From Checking HubSpot to Acting in Slack

The Situation

A B2B company running sales through HubSpot. Good inbound, decent pipeline, and one quiet problem: nothing happened until a human went and looked.

A new form submission, a deal moving stage, a contact hitting a lead-score threshold. All of it sat in HubSpot until someone opened HubSpot. On a busy day that could be tomorrow. On a Thursday afternoon it could be Monday.

The team was not slow. The system just had no way to tap anyone on the shoulder.

What I Built

A HubSpot-to-Slack layer with AI in the middle:

• Every relevant HubSpot event streams to Slack. New leads, stage changes, high-intent activity.
• Claude reads each one before a human does. It pulls the company, the role, what they asked for, and any history in the CRM, and writes a two-line summary.
• Routing by rules the team wrote themselves. Enterprise-looking lead goes to one rep, partner enquiry to another, support-shaped request goes to support, all with the right person tagged.
• Drafted replies for the ones that matter. When a lead matches an agreed profile, the notification arrives with a ready-to-send email draft in the rep's voice. They read it, edit if needed, send. Two minutes instead of twenty.

The Results

• Time-to-first-touch dropped from "whenever someone checks" to minutes. The team stopped losing leads to the calendar.
• Nobody checks HubSpot anymore. HubSpot checks them.
• Reps reply faster and more consistently, because the first draft is already there and already correct about who the person is.

Why It Worked

The team had already agreed on what a good lead looked like. They had just never written it down anywhere a system could read it. Most of the project was turning tribal knowledge into rules. The AI part was the easy bit.