Recipes / Content gaps from sales calls: buyer questions your website doesn't answer

Content gaps from sales calls: buyer questions your website doesn't answer

Fireflies + Firecrawl

IntermediateMarketing

Who it's for

Content and product marketers who want their next pages to answer what buyers actually ask on sales calls. Fireflies can be swapped for Gong or Granola, and Firecrawl for Tavily or Apify.

What you get

Gap table and the first content brief. Sample output: every company and person in it is fictional.

Calls read: 24 Fireflies sales calls, Jul 1 to Sep 30. Questions found: 61, in 9 topics.
Site: 142 URLs mapped on fernhill.example (Fernhill Software, fictional). 18 pages scraped after approval.

| Topic | Companies asking | Answered on site? | Best page now |
|---|---|---|---|
| SSO and SCIM setup | 9 | No | none |
| EU data residency | 7 | Partly | /security (one line, no regions) |
| Moving off spreadsheets | 6 | No | none |
| Seasonal user pricing | 5 | Partly | /pricing (no seasonal seats) |
| API rate limits | 4 | Yes | /docs/api/limits |

Brief 1: How SSO and SCIM setup works
- Who asks: IT leads in late-stage calls, for example at Quarry Lane Analytics (fictional) and Larkspur Freightworks (fictional)
- In buyers' words: "Do we have to set up SCIM ourselves?" (Sep 12, 'Quarry Lane security review'). "Which identity providers do you support?" (Aug 28, 'Larkspur technical call')
- What the site says now: /security mentions SSO in one bullet. Nothing on SCIM.
- Suggested format: a help article, plus a short section on /security
- Confirm with product: supported identity providers, typical setup time

Nothing was published or changed.

Time and cost

Estimate About 30–45 minutes, most of it reading the questions and briefs.

Assumptions: One quarterly run over about 25 calls and one website of about 150 pages, scraping about 20 of them.

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: ClaudeChatGPTCursor

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.

Developer setup

Auth setup and mcp.json for Claude Code and Cursor
  • Fireflies: Sign in with OAuth, or pass an API key as a Bearer token for headless use.
  • Firecrawl: Sign in with OAuth, or pass an API key as a Bearer token for headless use.

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": {
    "fireflies": {
      "url": "https://api.fireflies.ai/mcp"
      // headless alternative: "headers": { "Authorization": "Bearer YOUR_FIREFLIES_API_KEY" }
    },
    "firecrawl": {
      "url": "https://mcp.firecrawl.dev/v2/mcp-oauth"
      // headless alternative: "headers": { "Authorization": "Bearer YOUR_FIRECRAWL_API_KEY" }
    }
  }
}

Step-by-step prompts

  1. List my Fireflies meetings between {start_date} and {end_date} that were external sales calls with prospects: date, title, company and attendees. Leave out internal, onboarding and support calls, and give me the total. Wait for me to confirm the list.
  2. Search the transcripts of the confirmed calls for questions the prospect asked about the product, pricing, security, setup and use cases. Quote each question with the meeting title and date, group them into topics, and show the number of distinct companies per topic. Don't change anything in Fireflies.
  3. Use Firecrawl to map {website} and show me the URLs grouped by section. Suggest up to {max_pages} pages most likely to answer the top topics, with the credit cost. Scrape them only after I approve it.
  4. For each topic, say whether the scraped pages answer it fully, partly or not at all, and quote the line that does. Show me a table: topic, companies asking, answered, and best page now.
  5. For the top {n} gaps, write a content brief: working title, who asks and at what deal stage, the questions in buyers' words, what the site says now, a suggested format, and facts to confirm with product. Leave out buyer names, and keep it here as text.

Run it on a schedule (optional)

Where this goes wrong

Gotchas

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 content-gaps-from-calls.md