Ben Audry Get in touch
All work
GTM automation, Kuration AI

Automated LinkedIn Connection Messaging System

Turn warm connections into conversations. The system filters my LinkedIn network for founders and CEOs, validates the ICP fit, then runs a daily workflow that researches each contact's company and sends a personalised 5-message DM sequence, spaced 10 to 30 seconds apart so it reads like a human, not a bot.

Connection export from LinkedIn
ICP filter founders + CEOs
GTM research OpenAI per company
Craft 5 DMs personalised sequence
Sequence send 10 to 30s apart, daily

How it works

Pull connections

Export the full LinkedIn connection list to a Google Sheet. Raw data, no sorting, just every name, title, company, and profile URL in one place.

Filter founders + CEOs

Keep only titles that match the ICP, founder, co-founder, CEO. Everyone else stays out of the funnel.

Manual curation pass

Filtered contacts land on a staging tab for a human sanity check against the ICP definition. A sheet formula derives the LinkedIn slug from each profile URL. Confirmed contacts move to the live list, ready for outreach.

Daily trigger

An n8n schedule fires every day, reads the live sheet, filters rows where status is not Sent, and pulls each contact's LinkedIn profile through Unipile.

GTM research per contact

An OpenAI agent runs a go-to-market briefing on the contact's company website. The output, positioning, ICP, key pains, feeds the next step.

Craft + send 5 messages

A second OpenAI agent breaks the GTM research into 5 tailored messages. Unipile sends them sequentially with a 10 to 30 second wait between each, so the cadence looks human and avoids LinkedIn's spam flags. Sheet row updated, Slack pings on completion.

n8n workflow: schedule trigger reads contacts from a Google Sheet, filters not-sent rows, fetches LinkedIn info via Unipile, runs GTM research and craft message agents, then sends 5 DMs sequentially with wait nodes between each, updating the sheet at the end
Daily DM workflow: schedule trigger to 5-message sequence per contact.

Why this works

Cold DMs to strangers get ignored. Connections are warmer ground, but a generic template still kills reply rates. Two design choices move the needle here. First, the ICP filter plus manual curation means the workflow only touches contacts who actually fit, no spray and pray. Second, splitting one pitch into 5 short messages, each anchored in real research about the contact's company, reads like a human picking up a thread over a few days. The 10 to 30 second jitter between sends is small but matters, it keeps the pattern off LinkedIn's bot radar.