← jasperhaynes.com
← Back to Blog

How to Write a PRD in 90 Minutes (Not 6 Hours)

• PRD, product requirements, AI, product management, documentation

Let me describe a ritual that happens in product teams everywhere, every sprint:

Day 1: You finish a discovery session with clear insights. You think, "I should write this up while it's fresh."

Day 2: Three meetings and a production incident later, you haven't started.

Day 5: You finally open a blank document. Stare at it. Write a title. Get pulled into a Slack thread.

Day 8: The sprint planning meeting is tomorrow. You copy-paste from the last PRD, change the feature name, and scramble to fill in the details.

Day 9: Engineering asks 14 questions during sprint planning because the PRD is missing acceptance criteria, edge cases, and technical constraints.

Sound familiar? You're not alone. The average PRD takes 6-8 hours to write, yet 73% of engineers say the PRDs they receive are "incomplete or unclear."

Why PRDs Take So Long

The Blank Page Problem

Starting from scratch is cognitively expensive. Every new PRD requires you to:

No wonder the blank page wins. It's not laziness — it's cognitive overload.

The Scattered Context Problem

The information needed for a PRD exists in at least 5 places:

  1. Meeting recordings — stakeholder requirements and decisions
  2. Slack threads — technical feasibility discussions and constraints
  3. Figma comments — design rationale
  4. Research documents — user interview insights
  5. Your brain — the accumulated context from weeks of discovery

Synthesizing all of this into one coherent document is genuinely hard work.

The Completeness Problem

Every product manager has their blind spots:

No single person consistently covers every section of a PRD equally well.

The Intelligent PRD Co-Pilot

The Intelligent PRD Co-Pilot, part of the Jasper Toolkit roadmap, addresses all three problems. Here's how it works.

Step 1: Feed It Your Context (10 minutes)

Instead of starting with a blank document, you feed the Co-Pilot your existing context:

Meeting transcripts: Upload recording transcripts from Otter.ai, Fireflies, or your tool of choice. The AI extracts:

Slack conversations: Paste the relevant thread. The AI identifies:

Research documents: Upload interview summaries, competitive analysis, or survey results. The AI pulls out:

Rough notes: Even a quick bullet-point outline works. The AI expands abbreviated notes into full requirements.

Step 2: AI-Generated First Draft (5 minutes)

The Co-Pilot produces a structured first draft covering:

Problem Statement

User Stories

Acceptance Criteria

Technical Considerations

Edge Cases

Success Metrics

Dependencies

Step 3: Human Review and Refinement (60-75 minutes)

This is where the product manager's judgment is irreplaceable. You review the AI-generated draft and:

The key shift: instead of creating content, you're reviewing and refining content. This is faster, produces better results, and is cognitively much less draining.

Step 4: Completeness Check (5 minutes)

Before you publish, the Co-Pilot runs a completeness validation:

You address the warnings, and the PRD is ready for review.

Step 5: Consistency Validation

If you've written PRDs before, the Co-Pilot checks for consistency with your existing product:

The 90-Minute PRD

Here's the breakdown:

Step Time Activity
Feed context 10 min Upload transcripts, paste Slack threads, add notes
AI generation 5 min Co-Pilot produces structured first draft
Human review 60-75 min Validate, refine, add strategic context
Completeness check 5 min Address gaps flagged by validation
Total 80-95 min Complete, reviewed, validated PRD

Compare this to the traditional process:

Step Time Activity
Open blank document 5 min Stare at empty page
Structure setup 20 min Create sections, remember template
Context recall 30 min Search Slack, replay recordings, find notes
First draft 120 min Write everything from memory
Self-review 30 min Catch gaps and inconsistencies
Revision 60 min Rewrite unclear sections
Total 265 min 4.4 hours — and it's still probably missing something

That's a 65% time reduction with a higher quality output.

Tips for Getting the Most Out of AI-Assisted PRDs

1. Record Everything

The best PRD inputs are meeting transcripts. If you're not recording your discovery sessions, stakeholder meetings, and design reviews, start now. Every unrecorded conversation is context the AI can't extract.

2. Don't Skip the Review Step

The AI produces a strong first draft, not a finished document. Your job as a PM is to add the judgment, context, and strategic framing that only a human can provide. Skipping review is how you end up with technically correct but strategically wrong requirements.

3. Be Specific in Your Notes

"Users want better reports" gives the AI very little to work with. "Users want to export dashboard data as CSV with custom date ranges and department-level filtering" gives it everything it needs.

4. Feed It Contradictions

Real discovery produces conflicting requirements. Don't resolve contradictions before feeding them to the AI — let it surface the conflicts explicitly so you can make a conscious decision about which direction to go.

5. Use the Completeness Check Every Time

Even experienced PMs miss sections. The completeness check catches blind spots consistently. It takes 5 minutes and prevents 5 hours of back-and-forth with engineering.

The Death of the Blank Page

The blank PRD document is a relic. The question isn't "can I write this requirementa?" It's "given all the context I've already gathered, can an AI organize it so I can focus on the decisions that matter?"

The answer is yes. And it takes 90 minutes.

The Intelligent PRD Co-Pilot is coming to the Jasper Toolkit. Sign up to be notified when it launches.