AI web design, for the companies building the AI.

Whether you searched for AI web design or web design AI, the phrase gets used for two completely different things. One is a tool that generates a website for you — a builder or a generator, which is what most searches for AI web design are after, and which this page is not. The other, ours, is design and engineering for companies whose business is AI: a machine-learning engineering studio, a genomics platform, a computer-vision product. If you want the generator, any of the builders will do it in an afternoon and it will look like everyone else’s. If you are selling a hard technical capability to people who can see through a template, that is the work described here.

For a simple marketing site, honestly, sometimes. A generator will produce something coherent and cheap, and if the site’s job is to exist, that is a reasonable answer. Where it stops working is the part this page is about: deciding what the architecture should be, explaining a capability a general model does not understand, and giving a skeptical engineer specific evidence rather than confident-sounding copy. We use AI in our own production where it earns its place, and we publish what we measure about it — but a person decides the structure and is accountable for what ships.

Because they still look, and a thin site actively costs you. Coxit sells custom AI and ML development to technical founders and VP Engineering buyers on long sales cycles, and had already been through more than one attempt at a marketing site that never found the balance between polish and clarity for that audience. The fix was not more persuasion — it was giving each named engagement its own template so the evidence could be specific.

Yes, and they are separate jobs that can be done separately. For Stevens Industries the product was the work: a platform where estimators upload architectural PDFs and get automated counts, with an AI chat for querying the documents and a correction loop that feeds user feedback back into the model. Their manual review used to take two to six hours; with the model in place it runs in under ten minutes.

Then it is better to find that out before the contract. On Stevens we spent over fifty hours in evaluation and research before signing — testing available computer-vision approaches against the actual complexity of architectural drawings, which would have overwhelmed a conventional model without pre-processing. Pre-discovery is cheap compared with building the wrong thing.

It is the common starting point and it is fixable within the same engagement. Coxit had a logo and a couple of flat colours with no visual system behind them, which is not enough to carry an enterprise conversation. We extended it into photography, an illustrated icon set, gradients and card modules, applied consistently from the site through to printed collateral.

It depends on how specific the claim is. InheriNext, a genomics AI platform, went from no organic visibility to first-page rankings in four months — by publishing with specialist writers in a narrow field rather than competing on generic AI terms. Broad AI keywords are dominated by tool listicles and are usually not worth the fight.

Norml publishes its rates on Clutch — $50–$99 per hour with a $5,000 project minimum. A full engagement covering architecture, design system, content and build sits well above that minimum, and product design is scoped separately. We scope after discovery, not before.

You can run it yourself — the handoff includes the files, the system and a session with your team, and the structure is built so new service or industry pages are an extension rather than a rebuild. When a team would rather not, Norml Care covers updates, monitoring and small changes on a monthly plan.