Automated lead qualification with AI agents (HubSpot and Salesforce)

Inbound leads arrive faster than anyone can check them. This guide shows how to set up an Orkeste agent, or a swarm of two, that reads inbound emails, checks HubSpot or Salesforce, applies your qualification criteria, records the lead and drafts a reply. A rep approves every email and every CRM change you choose to keep under review.

The problem

Most teams qualify leads with a form score and a review once a day. Promising leads wait, poor fits take a rep's time, and the CRM fills with half-complete records. Rule-based automations score on form fields but cannot read what the prospect actually wrote.

How the workflow runs

Runs: Every hour, or when you start it.

  1. Read inbound emails: The agent reads recent emails in the sales inbox, in Gmail or Outlook.
  2. Check the CRM: In HubSpot it finds the contact by email and its deals; in Salesforce it searches the contact, the account and open opportunities.
  3. Qualify: It applies the criteria from its briefing and writes its verdict in one sentence.
  4. Approve: CRM changes and the reply draft wait in Validations for a rep.
  5. Record the lead: In HubSpot it creates the contact and a deal at the first stage, or adds a note. In Salesforce it logs an activity on the existing record.
  6. Hand off: Qualified leads are posted to the sales channel in Slack. Approved replies are sent from Outlook or saved as Gmail drafts.

The agents propose; a rep approves the CRM changes and every email before it goes out.

Who does what

What this workflow does not do

Works with your stack

Recipes in this guide

Questions about this workflow

How is this different from a Zapier or Make flow?

Those run fixed rules on structured fields. Here the agent reads the message the prospect wrote, checks your CRM and explains its verdict in a sentence, and the steps you choose wait for a rep.

Do I need HubSpot or Salesforce?

Either. With HubSpot the agent creates the contact and the deal itself; with Salesforce it logs activities on existing records and flags new companies for a rep to create.

How do I write the criteria?

In plain language in the briefing: company size, the needs you serve, the signals that matter such as a budget or a deadline, and what to do with poor fits.

What does it cost to run?

Each run uses credits from your plan, from 2 credits with Claude Haiku 5.5 to 235 with Claude Fable 5.1 depending on the model you choose, and you can set a daily and a monthly budget per agent.

Related workflows

Keep reading