Claude Code · PM & Founder Playbook

Real workflows.
Real results.
Zero code.

How product managers and founders are using Claude Code to do in 20 minutes what used to take 2 days — with exact step-by-step prompts.

50%
Productivity boost reported
20min
10-competitor analysis
45min
PRD → working prototype
0
Lines of code needed
Where to start
6 workflows PMs & founders
use every week
Click any card to see the full step-by-step walkthrough with exact prompts you can run today.
🔍
Competitive Analysis
Analyze 10 competitors in parallel — same depth, same framework, 20 minutes instead of a full day.
⏱ 20 min · was: full day
📋
PRD → Working Prototype
Write a spec in plain English, get a working prototype back in a single session. Dennis Yang does this daily at Chime.
⏱ 45 min · was: 2-week sprint wait
🎙️
Customer Research at Scale
Run 10 parallel agents across 10 interview transcripts simultaneously. Pain points, JTBD patterns, feature requests — extracted in minutes.
⏱ 5 min · was: 50 min
📊
Data & Bug Analysis
Upload a CSV of user feedback or bug reports. Get a severity/frequency matrix, prioritization table, and visual dashboard.
⏱ Under 1 min · was: afternoon
🚀
SEO Content Strategy
Solo founder use case: crawl your site, research competitors, rank keyword gaps, build a prioritized content plan — all real data.
⏱ 20 min · was: hired agency
📝
Docs & Release Notes
Turn meeting notes → agendas → release notes → stakeholder updates automatically. Slash the documentation tax.
⏱ 2 min · was: 1–2 hours

"The real unlock for product managers isn't in simply using AI to answer your questions. It's in building AI-powered systems and workflows that actually automate the work we do every day."

— Sachin Rekhi, founder of Notejoy & PM at LinkedIn, Microsoft
Use Case 01
Competitive Analysis
The #1 workflow PMs and founders cite. Analyze 10 competitors in parallel with the same framework applied to each — something that's impossible to do by hand without losing depth.
Without Claude Code
1 full day
Quality drops after competitor #4 from fatigue
→
With Claude Code
20 min
Same depth on competitor #10 as #1
🗂️
Step-by-step · ~20 minutes
Live Competitor Pricing Matrix
Use Claude's browser agent to visit competitor pricing pages in real time — not stale search results. Build an always-accurate pricing table.
🔥 Most requested 20 min Founders · PMs
⌄
1
Set up your folder
Create your Competitive Intel folder
Make a project folder with two files: one for your product, one listing competitors. Claude reads these every session automatically.
competitive-intel/
product-info.md ← your product
competitors.md ← list of targets
CLAUDE.md ← auto-loaded
2
Populate product-info.md
Tell Claude about your product
Plain English description of what your product does, your pricing tiers, your differentiators, and your target customer.
product-info.md ## Our Product: [Name]
What we do: [2 sentences]
Target customer: [who]
Pricing: [your tiers]
Key differentiators: [list]
3
List competitors
Populate competitors.md
List competitor names and their URLs. Claude will navigate to each one's pricing page live — not rely on outdated cached data.
competitors.md - Notion: https://notion.so/pricing
- Coda: https://coda.io/pricing
- Confluence: https://atlassian.com/...
- Obsidian: https://obsidian.md/pricing
[add as many as you want]
4
Run the command
Give Claude the instruction
One prompt kicks off a parallel analysis. Claude visits every URL live, extracts pricing, and compiles everything side-by-side.
Type into Claude Code Read product-info.md and competitors.md.
Browse each competitor's pricing page live.
For each one extract: plan names, prices,
key features per tier, free trial info.
Generate a comparison table and an
executive summary with strategic notes
about where we win and where we lose.
5
Get your output
What Claude delivers
  • A full side-by-side pricing matrix (Markdown table, copy straight into Notion/Slides)
  • Strategic commentary — "Your free plan is the most generous; Evernote's is most restrictive"
  • Feature gap analysis — what competitors have that you don't, and vice versa
  • Files saved locally so next month you just re-run the command to refresh

    • "In the past, anytime I asked Deep Research or ChatGPT to get competitor pricing, the results were woefully out of date. The only way to get accurate pricing data is to have the AI actually navigate to each pricing page in real time and extract the information directly."

    — Sachin Rekhi, Founder of Notejoy
    🔬
    Advanced · Reusable Skill
    Build a Repeatable Competitive Intelligence Skill
    Set it up once. Then run "Profile Notion" or "Compare Notion vs Coda" any time — same framework, no re-explaining.
    15 min setup 1-click after setup
    ⌄
    1
    Describe your process
    Tell Claude how you analyze competitors
    Have a conversation about what dimensions matter to you: positioning, pricing, target audience, content strategy, growth signals.
    Start the conversation I want to profile competitors. The
    dimensions I care about are: positioning,
    target audience, pricing model, top features,
    content strategy, and growth signals.
    Help me design a repeatable process.
    2
    Lock in the format
    Define your output template
    Tell Claude exactly what the final doc should look like — headings, sections, filename format. The more specific, the more consistent.
    In the same conversation Save each profile as:
    [company-slug]-[YYYY-MM-DD].md
    Include: Executive summary, pricing table,
    feature matrix, strategic notes, and
    a section on how we compare.
    3
    Create 3 skills
    Ask Claude to write the skills for you
    Claude writes its own instruction files. Ask for three: single profile, side-by-side compare, and track changes over time.
    Final prompt Based on our conversation, create
    three SKILL.md files:
    1. evaluate-company (single profile)
    2. compare-companies (side-by-side)
    3. trend-company (track changes over time)
    4
    Use it forever
    One-word trigger from now on
    Every future analysis uses the same framework. No drift from fatigue. When a new competitor appears, one command profiles them instantly.
    All future sessions — just say: Profile Notion

    Compare Notion vs Coda

    Trend Notion (compare to last month)
    What this system gives you
    • First setup: ~15 minutes of conversation
    • Every future analysis: under 1 minute to trigger, runs autonomously
    • All profiles saved as dated Markdown files — a historical record of how competitors evolve
    • Zero fatigue — #10 gets the same rigor as #1
    Use Case 02
    PRD → Working Prototype
    The workflow Dennis Yang, a PM at Chime, runs daily: write a spec in plain English markdown, open Claude Code, and have a working prototype in a single session. No sprint wait. No back-and-forth. Something real your team can react to.
    Without Claude Code
    2 weeks
    Write spec → sprint planning → wait → first build
    →
    With Claude Code
    45 min
    Write spec → run Claude Code → share screen recording in Slack
    📋
    The Dennis Yang workflow · Chime
    Spec to Prototype in One Session
    Walk through the exact session flow used by PMs at Stripe, Linear, and Chime to turn written specs into interactive prototypes.
    Used at Stripe · Linear · Chime 45 min
    ⌄
    1
    Write your PRD in Markdown
    Plain English spec file
    No special format needed. Just describe what the feature does, who it's for, and the key user flows. Save as a .md file in your project folder.
    feature-spec.md ## Feature: Onboarding Checklist
    Users: New signups (first 7 days)
    Goal: Get to first value moment faster

    User flow:
    1. See checklist on dashboard
    2. Check off steps as completed
    3. Show progress bar + celebrate at 100%

    Success metric: Day-7 retention +10%
    2
    Use Plan Mode first
    Let Claude plan before it builds
    Always use Plan Mode for feature builds. Claude shows exactly what files it will create, what logic it will write. You approve before it touches anything.
    Open Claude Code and type: Read feature-spec.md and plan how
    you'd build this as a working HTML
    prototype. Show me what files you'll
    create and what each will contain.
    Don't write any code yet.
    3
    Approve and build
    Execute the plan
    Once you're happy with the plan, one word kicks off the build. Claude creates all the files, wires up the logic, and confirms when done.
    After reviewing the plan: Looks good. Build it.

    [Claude creates files, wires logic]

    ✓ Created: onboarding-checklist.html
    ✓ Done. Open in browser to preview.
    4
    Iterate in plain English
    Refine until it's right
    Every change is a sentence. No Figma handoffs, no Jira tickets, no waiting. Your team reacts to something real — not a static mock.
    Keep going: Add a confetti animation when all
    steps are checked off.

    Make the progress bar green when
    above 50%, amber below.

    Add a "skip for now" link on each step.
    5
    Share immediately
    How to get team feedback fast
  • Record a 90-second Loom of the prototype — share in Slack before the stand-up
  • Team reacts to something real, not a description — feedback is 10x more specific
  • Engineers see the interaction model — handoff is a working reference, not a Figma link
  • The prototype file is in your repo — engineers can build directly on top of it

    • "Dennis Yang writes a PRD in markdown, opens a terminal, and types 'claude'. Twenty minutes later, the PRD is a running prototype. He shares a screen recording in Slack. The team is already discussing refinements based on something real."

    — Builder.io, Claude Code for Product Managers
    Use Case 03
    Customer Research at Scale
    Process 10 interview transcripts simultaneously with parallel agents. Extract pain points, JTBD patterns, feature requests, and sentiment — then synthesize everything into a single research brief.
    Without Claude Code
    50 min
    Process one at a time, re-explain format each time
    →
    With Claude Code (parallel)
    5 min
    10 agents run simultaneously — same framework on all
    🎙️
    Parallel agents · 10x speed
    10 Interviews → Research Brief in 5 Minutes
    Launch 10 sub-agents simultaneously. Each processes one interview with the same extraction template. One agent synthesizes all results.
    5 min PMs · Researchers · Founders
    ⌄
    1
    Organize your transcripts
    Drop all transcripts in one folder
    Copy your interview transcripts as plain text files (.txt or .md) into a folder. Claude can also read screenshots of interview notes.
    interviews/
    interview-01-sarah.md
    interview-02-marcus.md
    interview-03-priya.md
    ... (10 total)
    2
    Launch parallel analysis
    One prompt, 10 simultaneous agents
    Claude spins up sub-agents — one per interview. All run simultaneously with the same extraction template. Done in the time it used to take to do one.
    Type into Claude Code: Read every .md file in /interviews.
    For each one, extract:
    - Top 3 pain points (with severity 1-5)
    - Feature requests mentioned
    - Jobs-to-be-done patterns
    - Key verbatim quotes
    - Competitor mentions
    Save each analysis as [name]-analysis.md
    3
    Synthesize all findings
    Build the research brief
    After all 10 analyses are complete, one more command synthesizes everything into a single research brief with ranked themes and quotes.
    Then run this: Read all *-analysis.md files.
    Synthesize into a research brief with:
    - Top 5 pain points ranked by frequency
    - Most requested features
    - Representative quotes per theme
    - Recommended next steps for the team
    4
    JTBD deep dive (bonus)
    Run a Jobs-to-be-Done analysis
    Claude Code has a built-in JTBD skill that applies Christensen's full methodology — including Big Hire vs. Little Hire, forces of progress, and non-obvious competition.
    Advanced prompt: Using the jobs-to-be-done framework,
    analyze the themes across all interviews.
    Identify the functional, emotional, and
    social dimensions of each job.
    What are customers really hiring us for?
    What Claude delivers
    • 10 individual analysis files — one per participant, same structure
    • Master research brief with frequency-ranked pain points and feature requests
    • JTBD patterns — what customers are really trying to accomplish
    • Verbatim quotes organized by theme — copy straight into your PRD or investor update
    Use Case 04
    Data & Metrics Analysis
    Stop waiting for the data science team. Upload a CSV and get severity matrices, funnel analysis, feature impact estimates, and visual dashboards — in under a minute.
    Without Claude Code
    1 afternoon
    Pivot tables, VLOOKUP, waiting for data team
    →
    With Claude Code
    Under 1 min
    Plain English → matrix, chart, and recommendations
    📊
    Instant analysis · no data team needed
    Bug Severity / Feature Priority Matrix from CSV
    Export a spreadsheet from Jira, Linear, or Notion. Claude builds a severity/frequency matrix and tells you exactly what to fix first.
    Under 1 min PMs · Ops
    ⌄
    1
    Export your data
    Get a CSV from any tool
    Export bug reports from Jira/Linear, user feedback from Intercom, NPS data from Delighted, or a feature request list from wherever you track it.
    Any of these work: bugs.csv (from Jira export)
    feedback.csv (from Intercom/Zendesk)
    nps-responses.csv (from Delighted)
    feature-requests.csv (from Notion)
    2
    Drop it in your folder
    Move the CSV into your project folder
    No upload needed. Just move the file into the folder where you run Claude Code. Claude reads your local filesystem directly.
    product-data/
    bugs.csv ← just drop it here
    CLAUDE.md
    3
    Run the analysis
    One prompt, full analysis
    Claude writes Python to process the data, executes it, and gives you a structured output. You don't need to know any Python.
    The exact prompt to run: Analyze bugs.csv and build a
    severity/frequency matrix for the
    top 20 bugs. Categorize by:
    - Severity (user impact 1-5)
    - Frequency (how often reported)
    Then recommend the top 5 to fix first
    with reasons. Show as a table.
    4
    Build a visual dashboard
    Turn the data into a stakeholder-ready chart
    One follow-up prompt turns the table into an interactive HTML dashboard you can share directly with stakeholders.
    Follow-up prompt: Turn this analysis into an interactive
    HTML dashboard with a scatter plot
    (severity vs frequency), filterable
    by category, with a summary stats bar
    at the top. Make it shareable.
    What Claude delivers
    • Severity × frequency matrix — prioritized exactly like a PM would do it manually
    • Top 5 recommendation with reasoning — copy straight into your sprint planning doc
    • Interactive dashboard HTML — share the file or open it in a browser for stakeholder review
    • Saved analysis file — re-run next month on updated data in seconds
    Use Case 05 · Founder
    SEO Content Strategy
    The solo founder use case: crawl your own site, analyze competitors' content, research keyword gaps, and get a prioritized list of pages to build — all backed by real search data. No agency, no subscription tool.
    Without Claude Code
    Hire agency
    $2-5K/month or weeks of manual research
    →
    With Claude Code
    20 min
    Real keyword data, competitor gaps, prioritized plan
    🚀
    Solo founder · real data
    Full SEO Strategy in 20 Minutes
    The exact prompt a solo founder used to get Claude Code to crawl their site, research competitors, and return a prioritized content roadmap.
    20 min Real search data Founders · Marketers
    ⌄
    1
    Connect SerpAPI (free tier)
    Give Claude real keyword data
    SerpAPI gives Claude live Google search results. Free tier is enough to run this workflow. Add it to your settings.json as an MCP server.
    .claude/settings.json "mcpServers": {"{"}
    "serpapi": {"{"}
    "command": "npx",
    "args": ["serpapi-mcp"],
    "env": {"{"}"SERPAPI_KEY": "your-key"{"}"}
    {"}"}
    {"}"}
    2
    The master prompt
    One prompt does everything
    This is the exact prompt a solo founder used to get a full SEO plan in 20 minutes. Claude crawls, researches, compares, and prioritizes.
    The exact prompt to run: Help me create an SEO content strategy
    for [YOUR URL]. First, crawl my site
    and understand topics I've covered.
    Then research keywords in my niche
    using SerpAPI — what's ranking, what
    questions people ask, what competitors
    cover. Compare my content vs. gaps.
    Give me a prioritized list: page to create,
    keyword, search volume, ranking chance.
    3
    Review the plan
    Claude returns a prioritized roadmap
    Claude delivers a structured content plan with real search volumes, competitor gap analysis, and a rationale for why each page is worth writing.
    Sample output structure: 1. "How to [X]" — 8,400/mo searches
    Competitor coverage: weak
    Your advantage: [reason]

    2. "[Y] vs [Z] comparison" — 3,200/mo
    You have data nobody else has

    [continues for 10-15 opportunities]
    4
    Implement continuously
    Have Claude write the content too
    Once you've approved the plan, ask Claude to draft the highest-priority pages. It knows your site, your voice, and the target keywords from step 2.
    Follow-up: Draft the first article on the list.
    Use my existing site tone. Target the
    primary keyword naturally — don't
    keyword-stuff. Include an intro, 4
    sections, and a CTA at the end.

    "I got Claude Code to build me an entire SEO strategy. It researched my competitors, analysed keywords, looked at what content is ranking, and came back with a prioritised plan of exactly what pages I should add to my website. All backed by real data. And it took about 20 minutes."

    — Solo SaaS founder, NoCodeSaaS.io
    Use Case 06
    Docs & Release Notes
    The documentation tax is real — every PM spends hours per week on meeting notes, agendas, release notes, and stakeholder updates. Claude Code automates all of it, with context about your product baked in.
    Without Claude Code
    1-2 hrs/week
    Writing release notes, agendas, and updates manually
    →
    With Claude Code
    2 min
    All docs generated from raw notes with one command
    📝
    3 workflows · all automated
    Meeting Notes → Agenda → Release Notes → Stakeholder Update
    The full documentation pipeline. Raw notes in, polished deliverables out. Claude knows your product terminology and team structure from CLAUDE.md.
    2 min each PMs · Founders · Chiefs of Staff
    ⌄
    A
    Meeting Notes → Structured Summary
    Paste messy notes, get a clean summary
    Prompt: Here are my raw notes from today's sprint
    planning: [paste notes]

    Generate: a structured meeting summary
    with decisions made, action items (with
    owners), open questions, and a 3-sentence
    TL;DR I can post in Slack.
    B
    Release Notes → Multiple Formats
    One source, three audiences automatically
    Prompt: Read the git commits from this sprint
    (or paste the list). Generate release notes in
    3 formats:
    1. Technical (for engineers): exact changes
    2. Customer-facing: plain English benefits
    3. Exec summary: 3 bullets, business impact only
    C
    Investor/Board Update → Draft
    Turn your metrics into a narrative
    Prompt: Here are this month's numbers:
    MRR: $142K (+8%), Churn: 2.1%,
    New signups: 847, NPS: 61

    Key wins: [list]
    Key challenges: [list]

    Draft a monthly investor update email.
    Lead with the narrative, back with data.
    Confident but honest tone. Under 300 words.
    The documentation flywheel
    • Add your team structure and product terminology to CLAUDE.md once — all docs use it forever
    • Claude knows your product context — no re-explaining what "Workspace" vs "Project" means each time
    • Save prompt templates as CLAUDE.md skills — one command generates each doc type instantly
    • Entire documentation workflow runs in under 10 minutes per week