Intuit has undertaken a major technical overhaul of its AI infrastructure, moving away from a broad-capability multi-agent system to a granular, skill-and-tool-based architecture. The company embedded human experts directly into workflows alongside AI to meet rising demand for complex, agentic tasks that legacy systems could not handle.
According to Nhung Ho, VP of AI at Intuit, the transformation involved decomposing large agents into specialized components — effectively separating the “brain” from the “hands.” The company changed its orchestrator, planner, and the foundational “brain” of the system, while also shifting what every team across the organization builds.
The move reflects a broader industry trend: customer expectations have evolved from simple conversational interactions to sophisticated, multi-step AI tasks. Intuit’s new abstraction layer aims to support faster, more reliable execution of these complex workflows without sacrificing performance or accuracy.
Ho is scheduled to detail the technology decisions behind this infrastructure rebuild at VB Transform 2026 on July 14 and 15. Her talk will focus on the abstraction layer that underpins Intuit’s “system of intelligence” and how this architecture has enabled the company to handle increasingly complex agentic demands.
The counterargument: Some observers may question whether such a dramatic architectural shift introduces new operational risks, including integration complexity and potential downtime during the transition. If not executed carefully, the move could create fragmentation rather than cohesion, especially across a large organization like Intuit.