TL;DR
Use Clay to research each prospect with AI agents (Claygent) that pull company news, LinkedIn activity, and job postings, then write personalized first lines using an LLM. This produces genuinely researched personalization for 500 prospects in the time it used to take to personalize 20.
True personalization at scale was impossible until 2023. Now it's standard practice for outbound teams using Clay.
Why basic tokens aren't personalization: [First Name] and [Company] aren't personalization — they're mail merge. Everyone knows it's a template. Real personalization shows you actually paid attention.
The Clay-based personalization workflow:
Step 1: Build your base list with company and contact data from Apollo.
Step 2: Add a Claygent enrichment column. Clay's AI agent can browse the web, read LinkedIn profiles, and pull company news for each prospect. Typical research tasks: - "Find this company's most recent press coverage or funding announcement" - "Find the most recent LinkedIn post by this person and summarize it in 1 sentence" - "Check if this company is actively hiring SDRs or sales roles" - "Summarize this company's product in one sentence based on their website"
Step 3: Generate personalized first lines. Add an AI column (Claude or GPT-4) that takes the enrichment data and generates a custom first line for each prospect. Prompt example: "Write a 1-sentence cold email opener for an SDR tool company reaching out to [Name], VP of Sales at [Company]. Use this context: [Claygent output]. The opener should reference their specific situation, not compliment them or comment on their job. Sound natural, not generated."
Step 4: Human review. Export to a spreadsheet and have a human review the AI-generated lines before they go out. Catch hallucinations, awkward phrasing, or lines that miss the mark. Takes 1–2 minutes per 100 records with a skimming approach.
Step 5: Import with the personalized field into your sequence. The first line is a custom merge tag; the rest of the email is templated.
This workflow produces 6–12% reply rates vs. 1–3% for generic templates. We've seen it 5x reply rates within 30 days of implementation for clients.
From Cactus: Clay personalization is standard in every Cactus outbound program — our in-house Clay expertise is one of our key differentiators for clients who've tried generic outbound and seen declining results.
Cactus Marketing embeds with B2B tech startups to turn strategy into pipeline. We've worked with 60+ companies, supported 12 exits, and contributed to $7B+ in client valuations.
Book a free 30-minute call — we'll give you a concrete plan for your situation.
Book a free strategy call →How do I write a cold email that gets replies?
Write one sentence that's specifically about them, one sentence on their problem, one on your solution, and one CTA. The email should be under 80 words, reference something real about their company, and ask a yes-or-no question at the end.
How do I find leads for cold email?
Start with Apollo.io or LinkedIn Sales Navigator to build a list from your ICP criteria. Enrich it through Clay or a waterfall of data providers for verified emails. Target trigger events — recent funding, new hires, job postings — to catch companies in active buying moments.
What is a good cold email reply rate?
A good cold email reply rate is 3–8% for a broad ICP campaign and 8–15% for a highly personalized, trigger-event-based campaign. Anything above 15% with meaningful volume is excellent. Below 2% means something fundamental is wrong — ICP, targeting, or the email itself.
How many cold emails should I send per day?
New domains should warm up for 2–4 weeks before sending real campaigns, starting at 10–20 emails/day and capping out at 30–50/day per domain for sustained campaigns. With multiple warmed domains running in rotation, total volume can reach 500–2,000+/day without domain damage.