An AI sales brief for every contact, built from their real timeline
Before a call you want to know where things stand. The agent brief reads everything logged on a contact and gives you a summary, a score, a next step and a reply you could send right now.
What it reads
Every email, note and logged conversation on a contact sits on one timeline, and the brief is generated from that timeline rather than a form you fill in. There is nothing to prepare beforehand: open the contact, ask for a brief, and it reads what is already there.
To keep a request bounded, the brief looks at roughly the last fifty communications, each trimmed to about two thousand characters. Long-running accounts get the most recent shape of the relationship rather than the entire history back to day one.
What comes back
The output has a fixed shape: a short written summary, an engagement score out of 100, a recommended next action, and a draft reply. The structure is enforced on the model’s response, so you get all four fields every time rather than a paragraph of prose you have to parse for the useful part.
The draft reply is a starting point, not a send button. Read it, change what does not sound like you, and send it from the contact the same way you would any other reply.
Your choice of model
A workspace picks OpenAI or Anthropic as its AI engine, and the brief is requested in the same structured format from either one. Switching providers does not change what a brief looks like, only which model wrote it.
The same reasoning is available to Claude and ChatGPT directly through the MCP connector, which can pull the same brief for a contact and act on it inside the chat you are already using.
Where it sits
A brief lives on the contact record it was generated for, next to the same timeline it read. If the contact later comes through the shared inbox or gets booked through a booking link, that shows up in the next brief you ask for.
Common questions
- What does an AI sales brief actually contain?
- Four things: a short summary of the relationship so far, an engagement score out of 100, a recommended next action, and a draft reply you can edit or send as is. It is generated from the contact's own timeline, not a template.
- How much of the history does it read?
- Roughly the last fifty communications on the contact, each trimmed to about two thousand characters. That is enough to catch the shape of a relationship without sending an unbounded amount of text to the model on every brief.
- Does it work with OpenAI or Anthropic?
- Either. A workspace picks its provider, and the brief is asked for in a fixed structure either way, so the summary, score, next action and draft reply always come back in the same shape regardless of which model wrote them.
- What happens if the AI provider is down?
- The brief request fails with a clear error rather than a mangled or partial result, and nothing on the contact's record is touched. Nothing about the account or the CRM depends on the brief succeeding.