Speculative Web Builds for SMBs¶
Side hustle with James. Use AI to research small businesses with poor websites, speculatively build them a high-quality static page, then cold-send it with an offer: "We'll finish this properly and be your ongoing tech team for £500" (or whatever the number ends up being).
The Core Bet¶
Most small businesses have terrible websites - bad templates, outdated content, zero personality. They know it's bad but don't know how to fix it or can't justify the cost of a traditional agency. The usual sales pitch is "we could build you something great" - but that requires imagination from the buyer.
The insight here is that showing beats telling. You don't describe what you'd build; you build it and put it in front of them. A static HTML page that looks genuinely good - not functional, not hooked up to anything, but clearly a step change from what they have now. The prospect sees exactly what they'd get. That's a fundamentally different sales conversation.
Why This Works Now¶
This model would have been economically insane two years ago. A traditional agency can't afford to speculatively build pages for leads that mostly won't convert. But with Claude doing the heavy lifting on research, copywriting, and code generation, the per-lead cost drops dramatically. If you can produce a credible page in 30-60 minutes of combined effort, the maths starts to work even at modest conversion rates.
You and James have the specific skill set this needs: comfortable with AI tooling, can manage repos and deployments, understand what "good" looks like for a small business site. You're selling a capability gap - most SMB owners can't do this and don't know anyone who can do it cheaply.
Unit Economics (Needs Pressure-Testing)¶
The whole thing lives or dies on conversion rate vs. effort per lead.
- Speculative build time: ~30-60 mins per lead (AI-assisted research + page generation)
- If you send 20 per week and convert at 5%: 1 sale/week = ~£2k/month each
- If conversion is 2%: you need 50 sends/week for the same result, which changes the effort calculus significantly
The initial £500 (or whatever) is the hook, but the recurring value is in the "we'll manage everything" retainer. Hosting, domain, updates, email setup - the stuff that's trivial for you and mystifying for them. That's where the compounding revenue lives.
Open Questions¶
- Lead sourcing: How do you find the right targets at scale? Google Maps scraping for businesses with poor websites? Industry directories? There's a whole pipeline to build here before you send a single email.
- Cold outreach legality: B2B cold email is broadly legal under UK PECR (Privacy and Electronic Communications Regulations) with legitimate interest, but needs care. You're not spamming consumers - you're emailing businesses. Still, the framing and opt-out mechanism matter.
- What "managed service" actually means: Hosting where? Updates how often? Do you handle domain registration? Email setup? The scope of ongoing support needs defining before you can price it.
- Pricing sensitivity: £500 for a finished website plus ongoing support is aggressive. Could be a brilliant hook or could attract only the most price-sensitive customers who'll be the most demanding. Worth testing a range.
- Spec work perception: Some businesses will love receiving an unsolicited mockup ("wow, someone cared enough to do this"). Others will find it presumptuous or spammy. The framing of the outreach email is everything.
Connections¶
- Angle-First Generation for Vibe Design - the angle generation workflow could power variant creation for speculative builds. Generate 50 angles on what a business's site could be, verify diversity, flesh out the best 3, and send the strongest.
Smallest Testable Version¶
Before building any automation, pick 10 local businesses with visibly bad websites. Build each a page manually using Claude. Send a personalised email. See what happens. You'll learn more from 10 real sends than from weeks of building a lead pipeline. If nobody bites, the idea needs rethinking. If 1-2 respond, you know the core mechanic works and it's worth investing in the tooling.