Prompt Chain: Turn a Research Transcript into a Wireframe Brief

What This Builds

A four-step prompt chain inside a single Claude Project that takes raw interview transcripts and turns them into a wireframe-ready design brief. One project, one thread, four prompts. Instead of separately synthesizing notes, drafting a persona, mapping the journey, and writing a brief, you run the same source material through a fixed sequence and get all four artifacts from one pass.

Prerequisites

• A Claude Project already set up with your research notes or interview transcripts uploaded (covered in the Level 3 design-system assistant guide)
• Pro subscription at $20/month, needed for the longer context window this chain uses
• 3-5 interview transcripts or note sets from a single research round
• Total ongoing cost: $20/month a month, covers this chain plus any other Claude work you already do

The Concept

A prompt chain works like a relay race. Each prompt runs its leg, then hands its output to the next one instead of you starting cold every time. The first prompt reads the raw transcripts and pulls out themes. The second turns those themes into a persona. The third builds a journey map from the same themes. The fourth pulls all three together into a brief a developer or fellow designer could open and start working from. Because every step happens in the same Claude Project thread, later prompts can reference earlier outputs by name, so nothing gets lost in the handoff.

Build It Step by Step

Part 1: Set Up the Project and Load Source Material

1. Open Claude and go to Projects. Use an existing project already loaded with your research materials, or create one named for the current study (for example, "Checkout Flow Research, Q3").
2. Upload the raw transcripts or note files for this round. If you recorded and transcribed sessions, upload the cleaned transcript text, not the raw audio.
3. Add a short project instruction: "You are helping a UX designer synthesize research for [product/feature]. Reference uploaded transcripts directly and flag any claim you cannot trace to a source."

A note on participant data: Interview transcripts often contain a participant's name, employer, or other identifying details they gave you in the room. Before uploading, strip names and any detail that would let someone identify who said what, and check your research consent form covers sharing transcripts with a third-party AI tool. If it does not say so explicitly, keep transcripts out of the project and paste in de-identified notes instead.

What you should see: Your files listed in the project's knowledge panel, and the project ready to accept prompts that can read them.

Part 2: Run the Chain, Prompt by Prompt

Prompt 1, Theme Extraction:

Read all the uploaded interview transcripts for this study. Pull out the 5-8 recurring themes across participants. For each theme, note how many participants mentioned it and include one representative quote per theme. Label this output "THEMES."

Prompt 2, Persona Draft (references Prompt 1's output):

Using the THEMES you just identified, draft one primary persona for this study. Include: name, role, goals, frustrations tied directly to the themes, and a one-sentence "day in the life" summary. Ground every trait in a theme from THEMES, don't invent details the transcripts don't support. Label this output "PERSONA."

Prompt 3, Journey Map (references both prior outputs):

Using THEMES and PERSONA, draft a journey map for the [specific task, e.g. "checkout flow"] with these stages: Awareness, Consideration, Action, Post-Action. For each stage, list the persona's goal, pain points pulled from THEMES, and an opportunity for design improvement. Label this output "JOURNEY."

Prompt 4, Wireframe Brief (pulls everything together):

Using THEMES, PERSONA, and JOURNEY, write a one-page wireframe brief for [feature name]. Include: problem statement, target user (from PERSONA), the 2-3 highest-priority pain points to solve first (from JOURNEY), must-have screen states, and open questions for the team to resolve before wireframing starts. Keep it to one page.

What you should see after each prompt: A distinct, labeled output that the next prompt explicitly builds on. If a later output ignores an earlier one, add a line telling Claude to reference the specific labeled section by name.

Part 3: Test and Refine

Run the chain once on a completed research round you already know well. Compare the PERSONA and JOURNEY outputs against your own read of the transcripts. If Claude invents a pain point no participant raised, add to your project instructions: "Only include claims traceable to a specific transcript quote." Re-run the chain and check again before trusting it on a new study.

Real Example: Onboarding Flow Research

Setup: A UX designer at a mid-size SaaS company uploaded five transcripts from onboarding usability sessions to a Claude Project.

Input: Five 30-minute session transcripts, roughly 25 pages of text total.

Output: THEMES surfaced a repeated confusion point around a setup step users skipped without noticing. PERSONA and JOURNEY built on that theme, and the final brief flagged the skipped step as the top priority with a specific screen-state recommendation to make it unmissable. The designer took the one-page brief straight into a wireframing session that afternoon.

Time saved: Manual synthesis, persona drafting, and journey mapping for a round this size usually takes half a day spread across a week. The chain compresses the first draft of all three artifacts into roughly 20 minutes, leaving the rest of the time for review and refinement instead of a blank page.

What to Do When It Breaks

• Later prompts ignore earlier outputs → Reference the labeled section by name in the next prompt ("using THEMES above") rather than assuming Claude remembers context automatically.
• Persona or journey map invents details → Tighten the project instruction to require a traceable quote for every claim, then re-run just that prompt.
• Output runs long or repeats itself → Add a hard word or bullet limit to the prompt itself instead of editing after the fact.
• Project knowledge gets cluttered across studies → Start a new project per research round rather than piling transcripts from different studies into one. Cross-study contamination is the most common cause of a chain producing a persona that blends two different user groups.

Variations

• Simpler version: Skip the Project setup and paste transcripts directly into a single long chat, running the same four prompts in sequence. Works fine for a one-off study, but you lose the reusable knowledge base for future rounds.
• Extended version: Add a fifth prompt that turns the wireframe brief into a set of Jira or Linear ticket descriptions, one per must-have screen state, so the handoff to engineering is ready to paste in.

What to Do Next

• This week: Run the chain on your current research round and compare the brief against what you would have written manually.
• This month: Build a second chain for competitive teardown to strategy brief, reusing the same relay structure.
• Advanced: Connect the wireframe brief output to a ticketing tool so the final step files tickets automatically instead of you copying them over by hand.

Advanced guide for UX designer professionals. These techniques use more sophisticated AI features that may require paid subscriptions.