Prospector OS
Finds and qualifies local business leads automatically.
Discovery → qualification → website analysis → outreach → follow-up, as one pipeline.
The problem
Finding local businesses worth contacting is slow, and most of the time is spent on ones that were never a fit. The judgement is repeatable; the looking-up isn't worth a human's evening.
What it does
- Sources candidate businesses from public listings and search
- Enriches each with site, contact and technology signals
- Scores fit against explicit filters rather than vibes
- Drafts outreach grounded in what was actually found
- Tracks follow-up so nothing is chased twice or forgotten
How it’s built
Architecture — TypeScript · Supabase · LLM · Cost guardrails
Decisions & trade-offs
Cost guardrails before features
Every LLM call is metered and capped per run. An enrichment loop that silently costs £40 in a night is a worse bug than one that crashes — crashes are visible.
Qualification is deterministic, drafting is generative
Scoring uses explicit rules so a decision can be explained and re-run identically. The model only writes prose, where being wrong is cheap and obvious.
What it doesn’t do
- It does not send anything on its own — a human reviews every message.
- Scoring is tuned for UK local service businesses; it would need re-tuning for another market.
- It is not a CRM and does not try to be one.
Source is private. The architecture, decisions and limits above are the honest version of what it is.
Start here
Tell me what's eating your time.
No pitch deck, no “discovery call” funnel. Tell me the problem and I'll tell you honestly whether it's worth building — including when the answer is “don't bother”.
What happens next
- 01
You tell me the problem
A few sentences is plenty. No form maze, no discovery call funnel.
- 02
I reply within one working day
With an honest read on whether it's worth building — including when it isn't.
- 03
A fixed scope and a fixed price
Before any work starts, so you know exactly what you're getting and what it costs.