Most marketing teams sit on a goldmine they never open. Every discovery call, demo, and follow-up email contains the exact language your buyers use, the objections that kill deals, and the outcomes they actually care about. That intelligence almost never makes it into content or ad copy.

This guide walks you through a repeatable pipeline that changes that. Drop your call transcripts into a local folder, run a single prompt, and get back a structured analysis of every call, a prioritized content plan mapped to real objections, and platform-ready ad copy for Google and Meta — all grounded in what your buyers actually said.

Folder input Transcripts + skill files Read all transcripts CI analysis Produces CI report Content gap planner Blog + case study plan Ad copy skill Google + Meta formats Content plan content-gap-plan.xlsx Ad copy output RSAs + Meta variants

What you need before starting

Claude Code or Cowork. This pipeline runs inside either Claude Code (the CLI tool) or Cowork (the desktop app for non-developers). Both give Claude access to your local file system, which is what makes the transcript folder workflow possible. If you have not set up either yet:

The four files below — download all of them and keep them together in a folder you will reference as your working directory.

The files

Download all four and place them in a single folder on your machine — for example, /Users/yourname/call-pipeline/.

.MD conversation-intelligence-skill.md Analyzes transcripts and produces a structured CI report. Download .MD content-gap-planner-skill.md Reads the CI report and outputs a prioritized content plan as an xlsx file. Download .MD google-ads-copy-skill.md Reads the CI report and writes platform-ready Google RSAs and Meta ad copy. Download .MD pipeline-prompt.md The master prompt — paste this into Claude to run the full pipeline. Download

Step 1 — Install the skills

The three .md skill files need to be installed in Claude before you run the pipeline. Skills are instruction sets that tell Claude how to behave when it performs a specific task.

In Claude Code:

# From inside your project folder
claude skills add conversation-intelligence-skill.md
claude skills add content-gap-planner-skill.md
claude skills add google-ads-copy-skill.md

In Cowork:

Open the Skills panel, click "Add Skill," and import each .md file one at a time. They will appear in your skill library and activate automatically when Claude detects the right context.

You only need to install the skills once. They persist across sessions.

Step 2 — Set up your transcript folder

Create a dedicated local folder for your call transcripts. A clean structure looks like this:

/call-pipeline/
  ├── transcripts/
  │   ├── 2025-03-discovery-acme.txt
  │   ├── 2025-03-demo-globex.txt
  │   └── 2025-04-followup-initech.txt
  ├── outputs/
  └── (skill files live here)

A note on transcript format: Plain text (.txt) and PDF files both work. Most call recording tools (Gong, Chorus, Fireflies, Otter) let you export transcripts as one of these formats. If your tool exports in a proprietary format, copy the transcript text into a plain .txt file.

How often should you run this?

Weekly or monthly batches work best. Analyzing one call at a time is possible but misses the pattern-recognition that makes this pipeline valuable. When you drop in five or ten calls from the same ICP segment and run them together, Claude identifies which pain points came up repeatedly, which objections kept surfacing, and which features generated the most consistent enthusiasm. A pain point that showed up in three out of five calls becomes a headline. One that appeared once stays a footnote.

A practical cadence:

Step 3 — Run the pipeline

Copy the prompt below, fill in the three placeholders, and paste it into Claude Code or Cowork.

Fill in before running:

I have a folder at [FOLDER PATH] that contains:
- One or more call transcripts or email threads (plain text or PDF)
- The following installed skills: conversation-intelligence, content-gap-planner, google-ads-copy

Please run the full pipeline in order:

---

## STEP 1 — Read All Transcripts

Read every transcript or email thread file in the folder. If there are multiple files,
treat them as a batch — analyze them together, not one at a time. Note the filename
and any identifiable context (participant roles, date, deal stage) for each.

---

## STEP 2 — Conversation Intelligence Analysis

Using the conversation-intelligence skill, analyze the full batch and produce a single
consolidated CI report.

If there are multiple transcripts, synthesize findings across all of them. Where a pain
point, objection, or signal appeared in more than one conversation, flag it as a pattern
and weight it accordingly — patterns carry more copy and content weight than one-off
mentions.

The CI report must include all eight sections:
1. Pain Points
2. Desired Outcomes
3. Objections
4. Positive Reactions
5. Commitment Signals
6. Deal Risks
7. Unresolved Questions
8. Recommended Next Actions

Include Overall Sentiment, Deal Stage Read, and Confidence Level at the bottom.

Output the full CI report in the chat before proceeding to Steps 3 and 4.

---

## STEP 3 — Content Gap Plan

Using the content-gap-planner skill and the CI report produced in Step 2, generate a
content plan spreadsheet.

- Content types: Blog / SEO Articles and Case Studies only
- Map every piece to the specific objection or gap it addresses (not funnel stage)
- Assign priority based on frequency and deal impact
- Save the output as: content-gap-plan.xlsx

---

## STEP 4 — Ad Copy

Using the google-ads-copy skill and the CI report produced in Step 2, write ad copy
for [PLATFORM(S)].

- Audience temperature: [AUDIENCE TEMP]
- Pull hooks, objection-handling copy, and outcome language directly from the CI report
- Flag which specific lines or angles came from the transcripts (CI-sourced)
- For Google: produce one RSA ad group per major pain point or objection theme
- For Meta: produce Feed variants (pain-led, outcome-led, social proof-led) + Story/Reels

---

## DELIVERY

When all steps are complete, confirm:
- [ ] CI report is visible in chat
- [ ] content-gap-plan.xlsx is available for download
- [ ] Ad copy is formatted and ready to review in chat

If any transcript file cannot be read, flag it by name and continue with the remaining
files rather than stopping.

What you get back

Claude runs the four steps in sequence and produces three outputs:

1. The CI report (in chat)

A structured analysis of everything your buyers said across all the calls in the batch. Eight sections: pain points, desired outcomes, objections, positive reactions, commitment signals, deal risks, unresolved questions, and recommended next actions. Every finding is labeled as explicit (the customer said it directly) or inferred (implied by context), with supporting quotes.

Where a finding appeared across multiple calls, it is flagged as a pattern. These patterns are the highest-signal inputs for content and copy — they reflect what your market consistently believes, fears, and wants, not just what one prospect happened to say on one call.

2. content-gap-plan.xlsx

A downloadable spreadsheet with one row per content piece. Each row maps directly to an objection, pain point, or unresolved question from the CI report. Columns include the proposed article title, the angle and key message, target keyword or search intent, the recommended CTA, and priority (High / Medium / Low). A summary tab shows counts by gap type, content type, and priority.

The content types are intentionally narrow: blog/SEO articles and case studies only. Blog articles go toward searchable pain points and unanswered questions. Case studies go toward ROI objections and "prove it works" hesitation. The skill decides which based on whether the buyer expressed doubt about outcomes (case study) or how something works (blog).

3. Ad copy (in chat)

Platform-specific copy formatted and ready to review:

A few things worth knowing

The CI report outputs before ads and content run. This is intentional. Read it before the pipeline continues to the next steps. If the analysis missed something or misread the tone, you can flag it and course-correct before it flows downstream into your content plan and ad copy.

Patterns beat one-offs. A single call where a prospect mentioned pricing concerns tells you one thing. Five calls where pricing came up in the same way tells you it is a systemic objection your messaging is not handling. The pipeline surfaces patterns automatically, but the more transcripts you include in a batch, the stronger the signal.

Voice-of-customer copy is the point. The ad copy skill is specifically instructed to use the customer's actual language — not a polished, sanitized version of it. The most effective B2B ad copy often sounds like something a frustrated buyer would say to a peer, not something a marketing team wrote in a strategy session. That language lives in your transcripts. This pipeline puts it directly into your headlines and hooks.

Email threads work too. If your sales team runs deals primarily over email, paste the thread content into a .txt file and include it in the transcripts folder. The CI skill handles email threads the same way it handles call transcripts.

Built with Claude Code / Cowork. Skills created for B2B SaaS growth and demand generation teams.