1:1 coaching agenda: call habits from Gong and CRM cleanup for each rep, saved to Notion
Gong + Salesforce + Notion
Who it's for
Sales managers who want each weekly 1:1 to start from the rep's real deals: what Gong says about how those calls went, and what needs tidying in Salesforce. It's built for coaching, not scorekeeping. Gong can be swapped for Fireflies or Granola, Salesforce for HubSpot, Attio or Pipedrive, and Notion for keeping the agenda in chat.
What you get
One rep's 1:1 agenda, waiting for approval before it goes to Notion. Sample output: every company and person in it is fictional.
1:1 agenda: Alex Rivera (fictional), Thu Oct 22 Based on Alex's 3 largest open opportunities. Gong's answers are about these deals only, not a full picture of Alex's calls. What's working - Fernhill Software (fictional), $72K: Gong says the last two calls each ended with a dated next step the buyer agreed to. - Quarry Lane Analytics (fictional), $33K: Gong says Alex brought the buyer's security lead in on the second call, before procurement asked. Worth talking through - Larkspur Freightworks (fictional), $46K: Gong says pricing came up on the first call and no one with budget has joined since. Question: who else has to agree to the spend, and how could we meet them? Salesforce cleanup (nothing changed) - Larkspur Freightworks: close date Oct 10 has passed - Mossgate Labs (fictional), $9K: no next step - 2 opportunities with no contact roles Alex's topics - (left blank for Alex) Not included: talk ratio and other call metrics. Gong's MCP tools don't return them. Gong found all 3 deals. Notion page 'Alex Rivera 1:1, Oct 22' not created yet. Reply 'approve' to create it under 1:1 Notes.
Time and cost
Estimate About 10–15 minutes per rep, most of it editing the agenda.
Assumptions: One weekly run for a team of 5 reps, with 3 deals reviewed per rep.
- Gong: One ask_deal call per deal reviewed, so about 15 for 5 reps. Each call uses Gong credits based on the calls and emails analyzed. Every call uses credits. All three tools (ask_account, ask_deal, generate_brief) consume Gong credits based on the volume of calls and emails analyzed; briefs cost more per open-ended section. Same consumption as the equivalent API requests. Source: help.gong.io, checked 2026-10-16
- Salesforce: Reads only: one schema lookup and one or two SOQL queries for the team's open opportunities and their contact roles. Each call counts against the org's daily API request quota. Every call uses credits. No paid credits; every tool call counts against the org's daily Salesforce API request quota (one or more API calls per tool). Source: github.com, checked 2026-10-16
- Notion: One search for the parent page, then one page created per rep after you approve it. Credit use not confirmed by Notion.
Exact credit costs depend on your plan with each vendor. Check the tool pages before a large run.
MCP servers
Every server is confirmed in: ClaudeChatGPT
-
ClaudeChatGPTRead onlyEvery call uses creditsNeeds admin or IT
https://mcp.gong.io/mcp -
ClaudeChatGPTCursorRead + writeEvery call uses creditsNeeds admin or IT
https://api.salesforce.com/platform/mcp/v1/platform/sobject-all -
ClaudeChatGPTClaude CodeCursorCopilotRead + writeCredits not confirmed
https://mcp.notion.com/mcp
Set up in a chat app
Every server in this recipe is confirmed for Claude and ChatGPT. Connect each one, then paste the prompts below into one conversation.
- Connect Gong without code (Claude, ChatGPT)
- Connect Salesforce without code (Claude, ChatGPT)
- Connect Notion without code (Claude, ChatGPT)
Developer setup
Auth setup and mcp.json for Claude Code and Cursor
- Gong: Sign in with OAuth in your client (run
/mcpin Claude Code). Do it once interactively, and the client stores and refreshes the token. - Salesforce: Sign in with OAuth in your client (run
/mcpin Claude Code). Do it once interactively, and the client stores and refreshes the token. - Notion: Sign in with OAuth in your client (run
/mcpin Claude Code). Do it once interactively, and the client stores and refreshes the token.
Put this in .mcp.json (Claude Code, project root) or ~/.cursor/mcp.json (Cursor). Swap the example keys for your own, and don't commit real ones. Remove the // comment lines if your client requires strict JSON.
{
"mcpServers": {
"gong": {
"url": "https://mcp.gong.io/mcp"
},
"salesforce": {
"url": "https://api.salesforce.com/platform/mcp/v1/platform/sobject-all"
},
"notion": {
"url": "https://mcp.notion.com/mcp"
}
}
}
Step-by-step prompts
- Look at the Opportunity schema in Salesforce, then use SOQL to find open opportunities owned by {reps}. For each rep, flag any with a close date in the past, an empty next step, no contact roles, or no activity in {days} days. Show one table per rep. Don't change any records.
- For each rep, take their {n} largest open opportunities. Use Gong ask_deal once per deal to ask how the recent calls went: whether a next step was agreed, who from the buyer joined, and which questions or objections came up. Say which deal each answer is about, and tell me when Gong can't find a deal.
- Turn this into a 1:1 agenda per rep: two things that are working, one or two things to talk through framed as questions, the Salesforce cleanup list, and a blank slot for the rep's own topics. Don't add talk ratios, scores or comparisons with other reps. Show me each agenda.
- Search Notion for {parent page}. Create one page per rep under it, titled '{rep} 1:1, {date}', only after I approve it. Don't change any other page.
Run it on a schedule (optional)
- Run it the day before your 1:1s, so you have time to edit each agenda and the rep can see it ahead of time.
- In a scheduled run, stop at the drafts and send them to yourself. Create the Notion pages only after you've read and approved each one.
- Gong and Salesforce both need an admin to set up the connection first. Test one manual run before you schedule it.
Where this goes wrong
- Three deals read as a verdict on the rep. Gong's answers cover only the deals you asked about. Label every observation with its deal, ask the rep for context before drawing a conclusion, and drop anything Gong answered vaguely.
- A call habit Gong didn't actually report. It's easy for a summary to drift into numbers like talk ratio or question counts. Check that every line under 'What's working' and 'Worth talking through' traces back to a Gong answer for a named deal, and delete anything that doesn't.
- Quiet in Gong read as no calls. Gong excludes private calls and may not match every deal. List the deals Gong couldn't find separately instead of treating them as a coaching point.
- Another rep's deals in the list. Opportunity owner and account owner can differ. Filter on the opportunity owner, and check each rep's count against their pipeline view in Salesforce once.
Gotchas
- Gong's MCP server has three tools: ask_account, ask_deal and generate_brief. None of them returns per-rep call stats, so this recipe asks about each rep's top deals. It won't give you talk ratios or scorecards; use Gong itself for those.
- Gong needs admin setup: a Gong tech admin must create an MCP server integration first, and its authorization context (personal or shared) is fixed when it's created. Gong is read-only, excludes private calls, needs a paid seat and charges Gong credits per call based on how many calls and emails it analyzes.
- Salesforce needs admin or IT setup: its MCP server needs Enterprise Edition or above, and an admin must create an External Client App for each client. In ChatGPT it also needs developer mode. Ask for the read-only platform/sobject-reads server, and note each tool call counts against your org's daily API quota.
- Notion's help center names ChatGPT Pro as the supported ChatGPT tier. If you use ChatGPT on another plan, check that Notion connects before relying on this recipe.
- Share each agenda with the rep a day ahead and leave room for their topics. A 1:1 that opens with a list of their mistakes feels like an audit.
- Using HubSpot instead? HubSpot's connector covers CRM search and activities, so the cleanup checks work the same on deals and their logged activities. HubSpot also needs admin setup: non-admin users need a Super Admin to approve the connector first.
Download
Get the prompts, guardrails and expected output as one Markdown file. Paste it into a Claude Project or ChatGPT project's instructions, or save it as CLAUDE.md for Claude Code.
Download coaching-1on1-agenda.md