Intuit scrapped its own AI agent architecture twice in four months. At VB Transform 2026, its AI VP called that the fast path

At VB Transform 2026, Intuit VP of AI Nhung Ho said the company twice scrapped its agent architecture in about four months, moving from specialist agents to orchestration and then to a skills-and-tools model. The second rebuild took 60 days after Intuit found that natural-language handoffs between agents caused context loss and compounding errors.

Intuit’s experience shows why early agent systems are shifting away from rigid multi-agent chains toward more modular architectures. The broader trend is that production AI teams are prioritizing observability, fast iteration, and simpler execution paths over elaborate orchestration layers that can become failure points.