We run a menswear brand with AI - and the line we won't cross

 

If you want to build a company AI-first, the first thing to get right isn't the tools. It's the line. AI should run the everyday operating work of the company; people should own the product and every decision that defines its quality. Hold that line and the rest follows. We're Aarsag, a small premium menswear brand building the company this way from the start. This post is the principle behind it; the five steps to make a company AI-first is the practical how.

Key takeaways

• "AI-first" means AI runs the operating work (drafting, organising, summarising, tracking) while people own the product and every quality call.
• The split is the whole point: hand off the legwork and people get more time for the work that decides quality, not less.
• It doesn't take a platform or an engineering team. Ordinary tools (mostly Claude Code, a bit of ChatGPT) are enough to start.
• Everything that ships past AI is read by a person first, because the model gets things wrong in ways that read as right.

What does AI actually run, and what stays human?

AI runs the everyday operating work of the company. It drafts emails, posts, and product copy for a person to edit and approve. It summarises long threads so a decision isn't buried in a chain. It keeps the company's knowledge organised in our company OS, a single Git repository that collects everything the company knows in one place, so the right context is one search away. It tracks tasks and writes up decisions so the reasoning doesn't vanish into chat. And it proposes website changes as code we can review before anything goes live.

People own everything that decides whether the product is good. The fabric weight. How a collar sits after a wash. Whether a prototype is genuinely right or merely fine. The call to wait for a better version instead of shipping the one in hand. The relationship with our makers in Portugal who build the long-sleeve polo. A model can describe these things; it can't be accountable for them, so it isn't.

That division isn't a limit we're apologising for; it's the reason the work gets done well. Most small brands don't cut corners on quality because they stop caring. They cut them because the founders are buried in admin and never reach the slow, careful work the product needs. Handing the legwork to AI is what protects that time.

AI handles the everyday running of the company: drafting, organising, summarising, tracking. People and our makers in Portugal own the product, the taste, and every quality call.

Is this real, or just a claim?

Fair question. Every brand says it "uses AI" now. So, concretely: this post was drafted by an agent and edited by a person before you read it. That is the model in miniature. The agent did the legwork; a person made the calls, cut the lines that sounded right but weren't, and decided it was ready.

That last part matters, because the honest truth is the AI gets things wrong. It over-reaches. It writes confident sentences that fall apart on a second read. None of that is a reason to avoid it. It is the reason a person reads everything that carries weight before it ships. The system is built around that fact, not in spite of it.

Could anyone build a company this way?

Largely, yes, and that is the part worth saying plainly. We didn't buy a platform or assemble an exotic stack. The company runs out of that one repository, which the AI reads before it does anything, with a clear rule about what it may draft and what only a person may approve.

The tools are deliberately ordinary. Most of the work runs through Claude Code, with a bit of ChatGPT for image generation. No budget, no engineering team, nothing you couldn't get your hands on today.

The line we won't cross

So that's the whole of it: AI runs the company; people and our makers in Portugal own the product. We think that's the right way round. Done honestly, the work behind the scenes reinforces the quality of the garment rather than working against it, because it buys back the hours quality actually needs.

This is the start of a series on how we build and run Aarsag with AI; more to come.

FAQ

Does AI design the clothes? No. AI never designs, picks fabric, or judges fit. People and our makers in Portugal own the product and every quality call; AI handles the operating work around it.

What does "AI-first" actually mean here? That AI runs the everyday operating work of the company (drafting, organising, summarising, tracking) while people own every decision that carries weight. A way of working, not a product you buy.

What AI tools do you use? Ordinary, off-the-shelf ones: mostly Claude Code for the agent work and a bit of ChatGPT for image generation. Nothing specialised.

Are the product photos AI-generated? The current launch visuals are AI renders standing in as placeholders. Anything that influences a buying decision is replaced with real photography before the product is buyable.

---

Next: the five steps to make a company AI-first, the practical setup.