The phrase that kept catching me in client meetings was “we just want someone to handle it”. It was almost always said quietly, near the end of a scoping call, after the price had been talked about and the timeline had been laid out and the client had run the maths in their head and worked out that they couldn't quite afford either. They didn't want a quote. They didn't want a kickoff call. They wanted a website. Working. Now. Tended.
For twenty years I ran web agencies, including one I scaled to twenty-five staff. I sat in hundreds of those meetings. The clients we ended up working with were the ones who could afford a five-thousand-pound build and a two-thousand-pound retainer. The ones we couldn't — the florist on Mare Street, the plumber in Chiswick, the small-batch coffee roastery — went off to fight Wix and Squarespace, or paid a freelancer who quietly stopped returning emails six months later. I'd watch them leave the meeting and think about it for the rest of the week.
i.The clients we kept turning away
By 2022 we were turning away nearly half the leads that came in. Not because the work was bad. Because the maths was bad. A small UK business with a £150,000 turnover couldn't justify a five-figure invoice for a marketing site, however good. They'd go to the cheaper option, the site would launch, and then the part nobody had budgeted for would start: the slow rot. Hosting that wasn't monitored, content that wasn't updated, plugins that became security holes, SEO that quietly slipped, photographs that began to look dated. The site they'd paid for would be costing them money they couldn't see.
My agency would offer a maintenance retainer to fix that — usually £2,000 a month for a working set of hands. The clients who could afford it took it. The ones who couldn't — the majority — went without. Three years later, half of them would call back, asking if we could rescue the site that had quietly stopped working.
I sat in hundreds of meetings and watched the same pattern. The site launched, and then it rotted.
What I wanted to be able to offer those leads was a different shape of service. Not a project with a retainer attached. A subscription. The same care, the same editorial standards, the same tending — but priced like a phone bill, not a salary. The maths didn't work, though, because every part of the work — the design, the copy, the photographs, the structure, the SEO — needed real human time. You can't bill £15 a month and pay a designer to handle it.
ii.What changed in 2025
By the end of 2024 the AI-generation tooling crossed a threshold I'd been watching. Up to that point, AI-generated websites had a tell — a flatness, a templated feel, a copy register that smelled of a chatbot. The tooling could produce a site, but it was a site that any visitor would clock in three seconds. By the spring of 2025 that was no longer true, if you set the brief carefully and the post-generation quality work was rigorous.
More importantly, AI tooling crossed a productivity threshold for a single experienced operator. The work I'd previously needed a team of twenty-five to do — across design, copy, brand, SEO, infrastructure — became something one person could orchestrate, if that person had two decades of agency experience to draw on. The maths shifted. A subscription website service became technically possible to operate. The question was whether it could be done at editorial quality.
Editorial quality: the standard a small London magazine would hold itself to. Considered typography, real photographs, body copy that reads like a person wrote it, layout that breathes, and a point of view about what's worth saying. Not stock-photo, template-y, AI-Slop quality.
I spent eight months building the platform — the orchestration, the quality checks, the conversational editing tool, the autonomous improvement engine, the hosting infrastructure. The kind of operating system I'd have wanted on day one of every agency project I'd ever scoped, made into a service.

Plate 01. Chris Painter, at the desk · Brighton, May 2026
I work from a small studio outside Brighton, with the sea half a mile down the road. The tooling sits in front of me like a printed list of jobs that need doing, and the platform behind it does the work I'd previously have done by hand. The dispatch desk, I started calling it, half-jokingly, until the metaphor stuck. A desk where briefs come in, work goes out, and the operator's job is to make sure the standard holds.
The first Themesmith customer site shipped in February 2026. By April there were a hundred of them, all internal demos generated to test the platform across as many industries as I could think of — florists, plumbers, ceramics studios, law firms, coffee roasters, chiropractors, jewelers, restaurants, a CrossFit gym, an indie game studio. I wanted to know whether the quality held up across registers I'd never personally designed for.
It did. Different shops, same service.
iii.What we built
Themesmith builds, hosts, and continuously improves business websites. The shape of it has six moving pieces. The first is generation: a customer writes a paragraph about what they're making, and the platform produces a complete site in roughly twenty-five minutes — architecture, copy, type, palette, photographs, layout, navigation. Different visual register every time; no template folder, no drag-and-drop.
The second is conversation. After the site is live, an in-product assistant called Forge handles edits in plain English. The customer describes what they want changed; Forge proposes the change and shows a diff; the customer approves, edits, or rejects. Every change is logged and reversible. No website-builder vocabulary required.
The third is improvement. The platform watches the site and, on the highest tier, ships small improvements while the customer sleeps — clearer call-to-actions, better headlines on pages that aren't converting, fresh content. The customer sets the dial: from suggestions you accept by hand to fully automated changes with safety nets in.
The fourth piece is quality. Every site goes through automated checks before it can be published — the kind of small things that quietly break websites and that a Lighthouse audit doesn't catch. Buttons that look right but don't work on a particular browser. Animations that should play but don't. Contrast that fails on a Saturday-morning iPhone. Headings that skip a level after a CMS migration. Editorial review on top.
The fifth is infrastructure: enterprise-grade hosting, automatic SSL, daily backups, DDoS protection, a global delivery network. The standard the well-known names run on, included in every tier. The customer never has to think about it.
The sixth — and this matters more than you'd think — is ownership. The customer owns their domain, their content, their site. We earn the subscription each month by being worth it; we don't trap anyone. If they cancel, we send them a static snapshot of the site so they can host it anywhere. No lock-in by design.
iv.How decisions get made
Every product decision passes against four constraints. Quality: ship complete, not fast. We don't trade quality for speed. If something needs longer, it gets longer; the typical cycle is short enough that we don't need to compromise. Automation: if an agency does it every month — content updates, SEO tuning, performance monitoring — it should be a software workflow, not a billable hour.
Control: the customer owns the dial. From AI suggestions to fully automated improvements, your call, your pace. We never apply something that can't be undone. Outcomes: we measure real engagement, not vanity metrics. Publishing cadence, not feature count. Whether the site actually works for the person using it — not whether it looks good in a design-team review screenshot.
No exceptions. When something proposed for the roadmap can't be made to honour all four, we don't ship it.
v.The version of agency I always wished I could offer
The agency model I came up in was good at one kind of customer and bad at another. Good at the £30,000 brand-redesign project for an established business with a marketing director. Bad at the £150-a-month maintenance need of a working small business with no time. The first kind paid the bills; the second kind — the majority of the leads I ever spoke to — went away with nothing.
Themesmith is the version of agency I always wished I could offer to the second group. Same standards. A fraction of the cost. Editorial quality, by software, run by one person who's done this work for two decades and is determined to bring the bar with him.
If you're one of those leads I would have turned away in 2022 — write a brief. We'll have a kettle on.
Editor's noteThemesmith.ai Ltd opens to a first cohort of customers in May 2026. Pricing starts at £15/mo. The first hundred customers receive the founder offer (Business £29/mo, Autopilot £79/mo) locked in for the first twelve months. Begin a brief at begin a brief or read the full FAQ at frequently asked.