Sales Booking AI Agent

This sales agent reads emails as they come in, determines if they're a "Sales" email, and either responds, books a meeting, or waits for Human in the Loop (HITL) feedback.

Agent Stack

  • Email – Gmail
  • Calendar – Google Calendar
  • HITL Inquiry – Slack
  • HITL Feedback – n8n Form
  • LLM – OpenAI GPT 4.1-mini
  • RAG/Vector – Supabase

Alternative Stack Options

  • Email – Microsoft Outlook
  • Calendar – Microsoft Outlook
  • HITL Inquiry – Microsoft Teams, Telegram, WhatsApp
  • HITL Feedback – Slack, Teams, Telegram, WhatsApp, Custom
  • LLM – Claude Models, OpenAI GPT models, OpenRouter, Ollama
  • RAG/Vector – PostgreSQL, MySQL, Microsoft SQL, Google Docs/Drive, OneNote

Flow Options

  • Remove the HITL approval process, making it a completely autonomous agent
  • Refine prompts throughout LLM chains and secondary agents (ie: Reformat Email, Calendar Agent, Revision Agent(s), Email Agent)
  • Remove Vector store

Additional Options

  • Incorporate CRM – Hubspot, Salesforce, GoHighLevel
    • Add new contacts
    • Add notes to contact based on the conversation flow
  • Schedule follow-up sequences
Sales Booking Agent Screenshot

How the Sales Booking Agent Works

The Sales Booking Agent checks the target inbox every minute, looking for new emails.

When a new email is received, it checks the "from" field to make sure it's not sent from the user's own email. If not, it proceeds to read the contents of the email and determine whether it's a prospect sales-oriented email. If it is, the AI agent sends the body content through a series of LLM chain agents to reformat the message into programmatically readable information, and determines whether the email sender is needing a response or is looking to book an appointment.

If the sender is requesting more information, the AI agent checks the RAG vector store to provide a salient response to the email sender. Once it has understood the RAG information, it drafts a response and sends it to Slack for Human In The Loop (HITL) feedback. The human can then approve the draft, provide an alternative recommendation, or take over the email response from there. If the draft is approved, the AI agent will email the sender directly. If a revision is asked for, the AI agent will refine the original draft and provide it back in the Slack channel for another HITL revision. If this version is approved, it will email the sender directly. If not, it'll go through the revision process again.

If the sender is requesting to book an appointment, the AI agent will use the Calendar AI Agent to check the availability of the associated calendar. It will then draft an email with the calendar availability based on the email sender's request, and pass this draft to Slack for HITL feedback. If approved, the AI agent will book the time on the calendar, adding the email sender as an attendee to the meeting, and then email the sender the response. If the sender needs to reschedule, they can directly reply to the email and the AI agent will start the process of finding new availability on the calendar.

Most of this is done without the HITL component. It's necessary at the beginning to make sure the AI agents are working properly and providing the correct information back to the senders. Once everything is working as expected, the HITL piece can be removed, and this can be a completely automated Sales Booking Agent.

Ready to save hours and scale smarter?

Let's discuss how custom AI agents can transform your business operations and free up your time.