Gokul Rajaram joins SIERRA's Vignesh Ravikumar on the UNPAK3D Podcast to lay out what separates durable AI companies from the noise, drawing on his operator background at Google, Facebook, Square, and DoorDash, as well as his experience as an early investor.

His core lens is simple: winning companies start with obsessive founders, ship products that deliver fast, provable ROI, and build defensibility in a world where the big platforms keep moving up the stack.

Gokul points to three capability shifts unlocking the next wave of AI applications: materially better document processing, more reliable browser automation for legacy workflows, and voice plus conversational agents that can drive end-to-end tasks. While bullish on application-layer and security opportunities, he’s cautious on AI infrastructure and middleware, arguing the model providers are steadily absorbing those layers and that infra companies increasingly need consumer-grade product and brand muscles, not just developer adoption.

For application companies, he argues defensibility comes from domain-specific evaluation and a proprietary reward function that enables a self-improving system through reinforcement learning, paired with workflows and data advantages that are too niche for horizontal platforms to prioritize. The playbook to win enterprise spend is to start with a narrow, high-value use case, prove value in days or weeks, and expand in concentric circles rather than “boiling the ocean.”

Rajaram also pushes a new operating rule for the AI era: “compound or die.” Once a first product finds pull, companies should build additional products earlier than the old SaaS playbook would suggest, using a broader product suite to deepen customer relationships and block competitors. He emphasizes retention and expansion over topline growth, with usage as an even earlier signal than revenue, and discusses how both startups and incumbents should handle data access, privacy expectations, and the shift toward outcome-based pricing.

On talent, he describes the rise of AI-native execution norms: smaller teams moving faster, a premium on builders across functions, and increasing expectations for technical depth. He highlights emerging roles like AI ops that build internal automation for GTM and other functions, and he views forward-deployed engineers as necessary in the current “solution sale” phase until products become easier to deploy. He closes with a practical personal habit for staying effective: keeping the mind uncluttered by using a “second brain” system to capture everything quickly.

Listen on Apple Podcasts