The numbers and clients in the examples below are made up to show how a case study is put together. In your own case studies, write only real results, don't round in your favor, and publish them only with the client's permission.
The gist
Nobody hires an architect who just says "I know how to build houses." Good architects show photos of houses they've built, with the address, the cost and what the owner had to say. Your portfolio is the photos of your AI automations. Without it, you're just one more person who "knows how to do AI."
Key concepts
- Case study: a structured write-up of a project: Problem → Solution → Result
- Before/After ROI (return on investment): specific numbers before and after (hours, money, speed)
- Golden number: the one main number that answers "so what?" ($X a month saved, N hours freed up)
- Screenshots vs. descriptions: screenshots convince, descriptions explain; you need both
- 3 types of clients: small businesses, agencies, enterprise; each one wants something different
- Portfolio platforms: Notion, GitHub (a README page that describes a project), a personal website
Theory
Why words don't work and numbers do
Look at two ways to describe the same project:
Version A (words): "I built an email automation system for a real estate brokerage. The system intelligently handles incoming inquiries and responds to them."
Version B (a case study): "A real estate brokerage was getting 80+ inquiries a day and spending 4 hours on manual replies. I built a system that automatically answers the standard questions (65% of inquiries), routes anything urgent to a manager and logs everything in the CRM. The result: a team of 3 saved 18 hours a week, and the time to first response dropped from 4 hours to 3 minutes."
The second version sells. The first one only describes.
Case study structure: Problem → Solution → Result
1. Problem (2–3 sentences)
Who the client is, what hurts, and how it shows up in numbers:
Client: a digital marketing agency, team of 8 Problem: every Monday, about 7 hours went into putting together weekly reports for 12 clients. The data was collected by hand from 4 different sources: Google Analytics, Meta Ads, Google Ads and the CRM.
2. Solution (4–6 sentences)
What was built, with which tools, and how it works:
Built an automated pipeline: every Sunday at 11 p.m. the system connects to the APIs of all four platforms and pulls the week's metrics, Claude writes the narrative part with an analysis of the trends, and the system assembles a PDF and sends it to the clients. The account manager gets a Slack notification with a preview and can make edits or approve sending with one click.
3. Result (specific numbers)
Before: 7 hours a week ≈ 30 hours a month of manual work After: 15 minutes a week to review and approve all 12 reports ≈ 1 hour a month Saved: 29 hours/month = $1,450/month (at a manager's rate of $50/hour) Project cost: $2,800 Payback: under 2 months
The golden number: one number that decides it
A potential client is reading your case study. They give it 10 seconds. They need to see ONE number that answers "why would I need this?"
Examples of golden numbers:
- $3,200/month saved: for clients who think about money
- 22 hours freed up a week: for clients who think about time
- Response time: from 4 hours to 3 minutes: for clients who think about quality of service
- 87% of inquiries handled without a person: for clients who think about operational efficiency
The golden number belongs in the first paragraph of the case study and in the headline.
Case study headline template:
[Type of business] saves [Golden Number] with AI automation of [process] Examples: "Real estate brokerage saves 18 hours a week by automating incoming inquiries" "Digital agency cuts reporting time from 30 hours to 1 hour a month" "B2B company speeds up lead response from 4 hours to 8 minutes"
Screenshots vs. descriptions
A description explains the logic. A screenshot proves it's real.
What to screenshot:
- A chatbot in action, on a website or in Slack (showing real messages, with personal details blurred)
- A Google Sheet with a log of rows that were filled in automatically
- A finished PDF report the system generated (with client names blurred)
- A dashboard with metrics (if you built one)
- The workflow as a diagram (Claude Code can draw one in Mermaid, a text-based diagram format)
What to blur out:
- Real client names → replace them with "Client A," "Client B"
- Email addresses, phone numbers, personal details
- The client's financial data (show only relative numbers)
The rule of three screenshots: At least three screenshots per case study: before (the manual process or the empty system), during (the system at work), after (the result in the data or in the report).
3 types of clients and what matters to each
Small business (1–10 people):
- Values: the owner's time saved, simplicity, reliability
- Worries: too complicated, it'll break, I won't be able to figure it out
- Golden number: "X hours a week you can spend growing the business instead of on routine work"
- Case study format: short, conversational, with a specific scenario: "a day without the automation vs. a day with it"
Agencies (10–50 people):
- Values: scalability, integration with the tools they already use, no need to hire someone new
- Worries: depending on one developer, no documentation
- Golden number: "Capacity for X clients without hiring another manager"
- Case study format: technical details, integrations, SLA (a service-level agreement: the response times you commit to), handoff documentation
Enterprise (50+ people):
- Values: security, compliance, ROI the CFO can sign off on, scale
- Worries: security risk, vendor lock-in, no support
- Golden number: "Annual ROI of $X on an investment of $Y"
- Case study format: formal, with an ROI calculation, mentioning security and documentation
Where to keep your portfolio
Option 1: Notion (recommended to start)
Simple, good-looking, and you can publish it as a link. Structure:
Notion page your-portfolio
├── About (2 paragraphs + LinkedIn)
├── Services (what you offer)
├── Case Studies
│ ├── Case 1: Real estate brokerage, 18 hours/week
│ ├── Case 2: Digital agency, reports in 15 minutes a week
│ └── Case 3: B2B company, leads answered in 8 minutes
└── Contact (Calendly link)Option 2: A GitHub README
If your clients are technical: a GitHub repo with a well-formatted README. It shows your code, your architecture and your professionalism.
Option 3: A personal website
Makes sense once you have 3+ case studies and a steady flow of clients. Until then, Notion is faster and simpler.
Case study template (ready to use)
## [Headline with your Golden Number] **Client:** [Type of business, size, no name] **Industry:** [Real estate / Marketing / E-commerce / ...] **Timeline:** [X weeks from kickoff to launch] --- ### Problem [Client] was spending [X hours/period] on [specific process]. Pain point: [what exactly was wrong: slow, expensive, unreliable, tedious]. Cost of the problem: [X hours × $Y/hour = $Z/month lost]. ### Solution Built [type of automation] that [what it does, step by step]. Stack: [tools: Claude Code, Cloudflare Workers, Slack bot, ...]. Key decision: [one interesting technical detail]. ### Result | Metric | Before | After | |---|---|---| | Time spent on the task | X hours/week | Y minutes/week | | Cost of the routine work | $X/month | $Y/month | | Response time | X hours | Y minutes | **Project cost:** $X **Payback:** X months **Client testimonial:** "[Quote: only a real one, used with the client's permission]" [Screenshot 1: before] [Screenshot 2: the system at work] [Screenshot 3: the result]
Practice
Exercise: write your first case study
Pick a task you've already handled with AI: for yourself, at work or for someone you know (for example, from the "AI for everyday work" module: emails, meetings, presentations). If you've already built automations from the library lessons, use one of those
Fill in the Problem → Solution → Result structure:
Type this into the chat Problem: what task does the system take care of? Hours before: ___ a week Hours after: ___ a week Saved: ___ hours × $__ = $__ a month
Come up with your golden number: one number for the headline
Take 3 screenshots: before (how it was done by hand), during (the system or AI at work), after (the result: a log, a report, a spreadsheet)
Write the case study using the template, no longer than one letter-size page
Create a Notion page and publish it: click Share at the top → open the Publish tab → click Publish
Goal: a finished, published case study with specific numbers and screenshots. It's the first brick in your portfolio.
Case study template (copy and paste)
### Case study template (Problem → Solution → Result)
**Client:** [type of business, team size, NOT the name]
**Problem:** [the specific pain point WITH NUMBERS: hours, money, errors]
**Solution:** [what was built, which tools, how it works]
**Result:** [specific metrics: time before/after, money before/after, %]
**Stack:** Claude Code + [list your tools]
**Payback:** [X months]
**Testimonial:** "[Client quote, if you have one]"A filled-in example (made-up numbers):
**Client:** digital agency, 8 people
**Problem:** 30 hours/month on manual reports for 12 clients, 2-3 errors a month
**Solution:** Automatic data collection from GA + Meta Ads + Google Ads,
Claude writes the narrative part, the PDF is assembled and sent on a schedule
**Result:** 30 hours → 1 hour/month, 0 errors, reports arrive Monday morning
**Stack:** Claude Code + Cloudflare Workers + Slack bot + Google Sheets
**Payback:** 2 months
**Testimonial:** "On Mondays our marketer does real work instead of copying numbers"Common mistakes
- No before/after numbers. "I automated reports" is a description. "30 hours → 1 hour a month, $1,450/month saved" is a case study. Without before/after metrics, a case study doesn't sell.
- Writing a "case study" with no screenshots. Words describe, screenshots prove. At least 3 screenshots: BEFORE (the manual process), DURING (the system at work), AFTER (the result in the data).
- Showing the technical side instead of the business result. The client doesn't care that you used Cloudflare Workers. They care that they save $1,450 a month. Put the tech stack in small print at the bottom.
- Not asking for a testimonial right after delivery. The best moment to ask is when the client sees the working system for the first time. A month later, the excitement will have cooled off. (→ Delivery and retention: more on collecting testimonials)
How this connects to other lessons
- → Delivery and retention: coming up later in this module: how to get a testimonial and a case study out of every project you deliver
- → Packaging: already covered: case studies are the proof of value behind your packages
- → Real monetization case studies: already covered: worked examples with numbers, for inspiration
Tools and resources
- Notion: notion.com. Has a free plan (as of October 2026), good-looking public pages and portfolio templates
- CleanShot X (Mac): professional screenshots with annotations and blurring
- Loom: loom.com. A short video (2–3 min) showing the system at work is more convincing than a screenshot
- Carbon: carbon.now.sh. Good-looking screenshots of code, if you want to show the technical side
- Tally: tally.so. Free forms for collecting feedback and testimonials from clients
- Typeform: typeform.com. Polished forms for intake questionnaires (an alternative to Tally)
Key takeaways
Words describe; numbers sell. Every case study needs specific numbers: hours, money, speed.
The golden number is the one main number that answers "so what?" It belongs in the headline.
At least three screenshots. Visual proof clears up doubts that text alone can't.
A portfolio of three case studies matters more than a personal website. Case studies first, presentation later.
Next lesson
→ Cold email and LinkedIn outreach with Claude
You've already seen how to bundle your services into packages in the lesson Packaging.
The mark stays in this browser only and is never sent anywhere. My progress