@Sid Arora, while reviewing my notes from the AI PRD, I noticed in your example there isn't specific call outs of functional and non-functional requirements nor functional and emotional jobs (e.g. JTBD). I'm wondering why these elements from traditional PRDs don't get a lot more coverage in the AI PRD: Audio Journal.

@Sid Arora this makes sense. Will keep this in mind. BTW, I think it would be great to incorporate a PRD example for a B2B product. I view the Audio Journal example as a good consumer or Pro consumer use case. Thanks for a great course and excellent learning experience.

Here is my submission. This was a thoughtful and challenging activity.

here are my ten test cases. this was an insightful exercise.

Week 2 Activity - Love the detailed nature of this activity. Nice way to connect the learning with actual practice. Good insights.

Good Quality Control (What does “good quality” actually mean for this summarizer?) 1. Source Fidelity (Truth Over Fluency). Every output must be traceable back to the transcript. No invented concepts, no “helpful” additions, no synthesized frameworks. Fluency is secondary to faithfulness. If the speaker didn’t say it, it doesn’t exist. Quality starts with strict grounding. 2. Concept Extraction (Frameworks Beat Filler). The summarizer must prioritize mental models, decisions, and product insights over conversational noise. Logistics and small talk should disappear. Core frameworks (like deterministic vs. probabilistic) must surface clearly. If filler survives but concepts don’t, extraction failed. 3. Applied Insight (From Notes to Leverage). A strong summary converts ideas into practical implications. It should answer: What changes for me as a PM? If users finish reading without clearer direction on discovery, design, or measurement, the summarizer produced notes, not value. 4. Cognitive Efficiency (Designed for Two-Minute Recall). Formatting is part of quality. Use structured sections, crisp bullets, and scannable layouts so users can refresh the entire session in under two minutes. The goal isn’t completeness—it’s fast comprehension. Measuring Success: 1. Friction Index (How Much Did Users Have to Fix?). Measure how much users modify the output before keeping or sharing it. Minimal edits mean the summary matched intent. Heavy rewrites or deletions signal quality gaps. Low friction = real time saved. 2. Reuse Signal (Did It Become Working Material?). Track whether users copy sections into docs, notes, or follow-ups. When content leaves the product and shows up in real workflows, that’s stronger than any rating—it proves usefulness. 3. Steerability Rate (Can Users Recover Quickly?). Measure how often users successfully improve a weak result using regenerate, focus, or refinement controls. If users can course-correct and accept the next output, your recovery UX is doing its job. 4. Reference Benchmarking (Are We Matching Expert Output?). Maintain a small set of expert-written “reference summaries.” Regularly score AI outputs against them on coverage, accuracy, and actionability. This gives you a concrete baseline to track real improvement over time.