visual.

A hand-drawn process map titled "how an AI-native design studio works", split into human (warm) and AI (blue) columns. Section 1, the pipeline from brief to shipped, covers brief and presale, concept, design, build, QA and production - the human column holds client conversation, the founder in the room, hand-drawn pages and architecture decisions; the AI column holds meeting notes into context, sitemaps and wireframes in minutes, UI-kit extraction, code assistance and prompt-driven deploys. Below, a 2020 vs 2026 comparison of team shape and what happens after launch. Then "what we maintain" - 100+ shared skills, guardrails, one shared context, a person on every outcome - and "the numbers": wireframes from days to minutes, a creative marketing page from 84 to 16 hours, 100+ skills, 200 pages in half an hour.
The whole pipeline, brief to shipped - what the human does and what AI does at every step.

If you hoped it's “one prompt - whole website” - no. There's no magic like that in our process.

What AI took

Meeting notes go straight into context, sitemap in minutes (used to be days on wireframes), code assistance throughout, deploy + rollback by prompt.

What it didn't take

Concept. We went more analog there, not less - pencil, collage, clay, pastel. Still involve illustrators and artists. Because what? AI gets average, and we aim for above that.

And after launch clients can edit and vibe-code the site themselves, on the core we built. No more marketing wishes waiting in the queue.. (we call it WordPress Fuel)

Three things we haven't solved yet

//1 Freedom vs risk

We can launch 200 pages in half an hour. We can also drop the whole site with one wrong prompt. Inside the team guardrails work. Giving them to clients without babysitting - not yet.

//2 Quality ceiling

100+ skills, context for every project, design.md so AI doesn't invent new colors - and the output is still generic without a person. Maybe models will catch up. For now the last 1% is drawn by hand.

//3 Pricing

Everything got faster and we still charge hourly. So every automation we build cuts our own invoice. Funny. We're moving to another model now.

Agency ppl - when the hours disappear, what do you charge for? Fixed, value, subscription for the system? Tell me what works for you.

See more Norml in Google Search

Add Norml as a preferred source
Share
Zhenia Mohyla
Written by

Zhenia Mohyla

Co-founder · Design & Strategy

I lead design and strategy at Norml, shaping client work, internal systems, and the studio’s creative direction.

LinkedIn ↗