Turn one customer call into a published case study draft

A structured case-study draft built from the transcript of a happy-customer call, with every metric flagged for the customer to confirm before you publish a word.

What you'll have when you're done

A repeatable way to turn customer conversations into marketing assets: pull the transcript of a call with a happy customer, run it through a case-study template, and get back a clean draft, the challenge, what they tried, your solution, the results, with real quotes pulled from what they actually said. You edit, send it to the customer for sign-off on the quotes and numbers, and publish. The thing that used to mean hiring a freelancer or never getting done becomes a same-day draft.

Case studies are the asset you never get around to making

Customer stories are among the most effective marketing assets a B2B company has, and most companies have almost none, because making one is a project. You have to interview the customer, transcribe it, draft it, chase quotes. So it sits on the someday list while your best proof points evaporate after every call. I have lost count of the calls where a customer said something I would have paid a copywriter to invent, "you cut our onboarding from a quarter to three weeks," and I just... nodded, finished the call, and never wrote it down. The proof was free and I let it expire, every time, because capturing it felt like a project I did not have an afternoon for.

But if you run Granola, the interview already happened and is already transcribed, every happy-customer call is raw case-study material. The draft is a prompt away. The one hard rule, and it is non-negotiable: the customer approves every quote and number before anything goes public. AI drafting plus customer sign-off is the safe, fast path; skipping the sign-off is how you publish a metric they never agreed to.

What you need first

• A Granola transcript of a call with a genuinely happy customer (a QBR, a renewal, a thank-you call).
• A case-study template you like: Challenge, What they tried before, Solution, Results, in your format.
• A Claude Project on a business plan (customer data is sensitive).
• A way to get written approval from the customer (a simple email reply works).

Step-by-step

Step 1Pull the transcript of a happy call

Find a recent call where a customer said good things with substance, ideally specifics ("we cut onboarding time in half"). The Granola transcript is your source material. If the praise was vague, this works better after a short, intentional "can I ask how it's going?" call.

Step 2Draft against the template, quotes only from the transcript

Feed the transcript and your template to the Project:

Draft a case study from this transcript using my template (Challenge / What they
tried / Solution / Results). Rules: pull quotes verbatim from the transcript only.
For any metric or number, mark it [CONFIRM WITH CUSTOMER] and do not state a number
as fact unless they said it on the call. Flag anything that needs their sign-off.

The [CONFIRM WITH CUSTOMER] flag is what keeps you honest: the AI cannot round "about half" into "52%" and have it slip through.

Here is the shape of what comes back, illustrative:

Challenge. Before [Customer], onboarding a new analyst took "the better part of a quarter" before they could work a case unsupervised. [CONFIRM WITH CUSTOMER: roughly 3 months?]

Solution. "We started using [Product] to standardize the intake review, and the real change was that new people stopped guessing." (verbatim)

Results. Onboarding dropped to "about three weeks." [CONFIRM WITH CUSTOMER: exact figure?] Each analyst now handles [CONFIRM WITH CUSTOMER: how many?] more cases.

Every quote is lifted word-for-word, and every number wears a flag until the customer signs off on it. The draft is genuinely useful and genuinely unpublishable as-is, which is exactly the state you want: it did the extraction, and it refused to invent the one kind of thing that would get you in trouble.

Step 3Edit for narrative, keep the proof intact

Tighten the draft into a story, but do not touch the quotes or invent results. Your edit is about flow and framing, the AI's job was to extract the proof accurately. Resolve every [CONFIRM WITH CUSTOMER] flag into a clear question for the next step. The narrative move that makes these land is starting in the pain, not the purchase: open on what their world felt like before (the quarter-long onboarding, the analysts guessing), so the result has something to land against. A case study that opens with "Customer chose [Product] because of its robust features" is a brochure. One that opens with the specific, recognizable problem is a story a prospect sees themselves in.

Step 4Send to the customer for sign-off

This is the gate. Send the customer the draft (or just the quotes and numbers) and get explicit written approval. Beyond being the right thing, it protects you legally and relationally, a customer who sees a metric they did not authorize will not be a reference twice. Many customers also tighten their own quotes here, which makes the piece better.

Make the ask trivial to say yes to. Illustrative:

Subject: quick approval, your story

Hi [name], we'd love to feature [Company] as a customer story, draft attached. I only need you to confirm or fix the highlighted bits: two quotes and two numbers (onboarding time and cases per analyst). Reply with edits, or just "approved as-is," and we'll take it from there. Happy to anonymize the company name if you'd prefer.

The highlighted-bits framing matters: you are not asking them to proofread a marketing essay, you are asking them to verify the four things that carry legal and factual weight. That is a two-minute favor, not a project, which is the difference between getting the approval this week and chasing it for a month.

Step 5Publish and reuse

Once approved, publish it, and then reuse it. The same draft feeds your content engine (a quote becomes a post), your sales deck, and your outbound. One call, one approval, many assets. Concretely, one approved study becomes: the full page on your site, a two-line stat for a sales slide, a quote-card for social, a proof point dropped into a relevant outbound email, and a line in your next investor update. You did the capture and the approval once; the distribution is close to free from there.

How you'll know it's working

You actually ship case studies, plural, instead of having a graveyard of "we should write that up" calls. Sales starts pulling them into deals. And because every number was customer-confirmed, you never get the awkward "we didn't agree to that stat" email.

When it breaks

• The AI invented or rounded a metric. This is why Step 2's [CONFIRM WITH CUSTOMER] rule exists and Step 4 is a hard gate. Never publish an unconfirmed number.
• Quotes sound off. It paraphrased instead of pulling verbatim. Restate "verbatim from the transcript only."
• The customer balks at approval. Then you do not publish, full stop. Often they will approve a softened version; offer that, never override a no.
• The transcript has no real substance. The call was too vague. A short, intentional success-story call gives the AI something to work with.
• NDA or naming issues. Some customers cannot be named. Offer an anonymized version ("a $20M logistics company") with their approval.
• The draft reads like an ad, not a story. The AI defaulted to brochure tone. Tell it to lead with the customer's problem in the customer's words and keep your product out of the first third; proof persuades, praise does not.
• You have ten happy calls and no time to write ten studies. Run the extraction on all of them first, then publish the two with the sharpest confirmed numbers. The rest become quote banks for content and sales, which is value even unpublished.

Make it yours. A short, punchy logo-and-stat tile suits a website wall of proof; a long narrative suits a sales deck for a considered, high-ACV deal. Tell the Project which you are making, because the same transcript supports both. For regulated industries (healthcare, finance), add a line to the prompt to flag anything that even resembles a compliance-sensitive claim, so it goes to your reviewer before it goes to the customer.

Where this fits in your harness

This is a Granola pipeline workflow, same capture layer as interview scorecards and pre-meeting briefs. The output feeds your content engine and your outbound. And the same idea, run across many calls instead of one, becomes a product-roadmap signal from your customer conversations.