Gokul Rajaram - Lessons from Investing in 700 Companies - [Invest Like the Best, EP.456]

Episode Summary

Executive Summary: Patrick O’Shaughnessy interviews Gokul Rajaram on how AI is reshaping product building, company durability, hiring, ads, and leadership. Rajaram argues judgment, outcomes, and control of scarce assets will matter most as software becomes easier to create but harder to defend.

Main Topics: AI changes product development (Priority: 10/5): AI agents make product building bottoms-up, iterative, and far more hands-on for PMs. Judgment as the durable moat (Priority: 10/5): As code generation explodes, human judgment becomes the key scarce skill. Defensibility in AI software (Priority: 9/5): Durable products need data, network effects, hardware, regulation, or control points. Legacy software under pressure (Priority: 9/5): Seat-based and low-data-half-life software is most exposed to AI-native attack. Ads economics and platform risk (Priority: 8/5): Rajaram says only three ad models work, and agentic interfaces threaten ad engagement. Leadership, communication, and design (Priority: 8/5): Great leaders edit, simplify, and communicate clearly through recurring rituals. Hiring and career advice in AI (Priority: 7/5): He favors builders, long tenure, work projects, and managing AI agents over layers.

Key Arguments: PMs must be hands-on; product is now built bottoms-up with engineers, researchers, and design. Judgment is future-proof because AI slop makes choosing what to build and ship the hard part. Durability comes from scarce assets, control points, hardware, essential workflows, or network effects. Seat-based software like Zendesk is more exposed than system-of-record products like NetSuite. Agent companies must build migration tools and often full platforms, not just workflows. Only three ad business models work: own coveted users, drive outcomes, or serve exclusive demand. Great leaders simplify relentlessly; Jack Dorsey framed PMs as 'product editors.' The best hires are doers who can build AI-agent workflows, not managers with small spans of control. Self-serve products outperform because they scale widely and reveal unexpected product usage.

Data Points: Expense reviews automated by Ramp: 85% - Ramp claim in sponsorship copy Expense review accuracy: 99% - Ramp claim in sponsorship copy Company savings with Ramp: 5% - Ramp claim in sponsorship copy Timeframe of major product-development shift: December and January; December 25th and Jan 26th - Rajaram says the shift became clear over recent months AdSense share goal: less than 1% - Larry Page wanted Google involved in every ad on the internet, not just a large business Gmail storage at launch: one gigabyte - Used to illustrate Google’s technical ambition versus Yahoo Mail’s 10MB Yahoo Mail storage: 10 megabytes - Contrast with Gmail’s 1GB launch offer Workforce ratios: 1 to 20 - Rajaram says the designer/PM-to-engineer ratio is moving toward far more engineers per non-engineer Job tenure advice: minimum three to four years - He says that is needed to have real impact at a company Board/board-buddy cadence: once a month - He recommends board buddies meeting with management between board meetings DoorDash acquisition challenge: $10 or $20 - Tony Xu’s interview exercise to test whether candidates can acquire customers DoorDash acquisition target: 1,000 customers - Candidate exercise mentioned in the interview Ad engagement guardrail: X percent dip in engagement overall - Example of an engagement budget used to constrain monetization Google strategy review artifact: only images - Eric Schmidt asked for a strategy presentation with no words, only images Square onboarding threshold: 90 plus percent - Square accepted most applicants at the transaction level while managing risk in real time

Pivotal Quotes: "the notion of a long horizon, a long-running agent" — Gokul Rajaram: Describing the biggest change in product development under AI "Judgment is the number one thing that humans are going to bring." — Gokul Rajaram: On what remains future-proof in an era of AI-generated code "You need to lead with what is the outcome you can deliver or ideally even have delivered." — Gokul Rajaram: Advice for founders selling AI products and enterprise software

Implications: AI-native winners will likely combine durable data/control points with exceptional judgment; founders should prove outcomes early and build migration paths before legacy platforms shut them out.