← AIAP Tool GuidesAll programsHomeSearch
AIAP Tool Guides·AI for Design·41:09

AI for Designers: Product Research - Personas, Jobs to Be Done, Competitors and a Locked Value Proposition in Claude and ChatGPT

Amlan Designer, "a fellow builder and designer" with entrepreneurial stints; builds one expense-management app (FinLit India) across the series
▶LMS lesson
← PreviousGoogle Search Ads Fundamentals: Ad Rank and Quality Score, Extensions, Account Structure,
Lesson 1 of 5 in AI for Design
Next →Research Synthesis to First Build: a PRD by Prompt-for-a-Prompt, Mood Boards in ChatGPT an

The short version

  1. Framing: AI has collapsed idea-to-wireframe and research synthesis from days to minutes, so designers are judged on design thinking, not screens; a five-step build-along follows (0:00-0:05).
  2. One structured research prompt (context + three personas, three jobs to be done, three failure modes, one non-obvious insight) run in Claude and ChatGPT side by side for an expense app for families and students (0:05-0:16).
  3. Competitor teardown of five apps, three one-sentence value propositions, then the chosen one is locked and the whole study exported as a research document for later tools (0:16-0:40).

At a glance, three clicks deep

Skim here first: the closed row is the glance, open is the study card with the key points and timestamps, and the ↓ link drops to that concept's full write-up below.

01A fixed four-part research prompt, run in two modelsGive context (what, for whom, why), then always ask for three personas, three jobs to be done, three failur…›

Give context (what, for whom, why), then always ask for three personas, three jobs to be done, three failure modes and one insight most designers would miss - short answers, same prompt in two LLMs.

Prompt (0:05-0:07)

ChatGPT results (0:08-0:11, 0:13-0:16)

Claude insight (0:12-0:13)

↓ Full write-up of this concept

02Competitor teardown: one strength, one weakness, one missed opportunityIn the same chat, ask for the top five competitors with one thing each does well, one it does badly and one…›

In the same chat, ask for the top five competitors with one thing each does well, one it does badly and one opportunity it missed.

Prompt (0:16)

Five apps (0:18-0:23)

↓ Full write-up of this concept

03Stress-test three value propositions, lock one, export the researchAsk for three distinct one-sentence propositions (no hypotheticals), pick one, tell the model it is locked,…›

Ask for three distinct one-sentence propositions (no hypotheticals), pick one, tell the model it is locked, then compile everything into one formatted document.

Propositions (0:24-0:27)

Document generation and troubleshooting (0:27-0:39)

↓ Full write-up of this concept

The concepts in full

01

A fixed four-part research prompt, run in two models

how-to

Claude's insight: "The real competitor isn't another expense app. It's doing nothing."

Only the product-and-audience sentence changes between projects ("an expense management app for families and students who want to optimize their spend"). ChatGPT returned personas like Priya (family planner), Rohan (student who runs out by the 20th) and Sunita (semi-retired), and the insight that users want "permission to spend without guilt"; Claude stressed capturing expenses within the first 48 hours of onboarding. Merge the best of both; the trainer then sticks with Claude as "really robust".

Do it in this order
Why it matters

A repeatable research starter that keeps outputs comparable.

This idea elsewhereparallelCatalyst Session 5: the AI research chain for an ideal customer profile
02

Competitor teardown: one strength, one weakness, one missed opportunity

Do one niche function really well - that is the MVP lesson from five rivals.

Because it runs in the same conversation, earlier insights steer which apps are chosen. Claude covered YNAB (zero-based budgeting, about $109 a year, steep learning curve), PocketGuard (bank connection only, no manual entry), Monarch Money (collaborative, about $100 a year, no free tier, no approval flow before shared spend), Expensify (receipt scanning, weak on who-owes-what) and Rocket Money (subscription detection).

Why it matters

A fast map of where the gap is.

03

Stress-test three value propositions, lock one, export the research

how-to

The document is both an approval artefact and context for every later tool.

The chosen line promised one glance to know you are on track - "No setup, no spreadsheets, no guilt" - and a view of "what actually is costing your future self." Validate with real users (Maze is mentioned). Generating the docx took three tries: refresh if it stalls, start a new chat if it loops; Claude writes JavaScript to build the file.

Do it in this order
Why it matters

Turns a chat into a reusable artefact.

Tools referenced

ToolCoverageMomentContext
Claudedemonstrated
ChatGPTdemonstrated
Google Docsdemonstrated
Mazementioned
Figmamentioned
Figma Weavementioned
Claude Codementioned
Cursormentioned
v0mentioned
GitHub Copilotmentioned

Action items

    Resources mentioned

    Resources
    • docOutskill instructional handout with all the prompts

    Extraction notes

    This page was built from an auto-generated transcript, which garbles product and people's names. Those were corrected silently in everything above and logged here for transparency. The warnings flag claims that were true on the recording day but change fast.

    Transcript corrections applied

    The transcript saysThe trainer actually means
    Cloud / Plot / GodClaude
    Cloud CodeClaude Code
    VEV / ViviFigma Weave
    lock in my USBlock in my USP

    True on recording day — verify before relying

    ← PreviousGoogle Search Ads Fundamentals: Ad Rank and Quality Score, Extensions, Account Structure, Next →Research Synthesis to First Build: a PRD by Prompt-for-a-Prompt, Mood Boards in ChatGPT an