Automating Sales and CRM Work with AI
Five sales jobs that quietly decide whether a deal closes, and how much of each an agent can carry.
What ties these together
These share a failure mode: none of them is hard, and all of them get dropped when the week gets busy. The fifth follow-up, the enrichment nobody finished, the action item nobody assigned. An agent is useful here not because it is clever but because it is not busy — it does the fifth follow-up on the same day it did the first. The limit is equally consistent: the moment a human replies with something real, the agent's job is to get out of the way.
Each one below is written the same way, because these are the three things that decide whether an agent is worth building: what starts it, what it does, and the case where it has to stop and ask a person.
lead follow-up
The follow-up that closes the deal is usually the fifth one, and almost nobody sends it.
What starts it: A form is submitted, or a lead's status changes in the CRM.
What the agent does: watches for that trigger inside HubSpot, reads whatever arrived in whatever format it arrived in, and writes the result into Gmail, Calendly, Slack against rules you approved.
What it escalates instead of deciding: A reply that asks a real question, mentions pricing, or sounds annoyed. The agent's job is to keep the sequence going until a person is needed, then get out of the way.
quote generation
Quotes are assembled from known blocks with known pricing. That is a system, not a judgement call.
What starts it: A sales rep marks an opportunity ready to quote.
What the agent does: watches for that trigger inside HubSpot, reads whatever arrived in whatever format it arrived in, and writes the result into Google Docs, Stripe, Slack against rules you approved.
What it escalates instead of deciding: Any non-standard discount, a bundle with no list price, or terms outside the approval matrix. Quotes are commitments — the agent assembles, a person approves.
data enrichment
Enrichment done inconsistently is worse than none, because you cannot trust any of it.
What starts it: A record is created, or a scheduled sweep finds an empty field.
What the agent does: watches for that trigger inside HubSpot, reads whatever arrived in whatever format it arrived in, and writes the result into Airtable, Google Sheets, Slack against rules you approved.
What it escalates instead of deciding: A low-confidence match, or two plausible companies with similar names. Writing a wrong value into the CRM is worse than leaving the field blank, because nobody re-checks a populated field.
appointment scheduling
Scheduling back-and-forth costs more calendar time than the meeting.
What starts it: A booking request comes in, or a lead reaches the stage where a meeting is the next step.
What the agent does: watches for that trigger inside Calendly, reads whatever arrived in whatever format it arrived in, and writes the result into Google Calendar, Twilio, Slack against rules you approved.
What it escalates instead of deciding: A request to move something already confirmed with an external party, or a conflict with a meeting the agent did not create.
meeting notes
A summary nobody acts on is not worth the storage. The value is the assigned action.
What starts it: The call recording finishes processing.
What the agent does: watches for that trigger inside Fathom, reads whatever arrived in whatever format it arrived in, and writes the result into Slack, HubSpot, Notion against rules you approved.
What it escalates instead of deciding: An action item with no clear owner, or a commitment made to a customer. Assigning those wrongly is worse than not assigning them.
What has to be true before any of this is built
- Someone can state the rules out loud. If two people handle the same case differently and both believe they are right, that has to be settled first. Teams routinely find this is the most valuable part of the project, and are irritated that it came from an automation vendor.
- The escalations have somewhere to go. A named person, a queue, a channel — not an inbox nobody owns. An agent with no route for “I am not sure” makes a confident wrong decision instead, quietly, and by the time anyone notices the trust is gone.
- You know roughly what it costs today. Hours per week, written down before anything is built. Without a baseline nobody can say afterwards whether it worked — and the one randomised trial in this area found people believed they were 20% faster while actually running 19% slower.