LinkedIn outreach

Turn company website interest into LinkedIn conversations with Warm AI and FirstTouch

Published: 19 October 2026

Warm AI shows the companies taking an interest in your website. FirstTouch helps you find relevant people at those companies and prepare a LinkedIn conversation.

The short answer

Warm AI shows the companies taking an interest in your website. FirstTouch helps you find relevant people at those companies and prepare a LinkedIn conversation.

There are two ways to run this play: use an AI agent with Slack, or manage the LinkedIn steps in HubSpot workflows. Both start with company qualification and contact discovery. Both use one connection note and one bump after acceptance if there has been no reply.

Start with human approval required so you can review the selected person and message before sending.

Start with the company showing interest

A company reading your product pages or returning to your site gives your team useful context. Warm AI makes that company-level interest visible, including the available pages viewed and visit history.

Use the company domain to take the next step. Your ideal customer profile (ICP) defines which companies fit. Your personas are the job roles you want to reach. Warm AI already supplies company details and can provide named US visitor data where available. Reuse that information first. FirstTouch can fill missing qualification details or find a different target-role contact when needed.

Contact discovery does not establish which individual visited. The outreach should refer to interest from the company rather than claiming that a specific person was on your website.

Path 1: Use an AI agent

Choose this route when you want an agent to qualify the company and prepare the outreach.

  1. In Warm AI, open Integrations, connect Slack, and send company alerts to a dedicated channel with the company domain and available visit context.
  2. Connect your agent to Slack and FirstTouch's Model Context Protocol (MCP) tools. Confirm that the agent can monitor new channel alerts and that the monitoring task is actually configured.
  3. Give the agent your ICP, target roles, product positioning, LinkedIn sender, follow-up delay, and contact and credit limits.
  4. Have it check fit using available Warm AI data and select one relevant contact. Use FirstTouch only for missing information or to discover a target-role contact when none suitable was supplied.
  5. Prepare FirstTouch Dynamic Actions for that contact: a connection request with a note, then one bump on the connection-accepted branch. Keep human approval required on both.
  6. Review one real alert, selected contact, and action sequence before enabling ongoing processing.

MCP gives an agent access to FirstTouch's tools. FirstTouch Dynamic Actions are steps for an individually selected contact. The agent checks exclusions, existing connections, and overlapping outreach before preparing them.

FirstTouch's step-by-step joint guide includes the complete agent setup prompt, including the approval setting, messaging, and first test.

Path 2: Use HubSpot workflows

Choose this route when your team wants the LinkedIn steps in HubSpot's workflow builder.

  1. In Warm AI, open Integrations and connect HubSpot to sync the identified company and available company data.
  2. Compare the available Warm AI company data with your ICP. Use FirstTouch's MCP or API only for missing qualification details.
  3. Select one relevant contact using suitable Warm AI data where available. If you need a different persona or no contact was supplied, use FirstTouch's MCP or API for discovery. Have your agent or configured process create or update the qualified HubSpot contact and associate it with the company.
  4. Add FirstTouch's connection-request action in the HubSpot workflow builder. Choose the sender, add the note, skip existing connections, and keep approval enabled.
  5. Configure the acceptance check, follow-up delay, and reply-based suppression before the single FirstTouch message action. Keep approval enabled on that action too.
  6. Test with one qualified contact before enrolling more people. Prevent duplicate enrollment, stop on any reply, and end after the bump.

Reuse Warm AI company and contact data first. Any additional FirstTouch enrichment or contact discovery uses its MCP or API. The HubSpot integration runs the LinkedIn steps after qualified contact records exist. Record creation needs HubSpot write access, and this route needs the relevant HubSpot workflow features.

The two messages

A connection note should name your product, explain one outcome, and ask a clear question.

Replace the placeholders with your confirmed positioning and keep the note within the sender's supported limit.

That is the entire sequence. No acceptance means no bump. Any reply stops the sequence. After the bump, stop even if there is no response.

Review the first conversations

Check whether the companies fit your ICP and whether the selected people are relevant to your product. Review the note before sending, then use acceptance and reply rates from your own activity to refine the play.

A prepared action is not a sent message. Keep human approval required while you review the initial outreach, and verify acceptance and reply conditions before allowing the bump.

Supported platforms

FirstTouch supports clients including ChatGPT, Claude, Grok, Gemini, Codex, and Cursor through its supported connections. Automatic Slack listening depends on the agent's own monitoring features, plan, and permissions. HubSpot workflows are the alternative route for the LinkedIn steps.

Make the next company visit useful

Warm AI gives you the company-interest signal. FirstTouch helps turn that signal into a relevant person and a clear next step.

Read the full Warm AI and FirstTouch guide for the two setup paths, workflow visual, and copyable agent prompt. Or explore Warm AI to start with the companies visiting your site.