I Tested 5 AI Tools To Write A PRD—Here's The Winner Fireside Product Management podcast

Content provided by Tom Leung. All podcast content including episodes, graphics, and podcast descriptions are uploaded and provided directly by Tom Leung or their podcast platform partner. If you believe someone is using your copyrighted work without your permission, you can follow the process outlined here https://player.fm/legal.

Fireside Product Management »
I Tested 5 AI Tools to Write a PRD—Here's the Winner

2d ago 52:07

MP3•Episode home

TLDR: It was Claude :-)When I set out to compare ChatGPT, Claude, Gemini, Grok, and ChatPRD for writing Product Requirement Documents, I figured they’d all be roughly equivalent. Maybe some subtle variations in tone or structure, but nothing earth-shattering. They’re all built on similar transformer architectures, trained on massive datasets, and marketed as capable of handling complex business writing.

What I discovered over 45 minutes of hands-on testing revealed not just which tools are better for PRD creation, but why they’re better, and more importantly, how you should actually be using AI to accelerate your product work without sacrificing quality or strategic thinking.

If you’re an early or mid-career PM in Silicon Valley, this matters to you. Because here’s the uncomfortable truth: your peers are already using AI to write PRDs, analyze features, and generate documentation. The question isn’t whether to use these tools. The question is whether you’re using the right ones most effectively.

So let me walk you through exactly what I did, what I learned, and what you should do differently.

The Setup: A Real-World Test Case

Here’s how I structured the experiment. As I said at the beginning of my recording, “We are back in the Fireside PM podcast and I did that review of the ChatGPT browser and people seemed to like it and then I asked, uh, in a poll, I think it was a LinkedIn poll maybe, what should my next PM product review be? And, people asked for ChatPRD.”

So I had my marching orders from the audience. But I wanted to make this more comprehensive than just testing ChatPRD in isolation. I opened up five tabs: ChatGPT, Claude, Gemini, Grok, and ChatPRD.

For the test case, I chose something realistic and relevant: an AI-powered tutor for high school students. Think KhanAmigo or similar edtech platforms. This gave me a concrete product scenario that’s complex enough to stress-test these tools but straightforward enough that I could iterate quickly.

But here’s the critical part that too many PMs get wrong when they start using AI for product work: I didn’t just throw a single sentence at these tools and expect magic.

The “Back of the Napkin” Approach: Why You Still Need to Think

“I presume everybody agrees that you should have some formulated thinking before you dump it into the chatbot for your PRD,” I noted early in my experiment. “I suppose in the future maybe you could just do, like, a one-sentence prompt and come out with the perfect PRD because it would just know everything about you and your company in the context, but for now we’re gonna do this more, a little old-school AI approach where we’re gonna do some original human thinking.”

This is crucial. I see so many PMs, especially those newer to the field, treat AI like a magic oracle. They type in “Write me a PRD for a social feature” and then wonder why the output is generic, unfocused, and useless.

Your job as a PM isn’t to become obsolete. It’s to become more effective. And that means doing the strategic thinking work that AI cannot do for you.

So I started in Google Docs with what I call a “back of the napkin” PRD structure. Here’s what I included:

Why: The strategic rationale. In this case: “Want to complement our existing edtech business with a personalized AI tutor, uh, want to maintain position industry, and grow through innovation. on mission for learners.”

Target User: Who are we building for? “High school students interested in improving their grades and fundamentals. Fundamental knowledge topics. Specifically science and math. Students who are not in the top ten percent, nor in the bottom ten percent.”

This is key—I got specific. Not just “students,” but students in the middle 80%. Not just “any subject,” but science and math. This specificity is what separates useful AI output from garbage.

Problem to Solve: What’s broken? “Students want better grades. Students are impatient. Students currently use AI just for finding the answers and less to, uh, understand concepts and practice using them.”

Key Elements: The feature set and approach.

Success Metrics: How we’d measure success.

Now, was this a perfectly polished PRD outline? Hell no. As you can see from my transcript, I was literally thinking out loud, making typos, restructuring on the fly. But that’s exactly the point. I put in maybe 10-15 minutes of human strategic thinking. That’s all it took to create a foundation that would dramatically improve what came out of the AI tools.

Round One: Generating the Full PRD

With my back-of-the-napkin outline ready, I copied it into each tool with a simple prompt asking them to expand it into a more complete PRD.

ChatGPT: The Reliable Generalist

ChatGPT gave me something that was... fine. Competent. Professional. But also deeply uninspiring.

The document it produced checked all the boxes. It had the sections you’d expect. The writing was clear. But when I read it, I couldn’t shake the feeling that I was reading something that could have been written for literally any product in any company. It felt like “an average of everything out there,” as I noted in my evaluation.

Here’s what ChatGPT did well: It understood the basic structure of a PRD. It generated appropriate sections. The grammar and formatting were clean. If you needed to hand something in by EOD and had literally no time for refinement, ChatGPT would save you from complete embarrassment.

But here’s what it lacked: Depth. Nuance. Strategic thinking that felt connected to real product decisions. When it described the target user, it used phrases that could apply to any edtech product. When it outlined success metrics, they were the obvious ones (engagement, retention, test scores) without any interesting thinking about leading indicators or proxy metrics.

The problem with generic output isn’t that it’s wrong, it’s that it’s invisible. When you’re trying to get buy-in from leadership or alignment from engineering, you need your PRD to feel specific, considered, and connected to your company’s actual strategy. ChatGPT’s output felt like it was written by someone who’d read a lot of PRDs but never actually shipped a product.

One specific example: When I asked for success metrics, ChatGPT gave me “Student engagement rate, Time spent on platform, Test score improvement.” These aren’t wrong, but they’re lazy. They don’t show any thinking about what specifically matters for an AI tutor versus any other educational product. Compare that to Claude’s output, which got more specific about things like “concept mastery rate” and “question-to-understanding ratio.”

Actionable Insight: Use ChatGPT when you need fast, serviceable documentation that doesn’t need to be exceptional. Think: internal updates, status reports, routine communications. Don’t rely on it for strategic documents where differentiation matters. If you do use ChatGPT for important documents, treat its output as a starting point that needs significant human refinement to add strategic depth and company-specific context.

Gemini: Better Than Expected

Google’s Gemini actually impressed me more than I anticipated. The structure was solid, and it had a nice balance of detail without being overwhelming.

What Gemini got right: The writing had a nice flow to it. The document felt organized and logical. It did a better job than ChatGPT at providing specific examples and thinking through edge cases. For instance, when describing the target user, it went beyond demographics to consider behavioral characteristics and motivations.

Gemini also showed some interesting strategic thinking. It considered competitive positioning more thoughtfully than ChatGPT and proposed some differentiation angles that weren’t in my original outline. Good AI tools should add insight, not just regurgitate your input with better formatting.

But here’s where it fell short: the visual elements. When I asked for mockups, Gemini produced images that looked more like stock photos than actual product designs. They weren’t terrible, but they weren’t compelling either. They had that AI-generated sheen that makes it obvious they came from an image model rather than a designer’s brain.

For a PRD that you’re going to use internally with a team that already understands the context, Gemini’s output would work well. The text quality is strong enough, and if you’re in the Google ecosystem (Docs, Sheets, Meet, etc.), the integration is seamless. You can paste Gemini’s output directly into Google Docs and continue iterating there.

But if you need to create something compelling enough to win over skeptics or secure budget, Gemini falls just short. It’s good, but not great. It’s the solid B+ student: reliably competent but rarely exceptional.

Actionable Insight: Gemini is a strong choice if you’re working in the Google ecosystem and need good integration with Docs, Sheets, and other Google Workspace tools. The quality is sufficient for most internal documentation needs. It’s particularly good if you’re working with cross-functional partners who are already in Google Workspace. You can share and collaborate on AI-generated drafts without friction. But don’t expect visual mockups that will wow anyone, and plan to add your own strategic polish for high-stakes documents.

Grok: Not Ready for Prime Time

Let’s just say my expectations were low, and Grok still managed to underdeliver. The PRD felt thin, generic, and lacked the depth you need for real product work.

“I don’t have high expectations for grok, unfortunately,” I said before testing it. Spoiler alert: my low expectations were validated.

Actionable Insight: Skip Grok for product documentation work right now. Maybe it’ll improve, but as of my testing, it’s simply not competitive with the other options. It felt like 1-2 years behind the others.

ChatPRD: The Specialized Tool

Now this was interesting. ChatPRD is purpose-built for PRDs, using foundational models underneath but with specific tuning and structure for product documentation.

The result? The structure was logical, the depth was appropriate, and it included elements that showed understanding of what actually matters in a PRD. As I reflected: “Cause this one feels like, A human wrote this PRD.”

The interface guides you through the process more deliberately than just dumping text into a general chat interface. It asks clarifying questions. It structures the output more thoughtfully.

Actionable Insight: If you’re a technical lead without a dedicated PM, or you’re a PM who wants a more structured approach to using AI for PRDs, ChatPRD is worth the specialized focus. It’s particularly good when you need something that feels authentic enough to share with stakeholders without heavy editing.

Claude: The Clear Winner

But the standout performer, and I’m ranking these, was Claude.

“I think we know that for now, I’m gonna say Claude did the best job,” I concluded after all the testing. Claude produced the most comprehensive, thoughtful, and strategically sound PRD. But what really set it apart were the concept mocks.

When I asked each tool to generate visual mockups of the product, Claude produced HTML prototypes that, while not fully functional, looked genuinely compelling. They had thoughtful UI design, clear information architecture, and felt like something that could actually guide development.

“They were, like, closer to, like, what a Lovable would produce or something like that,” I noted, referring to the quality of low-fidelity prototypes that good designers create.

The text quality was also superior: more nuanced, better structured, and with more strategic depth. It felt like Claude understood not just what a PRD should contain, but why it should contain those elements.

Actionable Insight: For any PRD that matters, meaning anything you’ll share with leadership, use to get buy-in, or guide actual product development, you might as well start with Claude. The quality difference is significant enough that it’s worth using Claude even if you primarily use another tool for other tasks.

Final Rankings: The Definitive Hierarchy

After testing all five tools on multiple dimensions: initial PRD generation, visual mockups, and even crafting a pitch paragraph for a skeptical VP of Engineering, here’s my final ranking:

* Claude - Best overall quality, most compelling mockups, strongest strategic thinking

* ChatPRD - Best for structured PRD creation, feels most “human”

* Gemini - Solid all-around performance, good Google integration

* ChatGPT - Reliable but generic, lacks differentiation

* Grok - Not competitive for this use case

“I’d probably say Claude, then chat PRD, then Gemini, then chat GPT, and then Grock,” I concluded.

The Deeper Lesson: Garbage In, Garbage Out (Still Applies)

But here’s what matters more than which tool wins: the realization that hit me partway through this experiment.

“I think it really does come down to, like, you know, the quality of the prompt,” I observed. “So if our prompt were a little more detailed, all that were more thought-through, then I’m sure the output would have been better. But as you can see we didn’t really put in brain trust prompting here. Just a little bit of, kind of hand-wavy prompting, but a little better than just one or two sentences.”

And we still got pretty good results.

This is the meta-insight that should change how you approach AI tools in your product work: The quality of your input determines the quality of your output, but the baseline quality of the tool determines the ceiling of what’s possible.

No amount of great prompting will make Grok produce Claude-level output. But even mediocre prompting with Claude will beat great prompting with lesser tools.

So the dual strategy is:

* Use the best tool available (currently Claude for PRDs)

* Invest in improving your prompting skills ideally with as much original and insightful human, company aware, and context aware thinking as possible.

Real-World Workflows: How to Actually Use This in Your Day-to-Day PM Work

Theory is great. Here’s how to incorporate these insights into your actual product management workflows.

The Weekly Sprint Planning Workflow

Every PM I know spends hours each week preparing for sprint planning. You need to refine user stories, clarify acceptance criteria, anticipate engineering questions, and align with design and data science. AI can compress this work significantly.

Here’s an example workflow:

Monday morning (30 minutes):

* Review upcoming priorities and open your rough notes/outline in Google Docs

* Open Claude and paste your outline with this prompt:

“I’m preparing for sprint planning. Based on these priorities [paste notes], generate detailed user stories with acceptance criteria. Format each as: User story, Business context, Technical considerations, Acceptance criteria, Dependencies, Open questions.”

Monday afternoon (20 minutes):

* Review Claude’s output critically

* Identify gaps, unclear requirements, or missing context

* Follow up with targeted prompts:

“The user story about authentication is too vague. Break it down into separate stories for: social login, email/password, session management, and password reset. For each, specify security requirements and edge cases.”

Tuesday morning (15 minutes):

* Generate mockups for any UI-heavy stories:

“Create an HTML mockup for the login flow showing: landing page, social login options, email/password form, error states, and success redirect.”

* Even if the HTML doesn’t work perfectly, it gives your designers a starting point

Before sprint planning (10 minutes):

* Ask Claude to anticipate engineering questions:

“Review these user stories as if you’re a senior engineer. What questions would you ask? What concerns would you raise about technical feasibility, dependencies, or edge cases?”

* This preparation makes you look thoughtful and helps the meeting run smoothly

Total time investment: ~75 minutes. Typical time saved: 3-4 hours compared to doing this manually.

The Stakeholder Alignment Workflow

Getting alignment from multiple stakeholders (product leadership, engineering, design, data science, legal, marketing) is one of the hardest parts of PM work. AI can help you think through different stakeholder perspectives and craft compelling communications for each.

Here’s how:

Step 1: Map your stakeholders (10 minutes)

Create a quick table in a doc:

Step 2: Generate stakeholder-specific communications (20 minutes)

For each key stakeholder, ask Claude:

“I need to pitch this product idea to [Stakeholder]. Based on this PRD, create a 1-page brief addressing their primary concern of [concern from your table]. Open with the specific value for them, address their likely objection of [objection], and close with a clear ask. Tone should be [professional/technical/strategic] based on their role.”

Then you’ll have customized one-pagers for your pre-meetings with each stakeholder, dramatically increasing your alignment rate.

Step 3: Synthesize feedback (15 minutes)

After gathering stakeholder input, ask Claude to help you synthesize:

“I got the following feedback from stakeholders: [paste feedback]. Identify: (1) Common themes, (2) Conflicting requirements, (3) Legitimate concerns vs organizational politics, (4) Recommended compromises that might satisfy multiple parties.”

This pattern-matching across stakeholder feedback is something AI does really well and saves you hours of mental processing.

The Quarterly Planning Workflow

Quarterly or annual planning is where product strategy gets real. You need to synthesize market trends, customer feedback, technical capabilities, and business objectives into a coherent roadmap. AI can accelerate this dramatically.

Six weeks before planning:

* Start collecting input (customer interviews, market research, competitive analysis, engineering feedback)

* Don’t wait until the last minute

Four weeks before planning:

Dump everything into Claude with this structure:

“I’m creating our Q2 roadmap. Context:

* Business objectives: [paste from leadership]

* Customer feedback themes: [paste synthesis]

* Technical capabilities/constraints: [paste from engineering]

* Competitive landscape: [paste analysis]

* Current product gaps: [paste from your analysis]

Generate 5 strategic themes that could anchor our Q2 roadmap. For each theme:

* Strategic rationale (how it connects to business objectives)

* Key initiatives (2-3 major features/projects)

* Success metrics

* Resource requirements (rough estimate)

* Risks and mitigations

* Customer segments addressed”

This gives you a strategic framework to react to rather than starting from a blank page.

Three weeks before planning:

Iterate on the most promising themes:

“Deep dive on Theme 3. Generate:

* Detailed initiative breakdown

* Dependencies on platform/infrastructure

* Phasing options (MVP vs full build)

* Go-to-market considerations

* Data requirements

* Open questions requiring research”

Two weeks before planning:

Pressure-test your thinking:

“Play devil’s advocate on this roadmap. What are the strongest arguments against each initiative? What am I likely missing? What failure modes should I plan for?”

This adversarial prompting forces you to strengthen weak points before your leadership reviews it.

One week before planning:

Generate your presentation:

“Create an executive presentation for this roadmap. Structure: (1) Market context and strategic imperative, (2) Q2 themes and initiatives, (3) Expected outcomes and metrics, (4) Resource requirements, (5) Key risks and mitigations, (6) Success criteria for decision. Make it compelling but data-driven. Tone: confident but not overselling.”

Then add your company-specific context, visual brand, and personal voice.

The Customer Research Workflow

AI can’t replace talking to customers, but it can help you prepare better questions, analyze feedback more systematically, and identify patterns faster.

Before customer interviews:

“I’m interviewing customers about [topic]. Generate:

* 10 open-ended questions that avoid leading the witness

* 5 follow-up questions for each main question

* Common cognitive biases I should watch for

* A framework for categorizing responses”

This prep work helps you conduct better interviews.

After interviews:

“I conducted 15 customer interviews. Here are the key quotes: [paste anonymized quotes]. Identify:

* Recurring themes and patterns

* Surprising insights that contradict our assumptions

* Segments with different needs

* Implied needs customers didn’t articulate directly

* Recommended next steps for validation”

AI is excellent at pattern-matching across qualitative data at scale.

The Crisis Management Workflow

Something broke. The site is down. Data was lost. A feature shipped with a critical bug. You need to move fast.

Immediate response (5 minutes):

“Critical incident. Details: [brief description]. Generate:

* Incident classification (Sev 1-4)

* Immediate stakeholders to notify

* Draft customer communication (honest, apologetic, specific about what happened and what we’re doing)

* Draft internal communication for leadership

* Key questions to ask engineering during investigation”

Having these drafted in 5 minutes lets you focus on coordination and decision-making rather than wordsmithing.

Post-incident (30 minutes):

“Write a post-mortem based on this incident timeline: [paste timeline]. Include:

* What happened (technical details)

* Root cause analysis

* Impact quantification (users affected, revenue impact, time to resolution)

* What went well in our response

* What could have been better

* Specific action items with owners and deadlines

* Process changes to prevent recurrence Tone: Blameless, focused on learning and improvement.”

This gives you a strong first draft to refine with your team.

Common Pitfalls: What Not to Do with AI in Product Management

Now let’s talk about the mistakes I see PMs making with AI tools.

Pitfall #1: Treating AI Output as Final

The biggest mistake is copy-pasting AI output directly into your PRD, roadmap presentation, or stakeholder email without critical review.

The result? Documents that are grammatically perfect but strategically shallow. Presentations that sound impressive but don’t hold up under questioning. Emails that are professionally worded but miss the subtext of organizational politics.

The fix: Always ask yourself:

* Does this reflect my actual strategic thinking, or generic best practices?

* Would my CEO/engineering lead/biggest customer find this compelling and specific?

* Are there company-specific details, customer insights, or technical constraints that only I know?

* Does this sound like me, or like a robot?

Add those elements. That’s where your value as a PM comes through.

Pitfall #2: Using AI as a Crutch Instead of a Tool

Some PMs use AI because they don’t want to think deeply about the product. They’re looking for AI to do the hard work of strategy, prioritization, and trade-off analysis.

This never works. AI can help you think more systematically, but it can’t replace thinking.

If you find yourself using AI to avoid wrestling with hard questions (”Should we build X or Y?” “What’s our actual competitive advantage?” “Why would customers switch from the incumbent?”), you’re using it wrong.

The fix: Use AI to explore options, not to make decisions. Generate three alternatives, pressure-test each one, then use your judgment to decide. The AI can help you think through implications, but you’re still the one choosing.

Pitfall #3: Not Iterating

Getting mediocre AI output and just accepting it is a waste of the technology’s potential.

The PMs who get exceptional results from AI are the ones who iterate. They generate an initial response, identify what’s weak or missing, and ask follow-up questions. They might go through 5-10 iterations on a key section of a PRD.

Each iteration is quick (30 seconds to type a follow-up prompt, 30 seconds to read the response), but the cumulative effect is dramatically better output.

The fix: Budget time for iteration. Don’t try to generate a complete, polished PRD in one prompt. Instead, generate a rough draft, then spend 30 minutes iterating on specific sections that matter most.

Pitfall #4: Ignoring the Political and Human Context

AI tools have no understanding of organizational politics, interpersonal relationships, or the specific humans you’re working with.

They don’t know that your VP of Engineering is burned out and skeptical of any new initiatives. They don’t know that your CEO has a personal obsession with a specific competitor. They don’t know that your lead designer is sensitive about not being included early enough in the process.

If you use AI-generated communications without layering in this human context, you’ll create perfectly worded documents that land badly because they miss the subtext.

The fix: After generating AI content, explicitly ask yourself: “What human context am I missing? What relationships do I need to consider? What political dynamics are in play?” Then modify the AI output accordingly.

Pitfall #5: Over-Relying on a Single Tool

Different AI tools have different strengths. Claude is great for strategic depth, ChatPRD is great for structure, Gemini integrates well with Google Workspace.

If you only ever use one tool, you’re missing opportunities to leverage different strengths for different tasks.

The fix: Keep 2-3 tools in your toolkit. Use Claude for important PRDs and strategic documents. Use Gemini for quick internal documentation that needs to integrate with Google Docs. Use ChatPRD when you want more guided structure. Match the tool to the task.

Pitfall #6: Not Fact-Checking AI Output

AI tools hallucinate. They make up statistics, misrepresent competitors, and confidently state things that aren’t true. If you include those hallucinations in a PRD that goes to leadership, you look incompetent.

The fix: Fact-check everything, especially:

* Statistics and market data

* Competitive feature claims

* Technical capabilities and limitations

* Regulatory and compliance requirements

If the AI cites a number or makes a factual claim, verify it independently before including it in your document.

The Meta-Skill: Prompt Engineering for PMs

Let’s zoom out and talk about the underlying skill that makes all of this work: prompt engineering.

This is a real skill. The difference between a mediocre prompt and a great prompt can be 10x difference in output quality. And unlike coding or design, where there’s a steep learning curve, prompt engineering is something you can get good at quickly.

Principle 1: Provide Context Before Instructions

Bad prompt:

“Write a PRD for an AI tutor”

Good prompt:

“I’m a PM at an edtech company with 2M users, primarily high school students. We’re exploring an AI tutor feature to complement our existing video content library and practice problems. Our main competitors are Khan Academy and Course Hero. Our differentiation is personalized learning paths based on student performance data.

Write a PRD for an AI tutor feature targeting students in the middle 80% academically who struggle with science and math.”

The second prompt gives Claude the context it needs to generate something specific and strategic rather than generic.

Principle 2: Specify Format and Constraints

Bad prompt:

“Generate success metrics”

Good prompt:

“Generate 5-7 success metrics for this feature. Include a mix of:

* Leading indicators (early signals of success)

* Lagging indicators (definitive success measures)

* User behavior metrics

* Business impact metrics

For each metric, specify: name, definition, target value, measurement method, and why it matters.”

The structure you provide shapes the structure you get back.

Principle 3: Ask for Multiple Options

Bad prompt:

“What should our Q2 priorities be?”

Good prompt:

“Generate 3 different strategic approaches for Q2:

* Option A: Focus on user acquisition

* Option B: Focus on engagement and retention

* Option C: Focus on monetization

For each option, detail: key initiatives, expected outcomes, resource requirements, risks, and recommendation for or against.”

Asking for multiple options forces the AI (and forces you) to think through trade-offs systematically.

Principle 4: Specify Audience and Tone

Bad prompt:

“Summarize this PRD”

Good prompt:

“Create a 1-paragraph summary of this PRD for our skeptical VP of Engineering. Tone: Technical, concise, addresses engineering concerns upfront. Focus on: technical architecture, resource requirements, risks, and expected engineering effort. Avoid marketing language.”

The audience and tone specification ensures the output will actually work for your intended use.

Principle 5: Use Iterative Refinement

Don’t try to get perfect output in one prompt. Instead:

First prompt: Generate rough draft Second prompt: “This is too generic. Add specific examples from [our company context].” Third prompt: “The technical section is weak. Expand with architecture details and dependencies.” Fourth prompt: “Good. Now make it 30% more concise while keeping the key details.”

Each iteration improves the output incrementally.

Let me break down the prompting approach that worked in this experiment, because this is immediately actionable for your work tomorrow.

Strategy 1: The Structured Outline Approach

Don’t go from zero to full PRD in one prompt. Instead:

* Start with strategic thinking - Spend 10-15 minutes outlining why you’re building this, who it’s for, and what problem it solves

* Get specific - Don’t say “users,” say “high school students in the middle 80% of academic performance”

* Include constraints - Budget, timeline, technical limitations, competitive landscape

* Dump your outline into the AI - Now ask it to expand into a full PRD

* Iterate section by section - Don’t try to perfect everything at once

This is exactly what I did in my experiment, and even with my somewhat sloppy outline, the results were dramatically better than they would have been with a single-sentence prompt.

Strategy 2: The Comparative Analysis Pattern

One technique I used that worked particularly well: asking each tool to do the same specific task and comparing results.

For example, I asked all five tools: “Please compose a one paragraph exact summary I can share over DM with a highly influential VP of engineering who is generally a skeptic but super smart.”

This forced each tool to synthesize the entire PRD into a compelling pitch while accounting for a specific, challenging audience. The variation in quality was revealing—and it gave me multiple options to choose from or blend together.

Actionable tip: When you need something critical (a pitch, an executive summary, a key decision framework), generate it with 2-3 different AI tools and take the best elements from each. This “ensemble approach” often produces better results than any single tool.

Strategy 3: The Iterative Refinement Loop

Don’t treat the AI output as final. Use it as a first draft that you then refine through conversation with the AI.

After getting the initial PRD, I could have asked follow-up questions like:

* “What’s missing from this PRD?”

* “How would you strengthen the success metrics section?”

* “Generate 3 alternative approaches to the core feature set”

Each iteration improves the output and, more importantly, forces me to think more deeply about the product.

What This Means for Your Career

If you’re an early or mid-career PM reading this, you might be thinking: “Great, so AI can write PRDs now. Am I becoming obsolete?”

Absolutely not. But your role is evolving, and understanding that evolution is critical.

The PMs who will thrive in the AI era are those who:

* Excel at strategic thinking - AI can generate options, but you need to know which options align with company strategy, customer needs, and technical feasibility

* Master the art of prompting - This is a genuine skill that separates mediocre AI users from exceptional ones

* Know when to use AI and when not to - Some aspects of product work benefit enormously from AI. Others (user interviews, stakeholder negotiation, cross-functional relationship building) require human judgment and empathy

* Can evaluate AI output critically - You need to spot the hallucinations, the generic fluff, and the strategic misalignments that AI inevitably produces

Think of AI tools as incredibly capable interns. They can produce impressive work quickly, but they need direction, oversight, and strategic guidance. Your job is to provide that guidance while leveraging their speed and breadth.

The Real-World Application: What to Do Monday Morning

Let’s get tactical. Here’s exactly how to apply these insights to your actual product work:

For Your Next PRD:

* Block 30 minutes for strategic thinking - Write your back-of-the-napkin outline in Google Docs or your tool of choice

* Open Claude (or ChatPRD if you want more structure)

* Copy your outline with this prompt:

“I’m a product manager at [company] working on [product area]. I need to create a comprehensive PRD based on this outline. Please expand this into a complete PRD with the following sections: [list your preferred sections]. Make it detailed enough for engineering to start breaking down into user stories, but concise enough for leadership to read in 15 minutes. [Paste your outline]”

* Review the output critically - Look for generic statements, missing details, or strategic misalignments

* Iterate on specific sections:

“The success metrics section is too vague. Please provide 3-5 specific, measurable KPIs with target values and explanation of why these metrics matter.”

* Generate supporting materials:

“Create a visual mockup of the core user flow showing the key interaction points.”

* Synthesize the best elements - Don’t just copy-paste the AI output. Use it as raw material that you shape into your final document

For Stakeholder Communication:

When you need to pitch something to leadership or engineering:

* Generate 3 versions of your pitch using different tools (Claude, ChatPRD, and one other)

* Compare them for:

* Clarity and conciseness

* Strategic framing

* Compelling value proposition

* Addressing likely objections

* Blend the best elements into your final version

* Add your personal voice - This is crucial. AI output often lacks personality and specific company context. Add that yourself.

For Feature Prioritization:

AI tools can help you think through trade-offs more systematically:

“I’m deciding between three features for our next release: [Feature A], [Feature B], and [Feature C]. For each feature, analyze: (1) Estimated engineering effort, (2) Expected user impact, (3) Strategic alignment with making our platform the go-to solution for [your market], (4) Risk factors. Then recommend a prioritization with rationale.”

This doesn’t replace your judgment, but it forces you to think through each dimension systematically and often surfaces considerations you hadn’t thought of.

The Uncomfortable Truth About AI and Product Management

Let me be direct about something that makes many PMs uncomfortable: AI will make some PM skills less valuable while making others more valuable.

Less valuable:

* Writing boilerplate documentation

* Creating standard frameworks and templates

* Generating routine status updates

* Synthesizing information from existing sources

More valuable:

* Strategic product vision and roadmapping

* Deep customer empathy and insight generation

* Cross-functional leadership and influence

* Critical evaluation of options and trade-offs

* Creative problem-solving for novel situations

If your PM role primarily involves the first category of tasks, you should be concerned. But if you’re focused on the second category while leveraging AI for the first, you’re going to be exponentially more effective than your peers who resist these tools.

The PMs I see succeeding aren’t those who can write the best PRD manually. They’re those who can write the best PRD with AI assistance in one-tenth the time, then use the saved time to talk to more customers, think more deeply about strategy, and build stronger cross-functional relationships.

Advanced Techniques: Beyond Basic PRD Generation

Once you’ve mastered the basics, here are some advanced applications I’ve found valuable:

Competitive Analysis at Scale

“Research our top 5 competitors in [market]. For each one, analyze: their core value proposition, key features, pricing strategy, target customer, and likely product roadmap based on recent releases and job postings. Create a comparison matrix showing where we have advantages and gaps.”

Then use web search tools in Claude or Perplexity to fact-check and expand the analysis.

Scenario Planning

“We’re considering three strategic directions for our product: [Direction A], [Direction B], [Direction C]. For each direction, map out: likely customer adoption curve, required technical investments, competitive positioning in 12 months, and potential pivots if the hypothesis proves wrong. Then identify the highest-risk assumptions we should test first for each direction.”

This kind of structured scenario thinking is exactly what AI excels at—generating multiple well-reasoned perspectives quickly.

User Story Generation

After your PRD is solid:

“Based on this PRD, generate a complete set of user stories following the format ‘As a [user type], I want to [action] so that [benefit].’ Include acceptance criteria for each story. Organize them into epics by functional area.”

This can save your engineering team hours of grooming meetings.

The Tools Will Keep Evolving. Your Process Shouldn’t

Here’s something important to remember: by the time you read this, the specific rankings might have shifted. Maybe ChatGPT-5 has leapfrogged Claude. Maybe a new specialized tool has emerged.

But the core principles won’t change:

* Do strategic thinking before touching AI

* Use the best tool available for your specific task

* Iterate and refine rather than accepting first outputs

* Blend AI capabilities with human judgment

* Focus your time on the uniquely human aspects of product management

The specific tools matter less than your process for using them effectively.

A Final Experiment: The Skeptical VP Test

I want to share one more insight from my testing that I think is particularly relevant for early and mid-career PMs.

Toward the end of my experiment, I gave each tool this prompt: “Please compose a one paragraph exact summary I can share over DM with a highly influential VP of engineering who is generally a skeptic but super smart.”

This is such a realistic scenario. How many times have you needed to pitch an idea to a skeptical technical leader via Slack or email? Someone who’s brilliant, who’s seen a thousand product ideas fail, and who can spot b******t from a mile away?

The quality variation in the responses was fascinating. ChatGPT gave me something that felt generic and safe. Gemini was better but still a bit too enthusiastic. Grok was... well, Grok.

But Claude and ChatPRD both produced messages that felt authentic, technically credible, and appropriately confident without being overselling. They acknowledged the engineering challenges while framing the opportunity compellingly.

The lesson: When the stakes are high and the audience is sophisticated, the quality of your AI tool matters even more. That skeptical VP can tell the difference between a carefully crafted message and AI-generated fluff. So can your CEO. So can your biggest customers.

Use the best tools available, but more importantly, always add your own strategic thinking and authentic voice on top.

Questions to Consider: A Framework for Your Own Experiments

As I wrapped up my Loom, I posed some questions to the audience that I’ll pose to you:

“Let me know in the comments, if you do your PRDs using AI differently, do you start with back of the envelope? Do you say, oh no, I just start with one sentence, and then I let the chatbot refine it with me? Or do you go way more detailed and then use the chatbot to kind of pressure test it?”

These aren’t rhetorical questions. Your answer reveals your approach to AI-augmented product work, and different approaches work for different people and contexts.

For early-career PMs: I’d recommend starting with more detailed outlines. The discipline of thinking through your product strategy before touching AI will make you a stronger PM. You can always compress that process later as you get more experienced.

For mid-career PMs: Experiment with different approaches for different types of documents. Maybe you do detailed outlines for major feature PRDs but use more iterative AI-assisted refinement for smaller features or updates. Find what optimizes your personal productivity while maintaining quality.

For senior PMs and product leaders: Consider how AI changes what you should expect from your PM team. Should you be reviewing more AI-generated first drafts and spending more time on strategic guidance? Should you be training your team on effective AI usage? These are leadership questions worth grappling with.

The Path Forward: Continuous Experimentation

My experiment with these five AI tools took 45 minutes. But I’m not done experimenting.

The field of AI-assisted product management is evolving rapidly. New tools launch monthly. Existing tools get smarter weekly. Prompting techniques that work today might be obsolete in three months.

Your job, if you want to stay at the forefront of product management, is to continuously experiment. Try new tools. Share what works with your peers. Build a personal knowledge base of effective prompts and workflows. And be generous with what you learn. The PM community gets stronger when we share insights rather than hoarding them.

That’s why I created this Loom and why I’m writing this post. Not because I have all the answers, but because I’m figuring it out in real-time and want to share the journey.

A Personal Note on Coaching and Consulting

If this kind of practical advice resonates with you, I’m happy to work with you directly.

Through my pm coaching practice, I offer 1:1 executive, career, and product coaching for PMs and product leaders. We can dig into your specific challenges: whether that’s leveling up your AI workflows, navigating a career transition, or developing your strategic product thinking.

I also work with companies (usually startups or incubation teams) on product strategy, helping teams figure out PMF for new explorations and improving their product management function.

The format is flexible. Some clients want ongoing coaching, others prefer project-based consulting, and some just want a strategic sounding board for a specific decision. Whatever works for you.

Reach out through tomleungcoaching.com if you’re interested in working together.

OK. Enough pontificating. Let’s ship greatness.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit firesidepm.substack.com

112 episodes

#Tech #MBA #Business #Tom Leung #Productdevelopment #Productmanagement #Productivity