I’d had some generous assistance giving Serve The Team’s website a visual overhaul a few years ago, but I’d never really gotten around to tightening up the marketing copy. I’d also made a few tweaks to the user interface in the time since that, while done with good intentions, left me feeling that I’d hurt the user experience a bit.

Every few months I’d come back, have a look at the website and consider making changes. But I constantly ran into analysis paralysis: what did I really need to change, and why? What was critical, what was nice to have? I needed to make a plan, but I didn’t really have the time; with a few minutes to spare I could definitely have nibbled away at a change, but making a proper plan required real time and concentration.

Enter Claude, specifically Claude Code. Despite its name, Claude Code can do a lot more than just, well, code. I fired it up in the Terminal, checked that it had access to a skill that would allow it to take website screenshots, and then got ready to put it to work. I’m including the full prompt here, in case you want to see exactly how I set things up, but the key thing to take away if you don’t want to wade through it is that I asked Claude to play two roles — Marketing and User Experience experts — and write up its findings:

We’re going to do some research work to help drive planning. Use your installed skill to open up the Serve The Team website and review — at least — the homepage and the playbook page.

For this initial review, I want you to come at this from two points of view. For the first, imagine you are my Head of Marketing. With a particular emphasis on the homepage and the other non-playbook pages, produce a critique of the approach to marketing Serve The Team. Bear in mind that the role of the non-playbook pages is primarily to encourage prospective new users to sign up, and secondarily to provide background information that builds their confidence in the value of the product, encouraging them to sign up. Please record your critique, which should also include a record of any positive aspects of the site, in MARKETING_CRITIQUE.md.

Next, focusing on the homepage and the playbook, write a critique that’s more focused on the user experience. This is especially relevant to the playbook pages. In this scenario you are playing the role of my Head of Product Design. The intent of the playbook is a) for non-subscribers, to help them understand what the product does and encourage them to sign up, b) for both subscribers and non-subscribers, to help them find content that gives them advice or tools they can use to address a leadership challenge they may be facing. Please record your critique, which should also include a record of any positive aspects of the design, in USER_EXPERIENCE_CRITIQUE.md.

I specifically told it where to put the output as I wanted to make sure I could easily find it again, and because I didn’t want it randomly creating new directories in the STT source code. But that’s the only vaguely technical advice I gave. It ticked away for just six minutes before giving me this summary:

Marketing critique — 11 issues, 7 priorities:

  • The biggest problem is the hero: “Put your team first” is a values statement, not a value proposition. A first-time visitor can’t tell what the product is. The sub-copy compounds this with internal language (“thought starters”).
  • “Hundreds of” is weak social proof that shrinks the implied audience.
  • The sign-up page is severely under-powered — it asks for an email with almost no reinforcement of why to give it.
  • The About page’s anonymous founders section and decade-old stats undermine credibility.
  • The “What We Do” page references subscription tiers that no longer exist.

UX critique — 12 issues, 9 priorities:

  • The biggest problem is that there’s no sign-up prompt on the playbook index — the main conversion surface for someone browsing the product has no CTA.
  • The card context labels are raw slugs (“personal-development”, “one-to-ones”) — looks unfinished and is a quick fix.
  • The two filter groups (topic vs situation) are visually identical with no labelling.
  • Single-column card list on desktop wastes a lot of screen real estate — a 2-column grid would halve the scroll distance.
  • Card pages have no “back to playbook” link, so users arriving via email or search are stranded.
  • The sign-up button on card pages is orange — inconsistent with the teal palette everywhere else.

If you’re curious, you can find the full reports here (Marketing) and here (UX).

It took me five times longer to actually read and digest the full reports than it did for Claude to research and write them. They weren’t perfect — it reported on a page that wasn’t actually linked to on the website, for example, because I’d deliberately unhooked it when I got rid of paid subscriptions — but overall they were genuinely solid. Would I have gotten more in-depth critique and advice if I’d paid for it? Almost certainly. Could I afford to? Absolutely not.

I had a little bit of back and forth, clarifying some of its feedback, dismissing one or two things that I fundamentally didn’t agree with, and encouraging it to be meaner about one or two areas I felt it had overlooked, before asking:

I’m nearly ready to wrap up for today, but to set ourselves up well for next time, please can you review the two documents and use them to build out a detailed plan.

Two minutes later — and less than an hour after we started — it had created a comprehensive plan of attack to address the issues in the reports. Compare this to where I was before I got a Claude subscription. The difference isn’t that Claude did this work faster than I could have. It’s that this work wasn’t getting done at all. I’d been staring at my website for months, knowing it needed improving, but unable to find the energy even to figure out what to improve. Claude’s audit didn’t replace something I was already doing slowly; it made possible something I’d almost given up on entirely. If you’ve ever had an initiative at work or a side project that couldn’t get off the ground because the upfront analysis felt too expensive, it’s time to look at it again.

I really did have to wrap up that session when I did, but I’ve come back several times over the past month to step through the plan, with Claude doing most of the driving, and the website is now much stronger on both a marketing and UX front. If I’d been working with an engineering team then I could have quickly handed the plan off to them, but in practice Claude was able to deliver most of the implementation work too, in a total elapsed time of maybe a day. Most of that was me thinking about choices around content or style that I didn’t want to outsource to the AI.

I included the prompts because this kind of format worked well for me here and in other contexts, and you’re welcome to try it yourself: give the AI a specific role, a clear lens to evaluate through, and a structured output format. I used “Head of Marketing” and “Head of Product Design,” but you could just as easily ask for a critique of your onboarding flow, your internal documentation, your pitch deck, or your team’s runbook. The key is that a named role gives the AI a point of view, which produces much sharper feedback than a generic “review this for me.”

Take a stab at it yourself — what have you been meaning to improve but keep putting off? Pick a page, give Claude a role and a brief, and see what comes back. You might be surprised how much a fresh pair of eyes that “knows” what good looks like can unblock you.