Mine closed-won calls for winning narratives and map the content gaps
A Slack bot connected to call recordings and the CRM finds the stories that keep coming up in won deals, then the team maps which ones lack a one-pager, slide or page. Reps get content built from what buyers said.
Paste this into ChatGPT, Claude or your agent tool. See the prompt
- Content and Collateral
- beginner
- Sales leader
- Founder
- SDR manager
- Category
- Content and Collateral
- Difficulty
- Beginner
- Est. cost
- Not stated by source
Works with
Problem
Content plans are built from keywords and opinions rather than from what wins deals, so reps improvise on their best messaging. Nobody checks which winning narratives have no asset a rep can send.
Setup
An internal Slack bot connects Fathom call recordings with HubSpot, so a marketer can ask in plain language for the narratives that recur across 12 months of closed-won calls and get back a ranked list with deal links, call references and quotes. The narratives go into a spreadsheet matrix against nine asset types, from an indexed web page to a sales slide and one-pager. Reps rate which assets they actually use in five selling situations, and gaps are prioritized by how often a narrative wins times how many assets are missing. A follow-up prompt turns each narrative into buyer-phrased search and AI-assistant queries to track each quarter. Teams without an integration can export closed-won transcripts into a Claude Project and run the same prompts.
Prompt
Starter prompt by The GTM Agent Weekly, written from the setup described on this page. It is not the source's own prompt; see the source link below for their write-up.
# ROLE
You analyze our closed-won sales calls to find the narratives that win deals, then help map
where our content is missing.
# INPUTS
- {{TRANSCRIPTS}}: closed-won call transcripts from the last {{MONTHS}} months, each with deal name and link.
- {{ASSET_TYPES}}: the content types we track (e.g. web page, sales slide, one-pager, case study).
- {{ASSET_INVENTORY}}: what we already have for each narrative, if known.
# STEP 1: WINNING NARRATIVES
Find the narratives (problem, reason to buy, proof point) that recur across deals. For each:
- a one-line name and description,
- how many deals it appears in, with deal links,
- two or three verbatim buyer quotes with call references.
Rank by number of deals.
# STEP 2: CONTENT MATRIX
Make a table: narratives as rows, {{ASSET_TYPES}} as columns. Mark each cell "have",
"missing" or "unknown" from {{ASSET_INVENTORY}}. Prioritize gaps: narratives that win most
often with the most missing assets come first.
# STEP 3: BUYER QUERIES
For each top narrative, write five search or AI-assistant queries a buyer would type, in
their words, to track each quarter.
# GUARDRAILS
- Quotes must be verbatim with a call reference. Count only deals where the narrative clearly appears.
- Do not use lost deals in this analysis. Replace every {{PLACEHOLDER}} before running. Works with any capable LLM and runner.
Tools
Steps
- Ask for the narratives and value propositions that recur across the last 12 months of closed-won calls only, ranked by frequency, with quotes and deal context.
- Build a matrix with narratives as rows and asset types as columns, marking each cell strong, needs work or missing.
- Spend 30 minutes with reps on which assets they actually send for cold outreach, discovery follow-up, objections, multi-stakeholder buy-in and stalled deals.
- Prioritize the most frequent winning narratives with the most missing assets, starting with assets that can be cut into several others, such as a webinar with a transcript.
- Turn each narrative into buyer-phrased queries, check them in AI assistants each quarter, and tell sales whenever a new asset ships.
Results
The author reports 2 to 4 hours with a working integration, half a day to a day on the manual path, and about 15 minutes to turn the finished matrix into a backlog. Reps preferred short video clips over full webinars and still asked for one-pagers. No pipeline impact numbers are given.
What goes wrong
Gotchas
- Use closed-won deals only; mixing in good calls that went nowhere blurs which narratives actually win.
- Reps did not share full webinar recordings, while short clips and one-pagers got used, so ask reps before producing anything.
- The output names clients and quotes them, so strip or clear those details before any of it goes into public content. (editor note)
- A bot's ranking is only as good as the transcripts behind it; spot-check the quotes against the actual calls. (editor note)
Get 5 like this every week, free.
Real agent setups with sources and failure modes, in one short weekly email.
Free. One email a week. Unsubscribe anytime.
Source and credit
Yulia Olennikova, N.Rich, in GTM Strategist (Maja Voje) · newsletter
Source dated 2026-10-02 · last reviewed 2026-10-12. Summarized in our own words; see the source for details.