AI-Personalised Email Sequences

Why Personalisation Matters (and Why It's Hard at Scale)

Generic email sequences have falling open rates and even lower reply rates. Personalised emails — referencing the recipient's company, their specific role, their industry challenges — perform significantly better. The problem is that writing personalised emails manually does not scale. AI makes it possible to generate genuinely personalised emails at scale, with n8n handling the delivery and sequencing automatically.

The Workflow Architecture

1. Trigger — a new row added to an Airtable base or Google Sheet (via polling or webhook), or a new contact in HubSpot. This represents a new lead entering the sequence.
2. Airtable / Sheets node — fetch the lead's data: name, company, job title, industry, and any recent context (recent funding, product launch, or tech stack if enriched).
3. Basic LLM Chain — personalise the email body. Pass all lead data as context in the prompt:

   "Write a friendly, professional outreach email to {{ $json.first_name }} who is a {{ $json.job_title }} at {{ $json.company }} in the {{ $json.industry }} industry. Reference one specific challenge common to {{ $json.industry }} companies at their stage. Keep the email under 120 words. Do not be salesy — focus on being genuinely helpful."

4. Gmail or SendGrid node — send the email. Use a personalised subject line (also generated by the LLM Chain or templated with the lead's name and company).
5. Wait node — pause the workflow for 3 days before the follow-up.
6. Basic LLM Chain (second call) — generate a follow-up message that references the first email without repeating it.
7. Gmail or SendGrid node — send the follow-up.

The Wait Node for Sequence Timing

The Wait node pauses a workflow execution and resumes it after a specified amount of time. Use it between email steps to space out messages naturally. Set the first Wait to 3 days, the second to 5 days. Each execution is paused independently, so multiple leads can be in the sequence simultaneously without interfering with each other.

Generating Subject Lines That Get Opened

Add a separate LLM Chain call before the send step to generate the subject line. Pass the email body as context and ask for three subject line options — then use the first one, or add a Set node to pick the shortest. Avoid generic subject lines: the LLM should reference something specific to the recipient's company or role.

Stopping the Sequence on Reply

Add a check before each send step: query the Gmail API to see if the lead has replied to the previous email. If a reply is found, skip the current step and mark the lead in your CRM or spreadsheet. Continuing to send sequence emails after someone has replied is damaging — add this check from the start rather than retrofitting it later.

Review Before Scaling

Run the sequence for 10 leads manually — review every generated email before it is sent. After confirming quality, switch to automatic sending. Keep a sampling review in place (check 10% of emails weekly) so quality degradation is caught early.