How to prep for a customer call with your team's history and an AI agent
A five-step prep routine that uses an AI agent and your team's meetings, email and calendar, so you walk into every customer call knowing the history.
To prep for a customer call with an AI agent, ask it for three things: what your team has already discussed with this customer, what is still open, and what you should ask next. The agent is only as good as the history behind it, so the real work is giving it the whole team's history, not just yours.
Done well, this takes about five minutes and replaces rereading old notes.
Why customer prep goes wrong
Most prep fails the same way. You remember your own last call. You do not know what your colleague discussed with the same customer two weeks ago, or what the customer wrote in an email thread you were not on. So you ask a question that was already answered, or miss a commitment someone made.
An agent can close that gap, but only if it can see what the team has heard. If you only paste in your own notes, it will give you a tidy summary of an incomplete picture.
What the agent needs access to
For customer prep, three kinds of history matter:
- Meetings: what was said, by whom, and what was agreed.
- Email: what the customer wrote, and what you promised in reply.
- Calendar: who is meeting the customer and when, including colleagues.
If your setup only covers one of these, you will still get something useful, but you will miss things. How to give every agent on your team the same company context covers how to set this up.
The five-step prep routine
- Ask for the history. "Summarize everything our team knows about this customer, in date order. For each point, say which call or message it comes from."
- Ask what is open. "What did we promise them, and what did they promise us? List anything that has no follow-up yet."
- Ask who else has talked to them. "Which of my colleagues have spoken to this company, and when? What did they learn?"
- Ask for gaps and questions. "Based on this history, what do we not know yet that would help on this call? Suggest five questions."
- Check the sources. Open the two or three most important claims and read the quote behind them. If the agent cannot point to a source, treat the claim as unverified.
Step 5 is the one people skip, and it matters most. An agent can sound sure about something that was never said.
A prompt you can reuse
Paste this, replacing the bracketed parts:
I have a call with [customer] at [time]. Using everything our team has recorded or written about them, give me: (1) a dated timeline, (2) open commitments on both sides, (3) who on our team has spoken to them, (4) what changed since the last call, and (5) five questions to ask. Cite the meeting or message for every point and quote the key line.
Mistakes to avoid
- Prepping from your own notes only. The point is the team's history.
- Accepting summaries without sources. If it cannot cite, do not repeat it to the customer.
- Asking for everything. A timeline of one customer beats a dump of every mention. Keep the question narrow.
- Forgetting access rules. Your agent should see what you can see. If a colleague's private thread is not visible to you, it should not appear in your prep, and that is how it should be.
How to do it with Nysa
Nysa builds one company brain from your team's meetings, email and calendar. A notetaker joins the calls you choose as a visible participant, Gmail history is imported, and Google Calendar stays in sync. Nysa turns these into facts with sources, so the answer to "what did they say about the renewal?" comes with a quote and a date.
Connect your agent once, and use the routine above:
- Claude: Connect Claude
- ChatGPT: Connect ChatGPT
In ChatGPT, for example, you can ask it to find a person or company, see what Nysa knows about them, and update their details. After the call, agents can also write back to the brain, so the next teammate who preps for the same customer starts from the new facts.
Each agent sees only what its teammate can see, and raw email stays private to its owner; teammates get the facts drawn from it, not the messages. Setup takes under 5 minutes.
If answers look thin, run through saved isn't searchable to find out whether the call or thread actually reached your agent.
FAQ
Can an AI agent prep me for a call with a customer we have never spoken to?
It can help with general research and question ideas, but it cannot supply history your team does not have. The value of team history is highest for existing customers with several touchpoints.
Should I let the agent write my questions or just summarize?
Start with the summary and open commitments, because those are checkable. Use the agent's suggested questions as a prompt for your own thinking, not a script.
How do I know the agent is not making something up?
Require a source for every claim and open the important ones. Facts in Nysa carry a quote, a date and where they came from, so you can check any claim in seconds.