How to Turn a Call Transcript into a Case Study (Step by Step)

The fastest way to produce a B2B case study is to stop scheduling case study interviews. The material already exists: it's sitting in your call transcripts from Gong, Fireflies, Zoom, Fathom, Otter, or wherever your calls get recorded.

A discovery call contains the customer's problem in their own words. An onboarding call contains how they use the product. A QBR or renewal call contains the results. Together, those transcripts are a complete case study — unstructured.

This guide shows you how to structure it.

What you're extracting from a transcript

Every case study reduces to four components. Read (or process) the transcript hunting specifically for these:

1. The challenge. Look for the "before" language: "we were spending," "the problem was," "we kept running into," "it used to take us." The strongest challenge statements include a cost — time, money, or risk.

2. The decision and solution. Anything about why they bought and how they use the product: "what sold us was," "now we just," "every week we." Workflow descriptions beat feature mentions.

3. Results. Numbers first: percentages, hours saved, deals influenced, headcount avoided. Then directional statements: "way faster," "we stopped doing X entirely." Directional statements can often be firmed into numbers with one follow-up question.

4. Quotable lines. Short, emotional, specific sentences that sound like a human said them — because one did. "I genuinely don't know how we did this before" is worth more than any paragraph you could write.

The manual process (60–90 minutes)

1. Pull the transcript(s). One call works; multiple calls across the customer lifecycle work much better — discovery for the challenge, onboarding for the workflow, QBR for the results.
2. Highlight in four colours, one per component above. Don't edit yet — just tag.
3. Fill the skeleton: headline result → snapshot (industry, size, use case, metrics) → challenge → solution in practice → results → pull quote. (Full structure in our B2B case study writing guide.)
4. Clean the quotes lightly. Remove filler words and false starts, but keep the customer's phrasing. Over-polished quotes read as fake.
5. Send the draft for approval. Because everything came from the customer's own words, approval is usually a same-day "looks great" rather than a legal review cycle.

The AI-assisted process (minutes)

If you're doing this more than occasionally, extraction is exactly the kind of structured work AI does well — with two caveats.

A generic ChatGPT prompt ("turn this transcript into a case study") tends to produce plausible-sounding output that invents connective tissue, flattens quotes, and misses that the real results were mentioned in a different call. To get reliable output you need:

• Two-pass processing: first extract and verify the raw components (with quotes tied to the actual transcript text), then draft from the verified extraction. Single-pass generation is where hallucinated metrics come from.
• Multi-transcript ingestion: the challenge, workflow, and results usually live in different calls. Processing them together produces a complete story; processing one call produces a fragment.

This is the core of what Casily does: forward transcripts by email, webhook, or file upload, and it runs structured extraction and drafting, producing a case study in a consistent format with quotes traceable to the source calls. It also tags each proof asset (industry, use case, persona, result type) so your team can later match the right case study to the right prospect — which is where case studies actually earn revenue.

Which calls make the best source material?

Ranked by yield:

1. QBRs and renewal calls — results and ROI language, unprompted
2. Success check-ins — workflow detail and quotable enthusiasm
3. Onboarding calls — the "before" state, fresh and specific
4. Discovery/sales calls — the challenge and what they'd tried before
5. Support escalations that ended well — surprisingly strong "moment it paid off" stories

Do I need the customer's permission to use a transcript?

Yes — treat this carefully. Recording consent (which you should already have for the call itself) is not the same as publishing consent. The clean process: draft internally from the transcript, then send the customer the finished draft with an explicit request to publish, offering a named version and an anonymised fallback. Nothing goes public until they approve in writing. In practice, sending a finished draft gets approval far faster than asking for an interview ever did.