SDRaccount-research · beginner
30-second account research with Claude + Clay
Last reviewed: 2026-05-23 · saves ~20 min/prospect/run
Our take
The single highest-ROI workflow for any SDR. Replaces 20+ minutes of manual research with a 30-second Claude call that pulls LinkedIn + company news + earnings + 10-K snippets + funding events. Quality depends on prompt rigor — generic prompts give generic output. The version we ship below has been tuned across ~2000 accounts.
Tool stack
Steps
- Wire Clay to your prospect list. Each row pulls LinkedIn, company website, news, and funding data automatically.
- Add a Claygent column with the prompt below — it'll run for each row in batch.
- Output goes to a Slack channel or directly into Apollo's notes field.
- Review top 5 accounts manually before outreach — AI gets ~80% right, you fix the 20% that matter.
- After 30 days, audit reply rates by AI-research-quality bucket; iterate prompt.
Prompts
Research one prospect for outbound (Claygent) · Claude Sonnet 4.6
You are an SDR research analyst. Given a prospect's LinkedIn URL and company URL, return a 4-line research brief: LINE 1 (Why now): one recent event or signal that makes this account in-market this quarter. Source must be linkable. LINE 2 (Pain hypothesis): one specific job-to-be-done they likely have, based on their role + company stage. LINE 3 (Hook): one personal detail from their LinkedIn (post, comment, conference talk) we can reference naturally. LINE 4 (Avoid): one thing not to bring up (controversy, layoff context, etc.) Constraints: - Every line under 25 words. - If you don't have data for a line, write "UNCLEAR — verify manually." - Never invent quotes or numbers. - Use the prospect's name in line 3 to confirm it's them.
Pitfalls
- Generic prompt = generic output. Force specificity in the system prompt.
- Hallucinated quotes from public sources. Force source URLs.
- AI tone too formal for first-touch. Tell it to write in the prospect's voice based on their LinkedIn style.