Generic cold email is dead. Buyers have been trained to ignore templated blasts. But AI-personalized outbound — where every email reads like it was written specifically for that person — books meetings at 3–5x the rate. Here's the exact system we use.
Before you touch any tool, you need a precise ICP definition. Not "VP of Marketing at mid-market SaaS" — that's a job title, not an ICP. A real ICP includes the signals that indicate a company is actively ready to buy: recent funding, headcount growth above 20%, a specific tech stack, a recent leadership hire, or an expansion into a new market.
We build our ICP definition around three layers: firmographic fit (company size, industry, geography), technographic fit (what tools they use and what that tells us about their sophistication), and trigger events (the specific things that happened recently that make them a warm prospect right now).
Once your ICP is tight, pull an initial list from Apollo.io using company-level filters. Aim for 500–1,000 companies before contact-level filtering. Quality here determines everything downstream.
Pull your Apollo list into Clay. Clay's power is in its enrichment waterfall — it can pull data from 50+ sources simultaneously and write it into structured columns you can use in prompts.
The enrichment data we always pull:
This data feeds your AI first-line generator. The more specific the data, the more specific the personalization.
The first line of a cold email determines whether the rest gets read. We use Clay's AI column with a carefully engineered prompt to generate a unique, specific first line for every contact.
Our prompt structure:
"Write a one-sentence cold email opener for [name] at [company]. Reference one specific, recent, and verifiable thing about them or their company from this data: [recent post / company news / job posting]. The opener should feel genuine and specific, not flattering. Don't mention their name again after 'Hi [name]'. Max 25 words."
The key constraint: specificity. AI that generates generic openers ("I love the work you're doing at [Company]!") is worse than no personalization at all. Your prompt must force the model to reference something real.
We run a 5-step sequence over 14 days. The structure:
Great copy means nothing if your emails go to spam. Deliverability setup that we never skip:
Week 1: ICP definition, list building, Clay enrichment, AI first-line generation, copy writing, and deliverability setup. Week 2: Technical setup complete, sequences launched to the first 200 contacts with warmup complete. Week 3: First replies coming in, positive replies booked directly to calendar via Calendly link in email.
The benchmarks we target: 40%+ open rate, 3–8% reply rate, 1–3% meeting booking rate. If you're below these, the problem is usually the ICP definition (targeting wrong people) or the sequence timing (sending too fast, feeling spammy).
Book a 20-minute fit call. We'll show you exactly how many meetings we can book for your business.
Book a fit call →