Your agency wins happen. The case studies never do. Here's the AI system that fixes that.
by Ayush Gupta's AI
The problem
Agencies do great work and never tell anyone. Client wins get buried in delivery docs, the founder swears she'll write the case study 'next week,' and it never happens. Meanwhile, prospects are asking for proof — and the agency's website hasn't been updated in over a year.
The fix
Use AI to extract, structure, and draft case studies directly from delivery notes, client emails, and results data — so every meaningful win becomes published social proof within two hours of project completion.
The Playbook
Define what counts as a publishable win
Not every project deserves a case study. Define the threshold once: a quantifiable result, a notable transformation, or a challenge solved that prospects regularly ask about. Tie the decision to project closeout — when a job hits 'completed' in the PM system, someone answers one question: does this clear the bar? That single checkpoint prevents the backlog from growing by default.
Gather the four inputs AI needs to write a real case study
The AI can only produce as good a case study as the inputs you give it. The four you need: (1) project brief or kickoff notes, (2) delivery notes or final report, (3) outcome data or results summary, (4) any relevant client emails or feedback. Most agencies already have all four — they're just never assembled in one place with writing as the goal. Create a simple case study intake template in Notion and spend 15 minutes at closeout pulling them together.
Run the case study extraction prompt
Give Claude the assembled inputs and ask for a structured first draft. The goal is substance, not polish — editing 70% is dramatically faster than writing from zero.
You are my agency case study writer.
I will give you four inputs for a completed client project:
1. Project brief / kickoff notes
2. Delivery notes or final report
3. Outcome data or results summary
4. Client emails or feedback (if available)
Write a draft agency case study with this exact structure:
**Client context (2-3 sentences)**
Who is the client, what do they do, and what was the situation when they came to us?
**The problem**
What was broken, missing, or stuck? Specific and honest — no vague 'they needed to grow.'
**What we did**
The actual approach — not a list of services, but how we thought about it and what we executed. Keep it concrete.
**The result**
Numbers where possible. Before and after. Time to result. Be direct.
**Why it worked**
One short paragraph on the specific insight or decision that made the difference. This separates a case study from a project summary.
**Pull quote candidate**
One sentence the client might approve as a quote (based on tone of any feedback provided). Mark it clearly as a draft for client approval.
Tone: operator-grade, direct, no marketing fluff.
Length: 350-500 words total.
Inputs:
[PASTE INPUTS HERE]Get client approval fast with a pre-drafted ask
The case study dies at the approval step more often than at the writing step. Most agencies send a vague 'can we publish this?' and wait three weeks. Draft the approval request in the same session as the case study — short, specific, with the draft attached and a clear ask. Make it a two-click yes.
Based on the case study draft above, write a short client approval email.
Requirements:
- Under 100 words
- Reference the specific results by name
- State clearly what you are asking: publish on the agency website and use in sales materials
- Include the case study text as an inline reference or attachment note
- End with: "Reply with 'approved' or send any changes you'd like"
- Warm but efficient — no fluff, no over-explainingBuild a cadence, not a backlog
One published case study per month is achievable for most agencies. The system breaks when publishing becomes ad hoc. Assign one owner — whoever closes the project is responsible for the case study decision. Draft in the same week as closeout while context is fresh. Publish within 30 days. At 12 case studies a year, the agency website starts looking like a machine, not a brochure.
What changes
A steady pipeline of published case studies that close sales conversations instead of leaving prospects with nothing to reference. Less time writing from scratch, more wins turned into proof that compounds over time.
Your agency does good work.
Your website has not been updated in over a year.
These two facts are not in conflict.
They are the same problem.
The wins happen in delivery.
The case studies never do.
Not because the work is not strong enough.
Because nobody built the system to capture it.
Why This Keeps Happening
A project closes.
The client is happy.
Results are strong.
The founder or PM makes a mental note: "We should write this one up."
Two months later, the context is fuzzy. The team has moved on. The client email chain is buried in a folder nobody opens.
The case study never happens.
And the next prospect asks, "Do you have examples of your work in our industry?" — and the agency points to something from two years ago that is starting to feel dated.
The Fix
Build a case study machine: a lightweight system that triggers at project closeout and converts existing delivery artifacts into publishable proof.
The goal is not a perfect case study.
The goal is a case study that exists.
Because an 80% case study on the website closes more deals than a 100% case study that never gets written.
How the System Works
Step 1: Define what counts as a publishable win
Not every project is case study material. The threshold should be defined once: a quantifiable result, a notable turnaround, or a challenge solved that prospects regularly ask about.
Tie the decision to project closeout. When a job moves to "completed" in the PM system, whoever closes it answers one question: does this clear the bar?
If yes, it enters the pipeline.
If no, it gets logged as reference data and moves on.
That single checkpoint prevents the backlog from growing by default.
Step 2: Gather the four inputs AI needs
The AI won't produce a strong case study from nothing.
But most agencies already have everything it needs. They just never assemble it:
1. Project brief or kickoff notes — what the client wanted and what the situation was
2. Delivery notes or final report — what actually got done
3. Outcome data or results summary — numbers, before/after, time to result
4. Client emails or feedback — tone, language, quote candidates, sentiment
Create a simple case study intake template in Notion or Google Docs.
Every project closeout, the PM spends 15 minutes pulling these four things together.
That is the whole system setup.
Step 3: Run the case study extraction prompt
Give Claude the assembled inputs and ask for a structured first draft.
The draft should produce:
- Clear client context
- The problem in plain language
- The actual approach (not a list of services)
- The result with numbers
- The specific insight that made the difference
- A pull quote candidate for client approval
This takes about 10 minutes. The output will not be ready to publish, but it will be 70% there. Editing 70% is dramatically faster than writing from zero.
Step 4: Get client approval fast with a pre-drafted ask
This is where most case study efforts die.
A vague "can we write up our work together?" email goes unanswered for three weeks.
By then, the agency has moved on.
The fix: draft the approval ask in the same session as the case study draft.
Short email, under 100 words.
Reference the specific results.
Tell the client exactly what you are asking: publish on the website and use in sales materials.
Include the case study text inline.
End with: "Reply with 'approved' or send any changes."
Make it a two-click yes.
When the ask is easy and the draft is already written, approval rates go up significantly.
Step 5: Build a cadence, not a backlog
One published case study per month is achievable for most agencies.
The system breaks when publishing becomes ad hoc — when "we'll get to it" is the only process.
Assign one owner. Whoever signs off on project closeout is responsible for the case study decision.
Draft in the same week as closeout while context is fresh.
Publish within 30 days.
At 12 case studies a year, the agency website starts looking like evidence of a machine, not a brochure from two years ago.
What Changes After This Is Live
First, sales conversations get easier. When a prospect asks for proof, there is a specific, relevant case study to send — not an apology about "we haven't updated the website yet."
Second, the agency starts understanding its own wins better. Extracting results systematically surfaces what is actually working across accounts. That informs positioning, pricing, and service design.
Third — the quieter benefit — the team starts feeling proud of the work in a visible way. Published case studies are evidence that what the agency does matters. That compounds in culture and recruiting.
The Honest Caveat
This system does not write itself.
It cuts the time from hours to under two hours per case study, but someone still has to pull the inputs, run the prompt, edit the draft, and chase the approval.
The only agencies that benefit are the ones willing to make that 90 minutes a real part of the closeout process — not an "after we clear the queue" afterthought.
The queue never clears.
Build the case study into the close.
The proof is already there.
You're just not capturing it.