CTO Roundtable: Engineering Leaders Share Lessons on Building AI-Native Organizations

At the first CTO Circle held during Snowflake Summit 2026 in San Francisco, more than 350 CTOs from finance, telecom, retail, and technology sectors gathered to exchange experiences on building AI-native engineering organizations. The discussions converged on three themes: getting AI into production, balancing speed with risk, and redesigning engineering teams for AI. Snowflake's SVP of Engineering Vivek Raghunathan shared how the company treated developer productivity as a product, interviewing developers, mapping friction points, and establishing baseline metrics. Within 18 months, Snowflake's internal developer Net Promoter Score rose by over 30 points, with a 4:1 ratio of satisfied to dissatisfied developers. He emphasized that high leverage comes from deep usage and mastery, moving through stages of adoption, mastery, and optimization, and institutionalizing successful workflows. Jon McNeill, author of 'The Algorithm', urged leaders to reverse the typical approach: start with key business constraints and then rebuild engineering systems around them, measuring success not by code volume but by efficiency. Jeremy Burton, GM of Snowflake's Observability Business Unit, noted the shift from experiments to production, requiring proof of business value. AI agents generate more telemetry and demand more context, so observability must evolve to provide semantic layers, ontologies, and knowledge graphs, as well as standardized interfaces like APIs, CLIs, and Model Context Protocol (MCP). Netflix engineering manager Aditya Gaur explained that their root cause analysis automation succeeded due to data architecture investments made years before AI agents were introduced. Hex CTO Caitlin Colgrove discussed 'burning the boats' by eliminating its centralized AI team and distributing ownership to product teams, which accelerated delivery. Barclays MD Chris Kozlowski stressed that speed must be paired with governance and trust, while Dialpad's SVP Corey Burke and project44's VP Arun Rajamanickam described how AI agents reduce time spent on coding and increase time spent on defining intent and orchestrating agents. Cerebras EVP Qi Jin introduced 'Code Yellow', a framework for urgent organizational change focused on customer problems. Capital One's SVP Parvez Naqvi observed that as AI speeds implementation, engineering value shifts to judgment, planning, and architecture. CodeRabbit CEO Harjot Gill and Cursor COO Jordan Topoleski highlighted new bottlenecks in code review and quality assurance. The overarching conclusion was that AI-native engineering hinges on an organization's ability to learn, institutionalize, and continuously adapt, rather than on early tool adoption.