How NotebookLM Changed My Approach to Technical Reviews

The Context Problem

A typical design review at Google involves a 30-50 page design doc, references to multiple related docs, a codebase, and months of historical discussion in comments and threads. As a TPM, I need to understand all of this context well enough to ask the right questions — not the surface-level questions that waste everyone's time, but the deep questions that uncover risks and assumptions.

That's where NotebookLM became indispensable.

My Workflow

Step 1: Upload Everything

Before a major review, I upload the design doc, related PRDs, previous meeting notes, and any relevant post-mortems into a NotebookLM notebook. This gives the model full context about the project.

Step 2: Ask Structural Questions

I start with questions like:

• "What are the key assumptions in this design?"
• "What dependencies does this create on other teams?"
• "What's the rollback strategy if this fails?"
• "How does this compare to the approach we considered in [previous doc]?"

Step 3: Identify Gaps

The most valuable output isn't what NotebookLM tells me — it's what it can't tell me. When the model says "I don't see this addressed in the documents," that's often the most important question to bring to the review.

Step 4: Generate My Review Notes

I ask NotebookLM to create a structured set of review comments organized by theme: architecture, scalability, testing, dependencies, and timeline. I then edit these into my own voice and add my specific concerns.

Related: see The TPM's Guide to Evaluating AI Tools Without the Hype and Building Android at Scale: What Managing Camera & Media Taught Me About AI Integration.

The Result

My review prep time dropped from 3-4 hours to about 45 minutes. But more importantly, the quality of my reviews improved because I'm no longer spending cognitive energy on comprehension and can focus entirely on analysis and judgment.

The engineers I work with have noticed. One tech lead told me: "Your review comments are the most useful because you clearly read everything." I didn't correct them.