Last Tuesday I had a 47-minute customer call with a SaaS founder who has been using our automation workflow for six months. He told me, unprompted, exactly how it changed his weekly cadence, with numbers. Three years ago I would have spent the rest of the day hand-transcribing the call, hand-quoting the best parts, and stitching together a Word doc that hopefully looked like a case study. By the time I finished, the customer’s enthusiasm had cooled and he was slow to approve the final draft.

This time the call ended at 3:47pm. The case study was in the customer’s inbox for review at 4:38pm. Same day. 51 minutes total. The trick is a three-tool stack: Granola for the call, Claude for the draft, Notion for the storage and sharing. Here is exactly how each piece does its job.

The 3-tool stack at a glance

Before the breakdown, here is what each tool actually does in this workflow:

Total saved: a little over two hours per case study. Multiply that across a year of customer wins and the math becomes obvious.

Tool 1: Granola for the call recording and notes

Granola is a meeting notes app that runs quietly in the background on your computer during a call. There is no bot joining the meeting, no awkward “permission to record” moment, no second guest staring at your customer with a weird name like “NotetakerBot — Recording On.” The app just captures the audio off your machine and processes it after the call ends.

For a 47-minute call, Granola gave me three things I actually used:

1. A clean structured transcript with speaker labels (you and the customer)
2. Auto-generated AI notes grouped under headings like “Key challenges,” “What changed after,” and “Specific outcomes” — exactly the structure a case study needs
3. Direct quotes pulled from the transcript, formatted in a sidebar I could lift word-for-word

The auto-notes are not perfect. About once a call, Granola attributes a quote to the wrong speaker, or paraphrases something the customer said in a way that loses the exact phrasing you would want to publish. Always sanity-check before quoting. But as a starting point, it cuts the longest part of the old workflow — typing while listening — down to nothing.

Cost: free tier covers 25 meetings a month. The Pro plan at $14/month removes the limit and unlocks longer transcripts. For one or two case studies a week, free is fine.

Tool 2: Claude for the case study draft

Once Granola finishes, I copy three things into Claude:

1. The full transcript (this is why Claude over ChatGPT — Claude’s longer context window handles a 47-minute transcript without trimming it)
2. Granola’s auto-generated notes
3. Three framing questions: What was the customer’s problem? What did they try first that didn’t work? What specifically changed after they switched to our solution?

The exact prompt I use:

You are writing a customer case study for our blog. Below is a real customer call transcript and AI-generated notes.

Write a case study with this structure:
1. The customer's situation in one paragraph (industry, role, scale)
2. The problem they were trying to solve before us
3. What they tried that didn't work
4. What they changed when they switched to our product
5. The specific results, with numbers where the customer mentioned them
6. A pull-quote at the end (one sentence the customer actually said)

Use only language and details that appear in the transcript. Do not invent results. If a number is missing, say so explicitly. Tone: professional, concrete, no marketing fluff. 600-800 words.

Transcript:
[paste transcript]

AI notes:
[paste Granola notes]

Claude’s first draft is usually 80% there. The 20% I edit is mostly tightening the intro, picking a stronger pull-quote, and fact-checking any numbers against the transcript. The “do not invent results” instruction matters — without it, AI tools love to embellish with plausible-sounding but unsupported numbers.

Cost: Claude Pro at $20/month. The 200K context window is what lets a long transcript fit in one prompt without splitting.

Tool 3: Notion for storage and customer review

I keep every case study draft in a single Notion database called “Customer Case Studies.” Each row has fields for customer name, industry, deal size, status (draft, in review, approved, published), and the draft body itself.

The status field is the trick. When a draft moves from “draft” to “in review,” I share the page with the customer using Notion’s guest link feature. The customer can read it in their browser, leave inline comments, and request edits without ever creating a Notion account. When they approve, I flip the status to “approved” and the draft is queued for the marketing team to format and publish.

Why Notion instead of a Google Doc? Three reasons:

• The database view — I can see at a glance which case studies are stuck in review and which are approved but not yet published
• The customer link — guest sharing is cleaner than emailing PDF drafts back and forth
• Version history — Notion keeps every edit, so if a customer asks “wait, why did this number change,” I can show them

Cost: Notion Free plan handles this for a single user. If your marketing team also lives in Notion, the Plus plan at $10/seat/month is the right tier.

Putting it all together: the 51-minute timeline

Same afternoon, draft delivered. The customer reviewed it the next morning, requested two small changes, and we published the following Tuesday.

What I learned the hard way

Three things I would tell my past self before starting this workflow:

1. Always tell the customer the call is being recorded at the top, even though Granola is silent. Surprise transcripts make customers nervous and slow case-study approvals later. A simple “I am taking notes through a tool that helps me draft this up — totally fine if you’d rather I just take handwritten notes” gets a yes 95% of the time and protects the relationship.
2. Do not let Claude write the headline. AI-generated case-study headlines are uniformly terrible. Write your own headline based on the strongest customer quote.
3. Approve numbers before publishing, every time. Customers are sometimes loose with numbers in conversation, and a published case study with a number they later regret is the fastest way to lose the relationship. Always send the draft for review before it goes live.

The bottom line

This stack is not magic. It does not write a great case study from a mediocre call — if the customer didn’t say anything specific, no AI will rescue it. What it does is take a good call and turn it into a good draft fast enough that you actually publish the case study while the customer is still excited about the result.

For a B2B SaaS, professional services firm, or agency that produces case studies semi-regularly, the three-tool combination of Granola, Claude, and Notion turns a 4-hour task into a 1-hour task. The lift is not in any single tool — it is in the chain. Each tool is doing the part it is best at, and not trying to do the others’ jobs.

If you have been putting off writing up a customer story because the workflow felt heavy, this is the version that finally gets it shipped.

Related reading

• Three AI Tools That Turn Long Meetings Into 5-Minute Written Summaries
• Whisper vs Otter for Transcribing a 1-Hour Podcast Interview
• How to Use ChatGPT and Notion Together to Build a Second Brain from Scratch