Stop Collecting AI Tools. Start Using One Workflow.

The AI Tools Graveyard

Your Slack has a #ai-tools-pm channel with 47 recommendations.

Your bookmark folder has 12 "AI tools to try later." Later never comes.

You've signed up for:

• ChatGPT (for general stuff)

• Claude (because everyone's talking about it)

• Cursor (someone at your last company swore by it)

• Perplexity (research synthesis)

• NotebookLM (document analysis)

• Productboard AI (if you use Productboard)

• Notion AI (because you have a Notion workspace)

• And 6 others you forgot you subscribed to

You're paying $150/month for tools. You're using 1.5 of them.

This is the AI tools trap. And it catches 95% of startup PMs.

Why Having "The Right Stack" Doesn't Matter

Here's what doesn't work:

1. Collecting tools

2. Trying each one once

3. Feeling productivity guilt when you don't use them

4. Reading comparisons of "15 Best AI Tools for PMs"

5. Hoping that the perfect stack exists

The perfect stack doesn't exist. What matters is one workflow you actually use.

Not the number of tools. The depth of integration into how you actually work.

A PM with Claude + one workflow ships more than a PM with 8 tools and no system.

The PM Who Actually Wins

I know a founder at a Series A company who uses:

• Claude (in the browser)

• Cursor (for code + doc drafting)

• One Notion page (not a workspace)

That's it. Three tools.

But here's what makes it work: She's internalized a workflow.

When she needs to:

• Write a PRD → Claude + context file + template

• Analyze user research → Claude + NotebookLM equivalent (just in Claude)

• Stress-test a strategy → Claude with a specific prompt structure

• Draft a roadmap → Cursor for outline, Claude for narrative

• Prepare for a hard stakeholder call → Claude with her "executive prep" prompt

She doesn't browse a directory of tools. She doesn't wonder which is better. She has one thinking partner (Claude) that she's learned to work with like a colleague.

Her output quality has tripled. Her decision-making is sharper. Her time on drudgework dropped 70%.

Not because she has the best tools. Because she uses one tool deeply.

The One Workflow That Actually Works

Here's the minimal viable PM + AI system:

Step 1: Pick Your Main AI (Claude or ChatGPT)

• Claude is stronger for PM work (better at trade-off analysis, strategic thinking)

• ChatGPT is more intuitive if you're new to AI

• Pick one. Commit for 3 months. Stop evaluating.

Cost: Free or $20/month

Step 2: Build Your PM Context File

Create a single document (in your chosen AI) that contains:

[Your Product]

• What it does
• Who uses it
• Current size (revenue, users, team)
• Biggest constraint (money, time, technical debt, market fit)

[Your Market]

• 3 closest competitors
• Why you're different
• What's changing in the market

[Your Team]

• How many engineers/designers/PMs
• What they're good at
• What's bottlenecking them

[Your Goal This Quarter]

• One primary metric you're moving
• One secondary metric

[Your Decision Framework]

• How you actually prioritize (revenue impact? user love? risk? speed?)
• Past decisions you regret (so AI doesn't suggest those again)

This is 1 page. Not 10 pages. The only context your AI needs.

Step 3: Build Your Workflow Template

Create prompts for your most common PM tasks:

For PRDs:

Write a PRD for [feature]. Context: [dump your notes here] Constraints: [what we can't do] Output structure: Problem, why now, solution, success metrics

For prioritization:

I have 5 feature requests. Rank by impact using [your decision framework].

• Request 1: [details]
• Request 2: [details] For each, show: Impact on [your metric], timeline, risk.

For strategy decisions:

I'm debating between approach A and approach B. [Explain both] Help me think through: reversibility, downside risk, upside potential, and what information would change my mind.

These become your repeatable playbook.

Step 4: Run It

Each time you face a PM problem:

1. Open Claude (or ChatGPT)

2. Paste your context file

3. Paste the relevant template

4. Execute in 15 minutes

5. Output is 80% done. You refine it.

Step 5: Add One Specialization (Optional)

• If you do heavy research synthesis: Add NotebookLM ($0)

• If you're doing rapid prototyping: Add Cursor or v0 ($0-50/month)

• If you're doing data-heavy analysis: Add Perplexity ($0-20/month)

But don't add these unless you're actually doing that work daily.

Total cost: $20-40/month

Setup time: 2 hours (context file + 3 templates)

Monthly maintenance: 30 minutes (updating context as product evolves)

What NOT to Do

• Don't try "the best tool for each task" (you'll thrash)

• Don't use 4+ AI tools (context switching kills productivity)

• Don't build an 8-page context document (you'll never update it)

• Don't use separate tools for PRDs, research, prioritization, and strategy (use one tool for all of it)

• Don't read listicles comparing 15 tools and try to optimize (pick one, go deep)

The Real Unlock: AI as Your Thinking Partner

Here's what nobody tells you:

The value of AI isn't speed. It's clarity.

When you write a prompt to Claude asking "should we build X or Y," you're forced to articulate:

• What matters about this decision

• How you actually measure success

• What constraints you're working with

• What information would change your mind

Writing that out changes your thinking. The AI's response is secondary.

The PMs who win with AI aren't using it to generate more outputs. They're using it to think more clearly.

The ones who lose are cycling through 10 tools, trying to automate their way to better decisions.

You can't automate thinking. You can only make it structured.

Your Action Plan (This Week)

Monday: Pick Claude or ChatGPT. Commit.

Tuesday-Wednesday: Write your context file (1 page).

Thursday: Create 3 templates:

• PRD template

• Prioritization template

• Strategy template

Friday: Use it on one real problem from work.

Result: You've built your AI workflow. You're ahead of 95% of PMs.

Next: Use it on every PM task for 2 weeks. You'll integrate it into how you think.

Month 2: You notice decisions are sharper. You're spending less time on drudgework.

Month 3: You wonder how you ever worked without it.

---

The Uncomfortable Truth

Most PMs fail with AI not because they lack good tools.

They fail because they:

1. Try to use too many tools (cognitive overload)

2. Don't have a clear workflow (just asking random questions)

3. Don't give AI enough context (expecting magic)

4. Expect it to replace judgment (it doesn't, it clarifies it)

5. Give up after 2 weeks because the first output wasn't perfect

The PMs winning with AI right now are doing something boring:

• Using one tool

• Running a consistent workflow

• Feeding it good context

• Iterating on outputs like any other thinking tool

• Committing for the long term

It's not sexy. It's not exciting. It doesn't make for a good Twitter thread.

But it works.

---

The Tools That Matter (If You Must Pick)

For general PM work: Claude

For rapid prototyping: Cursor

For research synthesis: NotebookLM

For meeting transcription: Granola

For strategic thinking: Still Claude

But honestly? 80% of your AI PM work can happen in Claude with a good workflow.

The other 4 tools are nice-to-haves, not must-haves.

---

Stop Collecting. Start Building.

There are 121 AI tools for product managers.

You need 1 workflow.

Pick your tool. Build your context. Create your templates. Run it on real work.

Everything else is distraction.

The startup that wins won't be the one with the fanciest AI stack. It'll be the one with the clearest thinking.

AI is the multiplier. But your workflow is the foundation.

Start there.