AI Agent Deployment and Real ROI with MongoDB Field CTO Pete Johnson

In this episode of the Shift AI Podcast, Pete Johnson, Field CTO of AI at MongoDB, joins host Boaz Ashkenazy for a wide-ranging conversation on where AI agents are actually delivering ROI — and what still needs to happen before enterprises can trust them with customers.Pete shares his origin story as a self-taught programmer who got his start on a TRS-80 in 1981, traces how MongoDB was born into a world of cloud, mobile, and internet that relational databases were never designed for, and explains why vector search sits at the intersection of MongoDB's document model and modern AI use cases.The conversation digs into Pete's "Customer Agent AI World Tour," where he has met with enterprises in over a dozen cities and heard a consistent message: production-grade agents are real, ROI is measurable, but the deployments are employee-facing and human in the loop. Pete explains the three things blocking the jump to customer-facing agents at scale, governance, observability, and evaluations, and why that challenge mirrors the early days of HTTPS standards for e-commerce.Boaz and Pete also explore the growing conversation around sovereign AI and on-prem inference, why Apple's edge device ecosystem may be the quiet wildcard in the infrastructure debate, and how MongoDB's Atlas platform lets organizations deploy data across 125-plus hyperscaler data centers worldwide.The episode closes with a forward-looking discussion on the future of software engineering, Werner Vogels' five skills for tomorrow's engineer, and why Pete's two-word forecast for the future of work is "not doomsday" — backed by a compelling contrast between the bank teller and the tollbooth worker as a framework for thinking about automation and job transformation.This episode is essential listening for enterprise leaders, developers, and anyone thinking seriously about where agentic AI is today versus where it is headed.Chapters[00:00] From Intellivision to TRS-80: Pete's Tech Origin Story[03:13] What MongoDB Is and Why the Document Model Matters[06:23] Joining MongoDB and the Vector Search Opportunity[07:50] What Pete Is Hearing on His AI World Tour[09:06] Why Fortune 500s Start with Employee-Facing Agents[11:11] Security, Governance, and the Three Big Blockers to Customer-Facing AI[14:29] How Software Engineering Is the Canary in the Coal Mine[16:56] Sovereign AI, On-Prem Inference, and the Cost of Tokens[20:14] The Apple Edge Device Wildcard[21:19] How MongoDB's Atlas Platform Fits a Hybrid Cloud World[23:21] Using AI Agents Is Programming in English[25:50] Werner Vogels on the Five Skills Every Engineer Will Need[27:17] The Future in Two Words: Not Doomsday[28:08] Bank Tellers vs. Tollbooth Workers: Why Most Jobs Will Level UpConnect with Pete JohnsonLinkedIn: https://www.linkedin.com/in/petecj2/Connect with Boaz AshkenazyLinkedIn: https://www.linkedin.com/in/boazashkenazy/Email: [email protected]