Personalization at scale

Personalization used to mean "Hi [First Name]." Modern cold email uses enriched data, signal scraping, and AI to generate custom first lines at scale. Done right, every email feels hand-written even at 1000 sends per week.

The three tiers

Tier 1: Token replacement

Basic merge fields: first name, company, role, industry. Bare minimum. Doesn't actually read as personal.

Tier 2: Structured enrichment

Pull LinkedIn posts, company news, funding events, hiring signals, tech stack changes. Build a "trigger" field per prospect. Reference in first line.

Tier 3: AI-generated custom lines

For each prospect, feed structured data (LinkedIn activity, news, role) to an LLM. Get a unique first line per prospect. Routine with Clay + OpenAI/Claude integration.

The Clay workflow

The dominant stack in 2026:

  1. Export prospects from Sales Navigator, Apollo, or ZoomInfo to Clay
  2. Enrich each row with: LinkedIn activity, company news, hiring signals, tech stack
  3. Add AI column with prompt: "Write a 15-word first line referencing [most interesting trigger]"
  4. Push to cold email tool (Instantly, Smartlead) with personalized first line merged in

Result: every prospect gets a unique, specific first line. The rest of the email is a template with merge fields.

The AI first-line prompt

Clay AI column prompt
Given the following data about a prospect, write ONE first line of a cold email, 12-18 words, that references something specific and doesn't sound templated.

PROSPECT DATA:
- Name: {{name}}
- Role: {{role}} at {{company}}
- Recent LinkedIn post: {{latest_post}}
- Company news: {{company_news}}
- Hiring signal: {{job_posting}}

Rules:
- Pick the MOST specific trigger (not just the most recent)
- Sound like a peer, not a marketer
- No "Hope you're well" / "I came across"
- No generic congrats
- Reference the specific trigger by name or detail

Output: just the first line, nothing else.

The templated-but-personal email body

First line is per-prospect. Body is a template with 2-3 variable fields (role, company, industry pain).

Templated body with merge fields
Subject: {{subject}}

{{first_line}}

We work with {{role_plural}} at {{company_size}} B2B teams on the specific problem of {{role_specific_pain}}.

{{proof_point_for_segment}}

Worth 15 min next week? Tue 2pm or Thu 10am ET?

Sam

Four merge fields: subject, first_line, role-specific pitch, proof point. Everything else is a template. Each email reads as if written for them.

The quality tiers of enrichment

Fast, cheap

Medium

High-quality, slower

For enterprise targets ($50K+ ACV), tier 3 is worth it. For SMB volume plays, tier 2 is the right balance. Tier 1 doesn't work at all in 2026.

The quality check

Before scaling a personalization workflow to 1000+ prospects, spot-check 50 generated first lines manually. If more than 3 feel generic or wrong, fix the prompt. AI-generated spam at scale destroys more than unpersonalized sends.

The honest limit

No matter how good the AI is, real human review beats automation for top prospects. For your 20 most important targets per month, write the first line yourself. Save AI personalization for tiers below.

Related: cold outreach in direct response.