Clay workflow · Personal → seed-stage client, 2026
Fundraising as an outbound problem
A Clay workflow that turns one startup profile into a ranked investor pipeline with outreach already written: research, thesis fit, partner selection, enrichment and personalisation in a single automated run.
What came out of it
- 1 input
- A startup profile drives the entire pipeline
- 8 steps
- Positioning → firms → partners → list → email → drafts
- 2×
- Built for myself, then replicated for a seed-stage client
Context
Founders raise the way bad SDRs prospect: a long list, the same deck, and hope. The expensive part is not the sending, it is the judgement before it. Which firms invest at this stage, which partner owns the thesis, whether they already backed a competitor. Skipping that is what makes outreach ignorable.
What I built
Read our positioning
One input. The startup's own profile drives everything downstream.
Research investor firms
Filtered to a matching thesis. A partner who already backed an adjacent company is a conflict, not a lead.
Find partners at firm
Funds do not write cheques, partners do. Runs per firm to resolve who owns the thesis.
Build investor list
A Python handler I wrote inside Clay: normalises firm names and walks nested people objects, because the sources disagree and dedupe fails without it.
Find investor work email
The pipeline ends with a reachable person rather than a firm name
Draft email and LinkedIn
Written from both sides of the match: the startup's specifics and the partner's own portfolio
What I'd do differently
Two things earn their place. The portfolio check, because sourcing and enrichment are commodities, every tool does them, and deciding whether a partner has already backed an adjacent company is the judgement a founder would apply if they had the hours. And the Python step, because the moment real data arrives the tidy no-code path breaks: three sources spell the same firm three ways, and no amount of prompt-writing fixes a dedupe. Automating the judgement is the point; being willing to drop into code is what makes it survive contact with the data.