Lenny's Podcast · Nov 14Becoming an AI PM | Aman Khan (Arize AI, ex-Spotify, Apple, Cruise)
Aman's company started with ranking, regression and classification models, then moved toward LLMs. It automated internal data analysis instead of a chatbot, ran a hackathon where a support Slackbot failed for lack of context, and live-streams AI product teardowns every week or two.
Tips
- Do not copy the chatbot interface; ask what interface the problem needs▶ 31:41Arize automated internal data analysis and shipped just a small chatbot to understand how the technology worked.
- Run hackathons with a list of problems to learn where AI sticks▶ 37:29In Arize's hackathon, a Slackbot to route support questions to the right on-call person turned out hard because of missing context about who was working on what.
- Walk and chew gum: ship customer value while making room for AI exploration▶ 43:39Arize did a NotebookLM teardown webinar and plans to repeat it every week or two.
- Get into the details yourself to earn trust when the team is uncertain▶ 51:11He maxed out his five LinkedIn messages a day while the team was unsure which direction to take.
- Wander instead of waiting when the product direction is unclear▶ 59:09Arize's chief product officer Aparna told him that not being sure what the right decision is is what building zero to one really feels like.
- Understand the three types of AI PM before choosing a path▶ 6:56
