AI Pat Works Here Now: Why Agents Must Follow Human Rules with Pat Casey // CTO @ ServiceNow

Pat Casey joined ServiceNow in 2005 as the first engineer besides founder Fred Luddy, when the product was called Glide, and the office sat above a restaurant. Twenty years later, he's CTO of a $13B revenue company serving 85% of the Fortune 500, and until recently ran all 10,000 engineering employees. In this episode, Tobi gets a rare look inside one of the longest founder-era-to-public-company CTO runs in enterprise software.

Pat walks through the architecture behind ServiceNow's contrarian single-tenant bet 90,000 databases, 25B+ queries an hour, all managed by automation; why they bought a 15-person German startup to build RaptorDB on a Postgres fork; and why breaking apart a 20-year-old monolith is always harder than senior engineers assume.

On AI, Pat brings actual numbers: 7,000 Windsurf licenses bought a measured 15% productivity bump, but the real story is the small subset of engineers going 5–6x. His agent philosophy is the episode's core insight: create a user called "AI Pat," assign it work, and make it follow the same rules as humans, because you should not trust an LLM more than you trust a human being.

Key topics:

• Nerd path: Atari 400 → floppy-disk jockey at Aldus → first engineer at ServiceNow
• Scaling from a stuffed fish on a monitor to 10,000 engineers
• Single-tenant architecture, 90,000 databases, and RaptorDB
• Real AI coding data: Windsurf, Claude Code, 15% average vs 5x outliers
• "AI Pat" and the anthropomorphic model for enterprise agents
• Why SaaS incumbents may win the AI era,  and Pat's advice to CTOs