Ali Fakhar
فارسیGet in touch

How to personalise cold emails at scale with AI

  • Level: Intermediate
  • A few days to set up

Ali Fakhar · Updated: Sept 2026

The short answer

Research each account first and keep only verified facts, such as a market launch or a job posting. Give the model those facts, your offer and one reason to talk, and ask for a short email with one question. A person edits every message before sending and removes invented familiarity.

On this page

At a glance

The marketing problem
A company name in a template is not a reason to start a conversation. Research needs to support a relevant message without invented familiarity or assumptions about a prospect’s priorities.
Who this is for
B2B marketing and sales teams doing considered outreach to potential customers.
Tools involved
  • HubSpot
  • Claude API
  • n8n

What you need

Before you start

  • A reviewed research brief for each account (see the prospect research workflow)
  • A lawful basis for outreach in each market, such as PECR and UK GDPR in the UK
  • Clear ownership of approval, sending and follow-up

What goes in

  • Verified company context and the reason for contacting them now
  • The recipient’s role and what they are likely responsible for
  • A relevant offer and one piece of approved proof
  • Your voice guide

How it works

  1. 01 · TriggerStart from a reviewed research brief
  2. 02 · AI stepDraft a short email
  3. 03 · You decideEdit and approve
  4. 04 · ToolSend through your sequence tool
  5. 05 · OutputLearn from the replies
  • Trigger
  • Tool
  • AI step
  • You decide
  • Output

Step by step

  1. 01 Trigger

    Start from a reviewed research brief

    Use the checked brief and choose one relevant reason to contact this account now. If there is no good reason, do not send.

  2. 02 AI step

    Draft a short email

    Give the model the verified reason, the offer and your rules, and ask for a short draft with one question.

    Prompt to copy

    Write a cold email of under 90 words to [name, role] at [company]. Reason for writing: [the verified fact, with its source]. What we offer: [one sentence]. Proof: [one approved result or customer]. Rules: one question at the end, no flattery, no claims about their priorities, no “I hope this finds you well”. Write two subject lines of under six words. Follow this voice guide: [paste].

  3. 03 You decide

    Edit and approve

    Read every message for accuracy, tone and relevance. Delete anything that implies a familiarity you do not have.

  4. 04 Tool

    Send through your sequence tool

    Send through your approved tool with sensible daily limits, and honour opt-outs across every sequence.

  5. 05 Output

    Learn from the replies

    Review replies, including negative ones, to improve both the research and the message.

Where a person decides

A person reads and edits every message for accuracy, tone and relevance before it is sent.

What to watch for

  • Personalisation that is wrong, or too personal, does more harm than a plain email.
  • Do not claim to know priorities the recipient has not shared.
  • Deliverability depends on list quality and sending practice, not only on copy.

How to tell it is working

  • Every personalised line traces to a verified source.
  • Positive replies and meetings are measured, not open rates.
  • Negative replies and unsubscribes are tracked as a quality signal.

An illustrative example

Input

Illustrative: a fictional logistics company has announced a new warehouse in the Netherlands on its news page.

Output

A draft of under 90 words that mentions the announcement, links it to one relevant offer and asks one specific, low-pressure question, with two short subject lines.

Want this running in your team?

AI workflow implementation. Turn a repetitive marketing task into a workflow your team can use with confidence.

View service
All AI in marketing workflows ↗
Analytics preferences

Current preference: off