Ten years ago I started being handed the sort of problem that arrives as a Slack message and a deadline: a product nobody has quite defined, in a domain I’m not allowed to name, with no research budget and a stakeholder who has already designed it in his head. I’ve been doing it ever since: NatWest, Audiogum, and now My PT Hub and ClubWise. It is the best job there is.
I design end to end, research through to shipped interface. But the part I’m actually good at is the part that comes afterwards: making the decision stay made. Every product I’ve joined has had the same thing rotting underneath the brief (a hundred decisions taken properly once, then quietly remade, worse, six months later). That is what a design system is for. Not a component library: a company’s opinions, made expensive to ignore.
Every design team in the country has now been told to use AI. Most have responded by generating a great many design files nobody asked for. I spent last year on the unglamorous version: taking Muse (an open-source, agent-based content workflow) and putting it into production. Brand voice, audience, banned words, ICP. Two new agents and four skills. Then the part nobody screenshots for a case study: working with engineering to wire it into our internal APIs and instrument it end to end, so campaign content is measured rather than shipped into the dark. The marketing team has been running it without me for months. That is the whole ambition, really. Less exciting than a prototype, considerably more useful.
The same logic now runs the brand itself. My PT Hub’s colour rules, tone and voice exist as a reusable Claude skill, live across the company, so that anyone making anything starts on-brand instead of being corrected into it afterwards. A design system tells people what the answer is. Brand as code means the machine already knows.
Designers are encouraged to be modest about numbers, which is convenient, because most of us don’t have any. Mine: a pricing and onboarding redesign, +10% active users. Add-Ons, +15% feature adoption. A year of platform-wide usability work, +20% retention and +25% engagement. A ClubWise onboarding flow launched into a one-day sale, which produced the highest single-day sales in the company’s history and needed no staff training whatsoever to use (and I am prouder of the second half of that sentence than the first).
Before this there was Bankline, NatWest’s corporate banking platform: £1.2 trillion in payments, 100,000 business customers, 400,000 users at wildly differing permission levels, and an interface that had, when I arrived, the warmth and navigational logic of a multi-storey car park. We made it faster to use and markedly less likely to generate a phone call. Before that, audio interfaces at Audiogum, licensed to manufacturers who each needed them to look like their own idea.
Most of my work is under NDA, so I can describe it only in the manner of someone recounting a dream: it was very good, there were tables, you had to be there. What I can tell you is the pattern. I like simple, elegant solutions to complicated problems, and I like leaving a system behind rather than a screen. Something that carries on working long after I’ve stopped being in the room to defend it.
Claudia Loch 