Gong Automation

Overview of Gong Automation skill

What Gong Automation does

Gong Automation is a conversation-intelligence skill for using Gong through natural language via the Composio MCP integration. It helps an AI agent retrieve Gong call transcripts, recordings, analytics, speaker data, topics, trackers, and workspace information without requiring you to manually navigate Gong exports or API calls.

Best-fit users and jobs to be done

The Gong Automation skill is most useful for Revenue Operations, sales leadership, enablement, customer success, and analysts who need structured answers from Gong data. Common jobs include reviewing calls from a date range, finding transcripts for specific call IDs, summarizing deal conversations, comparing rep talk patterns, or pulling call evidence for coaching, forecasting, and pipeline inspection.

Why this is more than a generic prompt

A normal prompt can ask an AI to “analyze Gong calls,” but it cannot access Gong by itself. This skill defines Gong-specific workflows and tool usage through Composio, including date filters, call IDs, workspace IDs, pagination cursors, and transcript retrieval. That makes it better suited for repeatable Gong Automation usage where the agent must fetch real records before summarizing or analyzing them.

Important adoption considerations

Gong Automation depends on a connected Gong account and the Composio MCP server. It is not a standalone transcript parser and does not include extra scripts, reference files, or local data-processing code in the repository. Install it when you want an agent to operate Gong tools through MCP; avoid it if you only need offline analysis of transcripts you already exported.

How to Use Gong Automation skill

Gong Automation install and MCP setup

For a Claude Skills-style installation, use your skills installer against the repository path, for example:

npx skills add ComposioHQ/awesome-claude-skills --skill "Gong Automation"

Then configure the Composio MCP server in your client:

https://rube.app/mcp

When prompted, connect Gong using the supported OAuth or bearer-token flow. The skill requires the rube MCP connection; without it, the agent can read the instructions but cannot retrieve Gong data.

Inputs that make Gong Automation usage reliable

Strong Gong Automation prompts include the retrieval target, time range, timezone, workspace, and desired output format. For transcript pulls, use ISO-8601 dates when possible, such as 2025-02-01T00:00:00Z to 2025-02-10T23:59:59Z. If you know the call IDs, provide them instead of asking the agent to search broadly. If your Gong instance has multiple workspaces, include the workspace ID or workspace name to reduce irrelevant results.

Weak prompt:

“Analyze recent Gong calls.”

Stronger prompt:

“Use Gong Automation to retrieve transcripts for calls in workspace enterprise-sales from 2025-02-01T00:00:00Z through 2025-02-10T23:59:59Z. Summarize objections, competitor mentions, next steps, and rep coaching opportunities. Return a table with call ID, account, themes, risk signals, and recommended follow-up.”

Prompt patterns for common revenue workflows

For Revenue Operations, specify the business question before the retrieval request. Examples:

• “Find all Gong calls from last week for late-stage opportunities and identify pricing objections.”
• “Retrieve transcripts for these call IDs and compare customer pain points across them.”
• “Pull call analytics and speaker stats for a rep coaching review; flag talk-time imbalance and missed discovery questions.”
• “Get workspace-level Gong data needed to verify which team or workspace contains the calls before retrieving transcripts.”

This sequence helps the agent avoid jumping straight to summarization before it has the correct Gong records.

Repository reading path before first use

Start with composio-skills/gong-automation/SKILL.md. It is the only visible support file in the skill directory and contains the setup instructions, toolkit documentation link, workflow descriptions, and Gong tool names. Pay special attention to the parameter names shown for transcript retrieval, including filter__fromDateTime, filter__toDateTime, filter__callIds, filter__workspaceId, and cursor, because these determine whether the agent can paginate and narrow results correctly.

Gong Automation skill FAQ

Is Gong Automation for Revenue Operations?

Yes. Gong Automation for Revenue Operations is a strong fit when RevOps needs repeatable access to call evidence: transcript retrieval, call analytics, topic and tracker review, workspace checks, and sales-process inspection. It is especially useful when the output must support forecasting, coaching, deal reviews, or pipeline-quality analysis.

Does this replace Gong dashboards?

No. Gong dashboards are better for visual exploration inside Gong. The Gong Automation skill is better when you want an AI agent to retrieve specific Gong data and transform it into a written brief, table, coaching note, risk summary, or workflow-ready analysis.

What permissions or access are required?

You need a Gong account that can access the calls, transcripts, workspaces, or analytics you ask for. The MCP connection must be authenticated through Composio. If the connected user lacks access to a workspace or call, the skill cannot reliably retrieve it.

When should I not install this skill?

Do not install Gong Automation if your team does not use Gong, cannot authorize Composio MCP access, or only needs generic sales-call analysis from uploaded transcript files. It is also not the right tool for building a full custom Gong data warehouse; it is designed for agent-driven Gong tasks through natural language.

How to Improve Gong Automation skill

Strengthen Gong Automation prompts with business context

The best results come from pairing retrieval instructions with the decision you need to make. Instead of asking for “call insights,” say whether you are preparing a QBR, coaching a rep, auditing MEDDICC compliance, investigating churn risk, or validating forecast confidence. This helps the agent choose the right level of detail and avoid generic summaries.

Control scope with dates, IDs, workspaces, and pagination

Most failure modes come from broad or ambiguous retrieval. Use exact date ranges, call IDs, and workspace filters where available. For large ranges, ask the agent to paginate using the returned cursor and to report how many calls were reviewed. This is especially important for Gong Automation usage across active sales teams where a broad date range may return too much data.

Ask for evidence, not just conclusions

For executive or RevOps workflows, request citations from the retrieved call data: call ID, speaker, timestamp if available, customer quote, topic, or tracker reference. A useful prompt ending is:

“For every recommendation, include the supporting call ID and the transcript excerpt or analytic signal that led to it.”

This reduces unsupported coaching advice and makes the output easier to verify.

Iterate after the first output

Use the first result to narrow the next Gong Automation prompt. If the agent finds recurring objections, ask it to retrieve related calls or compare by rep, segment, workspace, or stage. If the output is too shallow, ask for a second pass focused on one dimension, such as competitor mentions, security concerns, pricing resistance, stakeholder mapping, or next-step quality.