Key takeaways
- Before calls, an agent reads the account and writes a short brief for the seller.
- During the day it finds contacts and drafts first messages, replies and follow-ups.
- After meetings it summarises threads and suggests fixes to records.
- Anything that reaches a customer or changes shared data should have a clear approval rule.
- Start with one bounded workflow and one owner, then widen the scope.
What does an AI agent do before the first call of the day
Before the first call, an AI agent prepares the seller: it reads each account on the calendar and writes a short brief covering what the company does, recent news, who is attending and what was said last time. The seller reads the brief in minutes instead of opening ten tabs.
Give it a goal such as "prepare me for today's four meetings" and it breaks that into research, search and summary steps. A wrong fact here costs little, because nothing has been sent.
How does an agent help with prospecting and outreach
During the day, an agent does the reading and writing around outreach that needs judgement but not a relationship. It finds companies that match your ideal customer, picks the right people at them and drafts an opening message for each, in your voice.
- Finding companies that match your ideal customer, then the right people at them.
- Drafting first messages, replies and follow-ups for the seller to check.
- Summarising long email threads and suggesting the next step.
- Flagging replies that need a quick human answer.
What does an agent do after calls and meetings
After a meeting, an agent turns notes into the follow-up email, the next task and suggested updates to the record. The seller checks the draft and the changes rather than writing them from scratch.
It can also tidy records the team rarely gets to: merging duplicates, filling missing fields and tagging segments. These changes touch shared data, so they belong behind an approval step at first.
Where do AI agents fit in revenue operations
In revenue operations, agents fit the repetitive judgement work that keeps the system clean: checking that new leads match the ideal customer, routing them, filling missing fields and preparing the weekly pipeline review.
Take an illustrative software company whose CRM fills with half-complete records after every event. An agent can read each company's website, add industry and size, flag records that do not fit, and leave a short note on why. A RevOps lead then reviews the flagged records instead of every one.
What should an AI agent not do on its own
An AI agent should not take actions that are hard to undo or that speak for the company without a person agreeing first. Sending to new contacts, quoting prices, changing contract terms and deleting records all belong behind an approval step until you trust the output.
Write these limits down before you switch anything on. A one-page list of what the agent may do alone, what it must ask about and what it must never do is more useful than any setting buried in a tool.
How do you start using an AI agent in sales
Start with one workflow that has a clear input, a clear output and a person who checks the result. Pre-call research is a good first choice: the agent reads the account, the seller reads the brief, and a wrong fact costs nothing because nothing was sent.
Once the output is reliable, move to drafting outreach with approval, then to wider tasks. Keep checking a handful of finished tasks every week, even after the agent has earned more freedom.
Where Pebbles Ai fits
- CapabilityAI workspaceIn Assistant chat, the Marketing and Strategy assistants take on research, prospecting and writing tasks. Auto, Plan and Manual modes set how much they do alone; the web app starts in Manual, where actions that reach the outside world wait for your approval.
- CapabilityProspecting & customer recordsDescribe your ideal customer; your assistant finds matching companies, then the right people at up to ten of them, and the leads land as contacts in the CRM (in Beta).
Common questions
How much of a seller's day can an AI agent take on?
Mostly the preparation: research, contact finding, first drafts, notes and record hygiene. Conversations, negotiation and judgement about the relationship stay with the seller, so the agent frees time rather than replacing the role.
Do AI sales agents need access to my CRM?
Most need at least read access to be useful, because the CRM tells them who you already know. Give write access only for the fields and actions you are happy for them to change, and review those changes at first.
Can an AI agent run sales without people?
Not well. Agents handle preparation and first drafts, but buyers still expect a person for real conversations, negotiation and judgement about the relationship.
About this page. Written by the Pebbles Ai team for people who run sales, marketing and growth. Last updated .
