# Conversation Intelligence > Ask what was said, agreed, or pushed back on across your calls, meetings, and emails. Returns a synthesized answer with citations. Consumes AI action credits proportional to work done. **Date:** 2026-07-29 **Source:** https://gtm.ai/docs/mcp/tools/conversation-intelligence --- ## Overview Conversation Intelligence turns your calls, meetings, and emails into answers. Ask what was said, agreed, or pushed back on in plain language, and it reads the relevant conversations and hands back a direct answer with citations, no scrubbing through recordings. Scope it to a single engagement, an account, or a contact. It is the double-click on the timeline that [Browse Engagements](/docs/mcp/tools/browse-engagements) builds. Browse tells you a meeting happened. Conversation Intelligence tells you what came out of it, and what to do next. ## What You'll Get A written answer to the question you asked, grounded in what was actually said. The response includes: - A synthesized answer, written from the actual conversations. - Inline citations linking back to the source meetings. - The raw transcript or email body, when you scope to a single engagement and request content. ## Best For - Walking into a meeting knowing what was discussed and committed last time. - Pulling requirements, pain points, and blockers out of a specific call. - Getting an account-health read across recent conversations without replaying a single recording. - Drafting a follow-up grounded in what the customer actually said, with quotes. Lean on it for the narrative: what was said, what was committed, and what to do next. Precise tallies across months of calls are not what it is for. For discovering which meetings or emails exist in the first place, use [Browse Engagements](/docs/mcp/tools/browse-engagements). For a firmographic or CRM-based account briefing, use [Account Research](/docs/mcp/tools/account-research). For a contact-level briefing, use [Contact Research](/docs/mcp/tools/contact-research). ## How It Works Conversation Intelligence works from the calls, meetings, and emails your organization already captures. Provide exactly one scope per call: - **An engagement ID** (from [Browse Engagements](/docs/mcp/tools/browse-engagements)) to analyze one call, meeting, or email. Request the content option to also receive the raw transcript or email body. - **A company ID** to synthesize across recent engagements with an account. - **A contact ID** to synthesize across recent engagements with a person. The tool pulls the most recent transcripts and email threads for your scope, up to five of each, then synthesizes an answer. **Connected sources** - **Web-conferencing:** Google Meet, Zoom, Microsoft Teams - **Revenue-intelligence:** Gong, ZoomInfo Chorus - **Email:** synced threads The more sources connected, the fuller the picture. A conversation that just wrapped can take a little while to index. Topic-based discovery, such as "every call where SOC 2 came up," is on the roadmap. To cover a whole book of business today, run the tool across your accounts in turn. ## Technical Details ### Credits **AI action credits.** Each call runs a sub-agent that retrieves and reasons over your conversation content, so pricing reflects the work involved. Most questions cost around 10 credits, scaling with how much context the answer draws on. See [Credits & Billing](/docs/mcp/credits). ### Requirements This tool only returns results if your organization has connected at least one calendar, email, meeting, or revenue-intelligence provider to ZoomInfo. Results are limited to conversations your user has permission to see. Your ZoomInfo admin can confirm what is connected. ### Automatic Tool Selection The AI invokes Conversation Intelligence when you ask what happened in a conversation, or what is going on with an account based on recent talks. To find which meetings exist in the first place, it reaches for [Browse Engagements](/docs/mcp/tools/browse-engagements). For a firmographic or CRM-based account briefing, it uses [Account Research](/docs/mcp/tools/account-research). The natural chain is Browse Engagements to find the meeting, Conversation Intelligence to analyze it, then a drafting step for the follow-up. ### Architecture Conversation Intelligence is a context agent, like [Account Research](/docs/mcp/tools/account-research) and [Contact Research](/docs/mcp/tools/contact-research). It returns a synthesized answer instead of raw transcripts, so results are ready to act on. Under the hood it runs a sub-agent that pulls the most recent transcripts and email threads for your scope, then returns only what serves your question. A single transcript can run to tens of thousands of characters; the sub-agent absorbs that volume so the primary orchestrator's context window does not have to. ## Example Prompts > *"What were the main objections on my last call with Acme Corp, and who raised the one about pricing?"* > *"Summarize where things stand with this account before my renewal call."* > *"Pull the requirements and any blockers the customer mentioned in this meeting."*