How to Use AI to Personalise Cold Outreach at Scale — Without Sounding Like a Robot

Personalisation is the most important variable in cold email reply rates. An email that references something specific about the recipient — their recent hiring announcement, a LinkedIn post they published, a product they just launched — converts dramatically better than one that could have been sent to anyone.

The challenge has always been scalability. Writing a genuinely personalised opening line for every prospect in a 500-contact list takes hours. Agencies and outreach teams have historically had to choose between quality (highly personalised, low volume) and quantity (high volume, generic).

AI eliminates that trade-off.

With the right workflow, you can generate genuinely relevant, personalised opening lines for hundreds of prospects in a fraction of the time it would take manually. This guide covers the specific tools, prompts, and processes that make it work — and the failure modes that make AI personalisation feel more robotic than manual outreach.

Why Most "AI-Personalised" Outreach Fails

Before covering what works, it is worth understanding why most attempts at AI personalisation produce disappointing results.

Failure mode 1: No input data. AI personalisation requires data about the prospect to personalise against. Feeding an AI model only a name and company produces generic variations of the same sentence. "Hi [Name], I see you're at [Company] — impressive work!" is not personalisation. It is a template with a merge field.

Failure mode 2: Over-reliance on tone modifiers. Telling AI to "write a warm, personalised opening" without giving it specific facts about the person produces plausible-sounding but hollow text. "I was impressed by your recent work in the [industry] space" — every recipient can tell this was not written specifically for them.

Failure mode 3: Not editing the output. AI output for cold email should be a starting point, not a final draft. A paragraph of AI-generated prose dropped directly into a cold email reads like AI-generated prose. The skill is in prompting precisely and then editing tightly.

Failure mode 4: Personalising the wrong part. Personalising the closing ("looking forward to connecting, [Name]!") is not personalisation. Personalisation has to be in the opening line — the first thing the recipient reads — and it has to be specific enough that they would know it was written about them, not about the last 50 people who received the same email.

The AI Personalisation Workflow

Here is the end-to-end workflow for generating personalised cold email opening lines at scale using AI.

Step 1: Collect Personalisation Data

For AI personalisation to work, you need data. The richer the data for each prospect, the more specific and relevant the personalisation can be.

Data sources to collect for each prospect:

• LinkedIn profile URL (for AI tools that can browse LinkedIn)
• Recent LinkedIn posts (last 30 days)
• Company website — specifically the "About" page, recent blog posts, case studies
• Any recent news mentions (press releases, industry news, funding announcements)
• Google Business Profile (for local business prospects) — review count, recent activity
• Any relevant signals from your list-building research (recent hire, recent expansion, etc.)

How to collect this at scale:

For most tools, this means enriching your list with additional data columns. Options:

• Clay (clay.com) — The most powerful enrichment platform for cold outreach. Clay integrates with LinkedIn, Apollo, Clearbit, and dozens of other data sources, and has a built-in AI column feature that can generate personalisation copy from the enriched data. Recommended for anyone doing AI-personalised outreach at volume.

• Phantombuster — Can scrape LinkedIn post content and recent activity for a list of prospects. Export to CSV, then process through AI.

• Manual research + a spreadsheet — For smaller campaigns (under 100 prospects), manually visiting each prospect's LinkedIn and noting one specific detail per person, then using AI to generate the opening line from that detail, is the most reliable approach.

Step 2: Generate Personalised Opening Lines With AI

With data collected, generate personalisation in one of two ways:

Option A: Clay's AI column (fastest at scale)

In Clay, create a new column with an AI formula. A prompt like:

"Based on the following information about [First Name], who works at [Company]: [LinkedIn post summary / recent news / company update] — write a 1–2 sentence personalised opening for a cold email. The opening should reference something specific and genuine about their situation. Do not be sycophantic. Do not use phrases like 'I was impressed.' Keep it under 25 words."

Clay runs this prompt for every row in your spreadsheet, generating a unique personalised opening for each prospect. Review and edit the outputs before importing to your sending tool.

Option B: GPT batch processing via a spreadsheet

If you don't use Clay, you can achieve a similar result using a Google Sheet with the ChatGPT or Claude API:

1. Build your prospect spreadsheet with columns: First Name, Company, LinkedIn insight (one specific fact per prospect)
2. Add a formula column that constructs the AI prompt: ="Write a 1-sentence personalised cold email opening for " & A2 & " at " & B2 & " who recently " & C2 & ". Keep it under 20 words. Do not use the phrase 'I was impressed.'"
3. Send the combined prompt column to GPT or Claude via the API (or manually, if volume is low)
4. Output column: the personalised opening line

Option C: Manual AI generation for high-value campaigns

For campaigns targeting a small number of high-value prospects (e.g., 50 ideal clients you want badly), manually research each prospect and generate a custom opening line:

1. Visit their LinkedIn; note one specific, recent, genuine detail
2. Open ChatGPT and prompt: "Here is a specific fact about a B2B prospect: [fact]. Write a cold email opening line that references this specifically and naturally. Keep it under 20 words. Do not use 'I was impressed' or any sycophantic language."
3. Review, edit lightly, and paste into the email

Step 3: Review and Edit AI Output

Do not send AI output without reviewing it. Common issues to catch:

• Too formal or generic: "I was fascinated to read about your company's recent expansion" — rewrite as something more conversational
• Factually incorrect: AI hallucinates. If it generates an opening referencing something that didn't happen, it will damage your credibility instantly
• Too long: AI tends to generate more words than cold email needs. Cut aggressively
• Clichéd phrases: "Impressive work," "incredible journey," "passion for innovation" — these are AI tells. Remove them
• Wrong tone for your ICP: formal language for a startup founder feels off; casual language for a law firm partner feels off. Review with your ICP's communication style in mind

A 10-minute review of AI-generated opening lines for a 100-person list will catch 15–20 outputs that need editing before they are ready to send.

AI-Enhanced Subject Line Generation

Beyond opening lines, AI can generate and test subject lines faster than manual brainstorming.

Prompt template:

"Write 5 short cold email subject lines for an email to [job title] at a [company type]. The email is about [brief description of what the email does or offers]. Subject lines should be under 7 words, feel like a personal email, and create curiosity without being vague. Do not use exclamation marks or phrases like 'exciting opportunity.'"

Generate 5–10 variants, pick 2–3 to A/B test in your first batch, and let the data determine the winner.

AI-Generated Follow-Up Variations

A significant advantage of AI in cold outreach is generating follow-up email variations that feel natural rather than mechanical. Most follow-up sequences use the same email body for every recipient — which is fine for early-stage campaigns but leaves conversion on the table.

AI can generate personalised variations of follow-up emails by pulling in the same context data used for the opening line:

"Write a brief 2-sentence follow-up email for a cold outreach sequence. The prospect is [First Name] at [Company]. In the first email, I referenced [what you referenced in email 1]. This follow-up should add a new angle or piece of value — specifically related to [their industry/situation]. Keep it under 50 words."

Full AI-Personalised Email: Worked Example

Prospect data:

• Name: Sarah
• Company: Maple & Stone Interior Design (8 employees, Manchester)
• LinkedIn signal: Posted two weeks ago about winning a commercial fit-out contract for a new Manchester hotel

AI prompt:

"Write a cold email opening line for Sarah, who runs an 8-person interior design studio in Manchester. She recently won a commercial hotel fit-out contract and posted about it on LinkedIn. The opening should reference this specifically. Keep it under 20 words. Do not use 'I was impressed.'"

AI output (edited):

"Congrats on landing the hotel fit-out — that kind of project is a different operational beast than residential work."

Full email (using this opening):

Hi Sarah,

>

Congrats on landing the hotel fit-out — that kind of project is a different operational beast than residential work.

>

I work with design studios scaling into commercial projects — mainly helping them handle the increased client communication and quoting workload without hiring extra admin staff.

>

Worth a 15-minute chat to see if that's relevant for you?

>

[Name]

This email is 75 words, highly relevant, and clearly written for Sarah — not for a generic "interior designer." It would take about 3 minutes to produce with this workflow.

For Businesses at Different Stages

If you're managing high-volume outreach: AI personalisation is the key to scaling quality. Without it, you either sacrifice personalisation for volume (lower reply rates) or sacrifice volume for quality (lower output). Clay or similar tools let you run 500+ personalised outreach emails per week with a team of 1–2.

If you're starting with a small pipeline: AI personalisation lets you punch above your weight. A 20-person startup running AI-personalised outreach to 200 high-fit prospects will generate more qualified conversations than a large team sending generic emails to a 2,000-person list. Relevance beats volume every time.

The Quality Control Rule

AI personalisation should be used to accelerate production — not to replace judgment. The final check before any email goes out should be a human asking: "If I received this email and knew the sender had generated it using AI, would I feel like the personalisation was genuine and relevant — or would I feel like it was a trick?"

If the answer is the latter, the output needs editing.

The goal is not "personalised" by the standard of a machine. The goal is genuinely relevant by the standard of a human recipient.

Systemify Automation builds AI-powered cold outreach systems — including enrichment workflows, Clay setups, and AI copy generation pipelines — for agencies and B2B businesses. If you want a system that produces personalised outreach at scale, get in touch.