The Person Is the Product
Find one person with an extraordinary workflow, shadow them, then compress their expertise into software
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 93%
Instead of researching an abstract user need, Martin embedded a single world-class practitioner of the target skill on the team, watched how he worked, measured how long his craft took him, and made compressing that time the product's north-star metric. The product's job became giving ordinary people the output of an extreme workflow without the extreme workflow.
Origin
Raiza Martin's approach to working with Steven Johnson — a 14-book New York Times bestselling author and journalist who joined the NotebookLM team as a non-engineer, non-PM, non-designer. Martin's own framing: 'Stephen, I think you're the product.'
Core principles
- 01Expertise you can watch daily beats expertise you interview quarterly.
- 02The product goal is to densify information the way an expert does, for people who are not the expert.
- 03Time-to-expert-output is the metric: watch how long the craft takes them, then crunch it down.
- 04The extreme workflow (8,000 highlighted quotes in Readwise) reveals the mechanism; the normie workflow (crumpled Post-its in a pocket) reveals the market.
- 05The embedded expert is also a sparring partner — disagreement is fine as long as you end aligned on the next step.
How to run it
- 1
Name the human capability your product replaces
Identify the specific cognitive craft your product is trying to democratise. For NotebookLM it was densifying information — turning large amounts of source material into shareable understanding.
Watch out Pick a craft, not a task. 'Summarising' is a task; 'making knowledge relatable' is a craft with an observable master.
- 2
Recruit a world-class practitioner onto the team, not into a research panel
Bring someone who is genuinely elite at that craft into daily working proximity — sitting with you, in the meetings, arguing about ideas. Johnson was not an engineer, PM or designer; his role was to be the exemplar and idea partner.
Pro tip Ideas people compound: Martin's routine with Johnson was a 'crazy idea of the day' that he would riff on and turn into 'how would people actually do this?'
Watch out Most PMs reject this as a wasted headcount — 'another chef in my kitchen'. That rejection is the reason the pattern is rare and valuable.
- 3
Shadow the workflow and instrument the time
Watch how the expert reads, annotates, connects and researches. Record how much time each stage of the craft costs them. Their elapsed time becomes the benchmark you are trying to collapse.
Pro tip Look for the workflow so extreme that no normal person would ever do it — that is where the mechanism worth automating lives.
- 4
Make time-compression the product metric
Set an explicit goal of crunching the expert's hours into a button. Ship the mechanism (source-grounded synthesis, an engaging conversational format) rather than the expert's tooling.
Watch out Do not ship the expert's tools to normal people; ship the expert's outcome. Nobody wants a Readwise with 8,000 quotes.
- 5
Extend the shadowing to your real users
Apply the same method beyond the in-house expert: sit with users for meaningful periods, regularly and intentionally. For NotebookLM this meant following students around, watching them do homework and study, and asking how studying makes them feel.
Pro tip Ask about feelings, not just steps — emotional friction is what the delightful format was designed to remove.
Watch out One-off user interviews do not produce this class of insight; the differentiator is regular, sustained, intentional co-presence.
In the wild
Martin told Johnson outright that he was the product, and that she would follow him around and watch everything he did — how he thought about language, information and knowledge, and how he researched his books — then figure out how to build technology that reproduced it.
→ The observations fed the Content Studio and the Deep Dive format: an opinionated, relatable synthesis of dense source material that any user gets in one click.
Beyond Johnson, the team sat with students, watching them study and asking how they felt while studying, rather than running one-off interviews.
→ Educators and learners became the product's first loyal demographic before professionals and enterprises followed.
Common mistakes
Treating the expert as a consultant instead of a teammate
The value came from daily proximity — watching the craft in motion, arguing about ideas, seeing the unglamorous parts. A scheduled advisory call surfaces the expert's self-report, not their actual workflow.
Copying the expert's tools rather than their outcome
Johnson's 8,000-quote Readwise flow is unreproducible for ordinary people. Building a better highlight manager would have missed the point; the product had to deliver the synthesis without the discipline.
Avoiding disagreement with your embedded expert
Martin and Johnson clashed often. The discipline that made it productive was ending each clash aligned on the next step even when they still disagreed on the substance — disagreement with no resolution is the actual failure mode.
Is it for you?
Best for
Product teams building AI tools that automate a knowledge craft, who can secure sustained access to one genuinely elite practitioner of that craft
Not ideal for
Commodity or infrastructure products where no single human exemplar of the outcome exists, or teams that cannot afford a non-shipping headcount
From the transcript
“maybe I watch Stephen and I look at the way that he does these things and I look at how much time it takes for…”
“I told him this I was like Stephen I think you're the product I think it's you and I'm going to follow you around I'm…”
“so I I I learned a lot just from watching Stephen of like his craft and thinking about okay how do I make people really…”
“with students just follow students around watch them do homework watch them study talk to them about how they feel when they study”
From the episode
Behind the product: NotebookLM
Raiza Martin (Senior Product Manager, AI @ Google Labs)