AI disclosure is a design problem
Since 2 August 2026 the EU AI Act requires you to tell people when they are dealing with a machine. Most teams will answer that with a footer, which is the one place nobody looks.
Sectors / Compliance and regulated software
Trust is the product. We design it in.




Why compliance
When the product is about risk, trust is the product. We design compliance, payments and regulated software so the cautious buyer can see it is safe above the fold, and the person using it every day can get the job done without a manual. Compliance software is sold to people whose job is to avoid a problem: an audit, a fine, a worker paid wrong, a sign that does not meet the regulations. They are not comparing feature lists. They want to know whether this product will keep them safe, whether it will hold up when the regulator asks, and whether their team will actually use it. Most compliance products are written as if they were feature lists anyway. The site leads with what the software does rather than what it protects against. The product shows every field and every setting because the engineers who built it know what each one is for. Trust signals sit in the footer. The result is a category that looks complicated and feels risky, which is the opposite of what it is selling. We design the other way round. Proof first: the volume handled, the checks done, the security, a named customer. Then the structure, so a cautious buyer can find the one claim that matters to them without a demo. Then the product itself, organised around the task (the pay run, the check, the lookup) rather than the database.
What we have built
Six hundred pages of UK law made searchable: search first, structure second, one screen instead of three tabs, with the regulation text, the diagram and the project side by side. 440 hours saved per team, per year.
Right to Work, DVLA and CIS checks, document expiry and the pay run, for logistics carriers and construction firms with self-employed workforces. More than 240,000 self-employed people paid through it. We rebuilt the website around two audiences and a compliance-first position, in time for the rebrand.
Customer panels for hosting providers, where a bad onboarding flow creates abuse, compliance and support load downstream. Designed from the information architecture through every state, with onboarding before the homepage.
Freelogo, an AI logo platform, designed so the user can see what the model did, how sure it is, and how to override it. The same human-in-the-loop pattern applies to any regulated product adding AI.
What we already know
We arrive fluent. Why reversibility, human sign-off and refusal states are design work, not legal's problem. Where a security claim belongs on a page and how to make it checkable rather than decorative. What a deadline does to a buyer (Wise's construction page was designed first because every construction firm was reacting to the 1 October 2026 Right to Work change) and how to use it where it is real rather than as a gimmick. How to show the regulation, the data and the decision on one screen so the person using the product can defend what they did. If your product needs a glossary and a statute before a designer is any use, that is our normal starting position.
The numbers
Every figure below has a client name on it and a case study behind it.
Testimonials
I recommend Grid Design. Sy and his team are brilliant at what they do. Fast, highly skilled, and extremely efficient. They are a rare diamond.
Ash MiriFounder, Air Landline & TSRGDReally impressed with the quality of Grid. Everything was backed with science and data, a strong review process, and deep sector understanding.
Andrew KillenCTO, ServeboltFAQs
Yes, and it is a growing share of what we do. Statutory regulations (TSRGD), workforce compliance and payments (Wise), and the hosting platforms where abuse and compliance load is designed in or out at onboarding. The thread is products where trust is the thing being sold.
That is where we are most useful. Audit trails, reversibility, human sign-off, refusal states and plain-language explanations are all design decisions, and designing them early is what gets a product through legal rather than stopped by it. We would rather have legal in the first workshop than the last review.
We design with realistic but invented data, never your customers'. Where research needs real users, it runs under your NDA and your data rules, and nothing leaves your systems. We will sign what you need before the kickoff.
A full engagement, from discovery through research, architecture, prototypes, interface and design system, is a twelve-week project priced as a fixed scope. We give you the number at proposal stage, before you commit to anything.
Yes, and for a regulated product it is usually the only sensible route. We rebuild the architecture underneath the screens so your team ships the new version in stages, each one checkable, rather than switching everything off for a big-bang launch that nobody can audit.
Yes, if the interface shows the working. What the model did, how confident it is, and how a person checks or overrides it. Reliability sets what the system may do alone; it does not decide whether you can ship. We design the human-in-the-loop steps first.
Related reading
Since 2 August 2026 the EU AI Act requires you to tell people when they are dealing with a machine. Most teams will answer that with a footer, which is the one place nobody looks.
From January, UK subscription rules mean a reminder before every renewal and an exit as easy as the sign-up was. Most hosting companies will send that to Legal, when it belongs to whoever owns the billing screens.
Most of your buyers decide before they fill anything in. The useful question is not how to capture them, it is what they cannot find on the site.