Reply To: What’s the Best AI Workflow to Turn Raw Notes into a UX Case Study?

Here’s the move: stop “writing” case studies and start assembling evidence. Let AI do the heavy lifting, you supply proof and decisions. Outcome first, details second.

The blocker

Raw notes are inconsistent, quotes are scattered, and metrics don’t align. You spend hours polishing paragraphs that don’t convince a hiring manager in the first 60 seconds.

Why this matters

Case studies are sales assets. Recruiters skim for three things: the problem, what you changed, and the measurable result. No clear KPI, no interview.

What I’ve learned

Speed comes from structure. Build an “Evidence Locker” first, draft second. Use AI to extract, normalize, and outline. You verify, cut fluff, and lead with numbers.

What you’ll need

• Raw artifacts: interviews, usability notes, screenshots, experiment logs, dates.
• An AI editor (GPT‑4‑style), and a basic spreadsheet (your Evidence Locker).
• A simple template: TL;DR, Context & role, Problem & goals, Research insights, Design decisions, Outcome & metrics, Lessons.
• 90 minutes for a complete first pass; 30 minutes to verify/polish.

Workflow that converts notes into a KPI-first case study

1. Create your Evidence Locker (15 min). Columns: Claim, Quote (verbatim), Metric (value + unit + date + source), Artifact link/ID, Confidence (Green/Amber/Red), Owner (you), Notes. Expect to fill 10–20 rows fast.
2. Chunk and label sources (10 min). Break notes into 200–400 word chunks. Labels: Interview_A, Usability_B, Metrics_2024‑08, Screens_App_v3.
3. Extract facts and quotes (10–15 min). Run the Evidence Extractor prompt on each chunk. Paste results into the Locker. Do not edit quotes yet.
4. Normalize metrics (10 min). Use the Metrics Normalizer prompt to convert vague claims into exact numbers with units and dates. Anything fuzzy becomes [UNVERIFIED] or an estimate with a range.
5. Outline outcome-first (5 min). Build the skeleton: TL;DR (2 lines), Problem (1 line), Role (1–2 lines), Top 3 insights (bullets), 3 design decisions (bullets), Outcome (3–5 hard numbers), Lessons (3 bullets). Expect a one‑page outline ready to draft.
6. Draft by section with constraints (20–30 min). Use the Section Draft prompt per section. Keep each section 80–120 words. Start every section with its outcome in bold text (you can style later).
7. Visuals that prove impact (10 min). Run the Visual Brief prompt for three visuals: before/after, flow fix, KPI chart. Capture suggested captions and alt text.
8. Audit like a skeptic (10 min). Run the Skeptic prompt on your draft. It will flag weak claims, unverified quotes, and missing dates. Fix or bracket.
9. Polish voice and scannability (10–15 min). Use the Voice Polish prompt to shorten sentences, surface numbers early, and de‑jargon. Target a 6th–8th grade reading level.
10. Ship and test (10 min). Export to your portfolio or PDF. Do a 60‑second skim test with a colleague: can they state the problem and outcome? If not, tighten TL;DR and Outcome.

Copy‑paste prompts

• Evidence Extractor: “You are a UX research analyst. From the text I paste next, extract: 1) three verbatim user quotes with speaker labels if available, 2) top three pain points (short phrases), 3) any metrics with value + unit + date + source, 4) notable anomalies or surprises. Mark unclear items as [UNVERIFIED]. Keep it concise, bullet format.”
• Metrics Normalizer: “Normalize these outcomes into explicit metrics. For each, return: metric name, value, unit, baseline, comparison period, sample size, method (e.g., A/B, cohort), and confidence (Green/Amber/Red). If data is missing, propose a defendable proxy and mark [ESTIMATE].”
• Section Draft: “Draft the [SECTION NAME] of a UX case study in 80–120 words. Start by stating the outcome in the first sentence. Use the following evidence only: [PASTE relevant rows from Evidence Locker]. Keep tone professional, concise, and free of jargon. Any uncertainty stays in brackets.”
• Visual Brief: “Given this draft and evidence, propose three visuals that prove impact (before/after, flow, KPI trend). For each: title, what to show, why it matters, caption (≤18 words), alt text. Prioritize clarity over aesthetics.”
• Skeptic Audit: “Be a skeptical hiring manager. Scan this draft and list: 1) unverified claims, 2) missing dates or baselines, 3) places where method overwhelms outcome, 4) jargon to remove, 5) opportunities to quantify. Return as actionable bullets.”
• Voice Polish: “Rewrite for clarity and brevity. Keep my voice confident and plain English. Front‑load numbers. Replace passive with active. Max 15–18 words per sentence.”

What to expect

• A complete first draft in 60–90 minutes with 3–5 hard metrics and 2–3 visuals.
• Higher credibility: every claim tied to evidence or bracketed as [UNVERIFIED]/[ESTIMATE].
• A skim‑friendly story: outcome in the first 2 lines, decisions justified by data.

Metrics to track

• Time‑to‑first‑impact (seconds until a reader sees the main KPI) — target < 30s.
• Quote accuracy rate (verbatim, sourced) — target 100%.
• Verified data coverage (% of claims with source/date) — target ≥ 90%.
• Readability (grade level) — target 6–8; keep it skimmable.
• Portfolio conversion (views → interview requests) — baseline, then aim for +25–50% lift.

Common mistakes and fast fixes

• Outcome buried: Put the top KPI in the TL;DR and again in the Outcome section header.
• Soft metrics only: Add absolute numbers, baselines, and dates. If unknown, add [ESTIMATE] with a range and next‑step to verify.
• Over‑index on process: Cap methods to one short paragraph; move details to an appendix.
• Paraphrased quotes: Keep verbatim. If edited for clarity, tag [EDITED].
• Mismatched screenshots: Use before/after with the same viewport and a single highlight callout per image.

1‑week plan to get one polished case study live

1. Day 1: Set up the Evidence Locker. Chunk and label sources. Run Evidence Extractor on two key interviews (45–60 min).
2. Day 2: Normalize metrics from experiments/analytics. Fill baselines and dates (30–45 min).
3. Day 3: Draft TL;DR, Problem, Role using Section Draft prompt (45 min). Verify quotes.
4. Day 4: Draft Research insights and Design decisions (60 min). Tie each decision to one insight.
5. Day 5: Draft Outcome & metrics. Add 3–5 numbers with baselines and periods (45 min).
6. Day 6: Generate Visual Brief, capture or annotate screenshots, finalize captions (60 min).
7. Day 7: Run Skeptic Audit and Voice Polish. Export and run a 60‑second skim test with one colleague. Publish (45–60 min).

Proof over prose. Lead with numbers, back with quotes, show the before/after. Your move.

— Aaron