Snowflake convenes 350 CTOs to share lessons on building AI-native engineering teams

Snowflake Summit 2026 in San Francisco hosted the inaugural CTO Circle, a convening of more than 350 CTOs from financial services, telecommunications, retail, and technology sectors. The event was created to provide engineering leaders a trusted forum to learn from peers navigating the transition to AI-native organizations, moving beyond the individual company conversations that competitive pressures and rapidly evolving technology typically keep siloed.

Vivek Raghunathan, SVP of Engineering at Snowflake, opened the discussion with a fundamental reframing: treating developer productivity not as an engineering or culture problem but as a product. The question "What if you treated your developers like customers?" became the foundation for Snowflake's transformation. Rather than assuming leadership knew what engineers needed, the company applied the same product management principles used for customer-facing products—interviewing developers to map friction points, establishing baseline metrics, and running experiments to measure impact. The approach combined executive sponsorship with bottom-up adoption, ensuring improvements reflected how engineers actually worked. The results were concrete: in 18 months, Snowflake increased its internal developer Net Promoter Score by more than 30 points, achieving a 4:1 ratio of satisfied to dissatisfied developers. Crucially, that improvement translated into faster, more adaptive software delivery.

Raghunathan outlined three stages in the journey from AI adoption to organizational mastery. Adoption occurs when developers begin using AI tools daily. Mastery develops as engineers discover repeatable workflows—prompting techniques, planning methods, debugging approaches—that consistently deliver better outcomes. Optimization happens when those patterns become institutional knowledge shared across the organization. Early AI adopters at Snowflake experimented internally; successful patterns were documented and made available to all engineers. The competitive advantage came not from deploying another AI assistant but from institutionalizing successful ways of working. This shift changes how organizations think about software development itself: as AI reduces the effort to transform ideas into code, product managers can prototype, designers can validate concepts directly, and domain experts can add guardrails. Code becomes the fastest way to test assumptions.

Jon McNeill, author of The Algorithm, urged leaders to reverse the typical approach to AI adoption. Rather than beginning with technology and searching for applications, successful organizations identify the few business constraints that matter most and redesign engineering systems around solving those problems. Success, he argued, will be measured not by who generates the most code or consumes the most tokens but by who builds the simplest, fastest, most effective engineering system—particularly by who eliminates the most unnecessary steps between idea and production.

Jeremy Burton, General Manager of Observability at Snowflake, highlighted a critical operational challenge: as AI becomes embedded throughout the software development lifecycle, organizations must invest in architectures that balance velocity with reliability. Early AI experimentation is giving way to production deployment, where measurable business value is expected. AI agents generate more telemetry, interact with more systems, and make decisions using information distributed across complex environments. Burton challenged the assumption that better models simply need better data; in reality, AI is only as effective as the context it can access. That context includes semantics (what data means), relationships (ontologies and knowledge graphs), and business context. Standardized interfaces—APIs, CLIs, Model Context Protocol (MCP)—allow AI agents to reliably retrieve and act on information, while intuitive interfaces let engineers validate AI-generated insights. Organizations storing logs, metrics, traces, and data across disconnected systems handicap AI's ability to reason accurately about production.

Aditya Gaur, Engineering Manager at Netflix, demonstrated this principle through Netflix's automated root cause analysis. Although often described as an AI initiative, its success depended far more on data architecture. Years before introducing AI agents, Netflix invested in connecting fragmented telemetry, modeling operational relationships through ontology and knowledge graphs, and creating a shared context layer. By the time AI entered the picture, the foundation existed. AI agents could reason over structured operational knowledge rather than searching disconnected logs, generating far more meaningful hypotheses during incidents.

Caitlin Colgrove, CTO at Hex, urged leaders to commit fully to AI nativity, describing it as "burning the boats." Hex initially created a dedicated AI product team, but the centralized approach created a bottleneck. The company disbanded it and distributed responsibility across every product team. Today, Hex delivers product rapidly and embeds AI throughout because ownership lives with engineers closest to the customer problem. Chris Kozlowski, Managing Director at Barclays, brought an enterprise perspective: speed creates value only when matched with governance and trust. In regulated environments, engineering teams cannot choose between innovation and control. Corey Burke (SVP Engineering at Dialpad) and Arun Rajamanickam (VP Engineering at project44) described how AI shortens the path from identifying customer problems to validating solutions—provided teams have infrastructure, shared context, and operational discipline to iterate confidently. As AI agents implement significant portions of features independently, engineers spend less time writing code and more time defining intent, orchestrating agents, and validating outcomes.

Qi Jin, EVP at Cerebras Systems, observed that organizations optimized for previous generations of software development often become the biggest barrier to AI adoption, as established processes and organizational boundaries were designed to scale proven ways of working, not to continuously adapt. The gathering converged on a conclusion: building AI-native engineering organizations requires more than tools—it demands rethinking team structure, leadership models, and where engineers create the most value.