Pre-Meeting Brief

The Grow2.ai AI agent prepares a short brief for the manager before each meeting in the calendar. Instead of reading through all correspondence before a call, scrolling the CRM, and checking three communication channels, the salesperson sees a ready-made summary — who the contact is, what was agreed last time, which questions are still open, and what preferences are known.

Automation removes the routine of gathering context from the salesperson and turns 15 minutes of preparation into 30 seconds of reading a ready-made brief.

What automation does

1. Tracks new events in the work calendar (Google Calendar, Outlook, or equivalent).
2. Identifies external meeting participants by email addresses.
3. Pulls the contact card and linked deal from the CRM.
4. Collects recent emails, messages, and call transcripts from communication channels.
5. Summarizes the interaction history: key topics, commitments, objections, pain points.
6. Extracts structured facts — the contact's role, company size, deal stage, open action items.
7. Generates a short brief of 150–250 words in the event description or in a separate channel — for example, a Slack notification 15 minutes before the meeting.
8. Attaches links to source materials so the manager can quickly verify a detail.

When the brief appears

Automation triggers on two events: when a new calendar event is created (the brief is generated in advance and updated when changes occur) and a set interval before the meeting — for example, 30 minutes before — to account for recent emails and messages. The delivery format is configurable for the team: some salespeople prefer to see the brief directly in the event description, others prefer it in personal messenger messages.

What automation does NOT do

1. It does not make decisions for the manager. The brief is a prompt, not a call script; the tone of the conversation, the offer, and the counterarguments remain with the human.
2. It does not replace the CRM. Automation reads data but does not update deals, move stages, or create tasks after a meeting — those are separate scenarios.
3. It does not guess facts that are not in the history. If a client has never discussed the budget in correspondence, the brief will not invent a number — it will honestly note that the information is absent.

Automation runs in the background and requires no new actions from the manager: the brief appears on its own in the event description and in the chat, and the salesperson simply reads it in those same 30 seconds.

The technical foundation of the briefing is a four-step chain: a calendar trigger, data collection from three sources, LLM summarization, and delivery of the brief to the channel where the manager will see it. The pipeline is built on a low-code platform.

Data flow

1. The calendar (Google Calendar, Outlook) sends a webhook or is polled on a schedule — the automation catches a new or modified event with an external participant.
2. The email addresses of external participants are matched against contacts and deals in the CRM (HubSpot, Salesforce, or equivalent). If there is no match, the agent marks the meeting as a 'cold contact' and generates a simplified brief based on public data and the company domain.
3. In parallel, the agent pulls the last N emails with the participant's address from the inbox, related threads from messengers (for example, the client's Slack channel), and transcripts of past calls from the recording tool, if it is connected.
4. The collected context is passed to the LLM with a prompt that describes the brief format, the priority of facts, and restrictions (do not fabricate numbers, do not repeat NDA data aloud).
5. The LLM response goes through structural validation (all required sections are filled in) and is posted to the event description in the calendar or to the specified channel — Slack, Telegram, or a corporate messenger.

Implementation steps

1. Inventory of sources — which email domains, CRMs, messengers, and call recording tools are actually used by the sales team.
2. Setting up access: calendar service account, CRM API key, OAuth connectors for email and messengers.
3. Building the pipeline on a low-code platform (workflow engine, Zapier, or equivalent) — trigger, three data collection branches, aggregation, and LLM call.
4. Prompt engineering and tuning: the team describes which facts are critical, which are secondary, and what brief length they find convenient. Iterations on 20–30 real meetings.
5. Pilot with 2–3 managers over 2 weeks with feedback after each meeting — what was useful, what was unnecessary, what was missing.
6. Rollout to the full team, setting up monitoring (how many briefs have been generated, the share of meetings with a brief, median time from trigger to publication).
7. Documenting the process: what to do if the brief did not arrive, where to report errors, who is responsible for the prompt.

Pipeline components

Quality and control

The automation includes three control loops: structural validation of the brief before publication, an explicit 'data not found' label for empty sections, and a feedback button in the message — the manager marks the brief as useful or not. Feedback is collected in a table and used to fine-tune the prompt. In the first few weeks, some briefs should be reviewed manually before publication to catch hallucinations and phrasings the team does not accept.

The automation connects to three systems, so prerequisites are divided into data, access, and team readiness.

Data and Access

• A work calendar (Google Calendar, Outlook, or compatible), where meetings are scheduled with external participants and it is clear from the email who the client is.
• A CRM with an active contact and deal history — records must be up to date at least in the basic fields: company, role, stage, recent activity.
• Client communication channels — corporate email and/or a messenger where correspondence is conducted. Access via OAuth or a service account.
• A call recording and transcription tool — optional, but significantly improves the quality of the brief if available.
• Access to LLM-API via a corporate account or an aggregator platform.

Team Readiness

• Sales managers have agreed that their correspondence and call recordings are read by the automation — this is stated in the policy.
• A process owner has been assigned: RevOps, Head of Sales, or an external automation partner. They are responsible for the prompt, monitoring, and fine-tuning.
• There is an agreed-upon brief format — which sections are needed, what length, and where to deliver it.

Timeline

Standard implementation takes 6–10 weeks: 1–2 weeks for source inventory and access setup, 2–3 weeks for assembly and prompt engineering, 2 weeks for a pilot with 2–3 managers, and 1–2 weeks for rollout and stabilization. Timelines grow if the CRM is maintained carelessly or the team is not ready to share correspondence — in that case, a separate data cleanup and policy alignment stage is needed before automation.