Intuit Rebuilt Its AI Agent Architecture Twice in Four Months — and Called It Progress

Intuit, the company behind TurboTax, QuickBooks, and Mailchimp, has revealed that it completely scrapped and rebuilt its AI agent architecture twice in roughly four months. Rather than framing it as a setback, Nhung Ho, Intuit's VP of AI, described the cycle of demolition and rebuilding as the fastest path forward at VB Transform 2026.

The Architecture That Kept Breaking

Intuit has been building toward an AI-driven platform since 2019. Its GenOS platform, anchored by a component called GenRuntime, functions as the operating system for its agents — receiving data, reasoning over it, and determining what action to surface to the customer.

The first rebuild moved the company from a fleet of large specialist agents to a central orchestration layer. That design held for about three months before structural problems forced another overhaul. The culprit was error compounding.

When agents pass outcomes to one another in natural language, each downstream agent must infer how the upstream agent reached its conclusion. That inference degrades with every additional handoff. As Ho put it bluntly: "If you have 10 agents and they all are passing to each other, every time that pass happens, error compounds." A 10-agent chain does not fail occasionally — it fails by design.

From Agents to Skills

The second rebuild decomposed large agents into granular skills and tools. Rather than each agent owning a broad capability, shared skills now run across every part of the product that needs them. The rebuild also decoupled Intuit's orchestration from specific model providers, letting the company swap in different models without rebuilding the entire stack.

The payoff has been substantial. TurboTax users with AI agent support now complete their returns 12% faster, with nearly half finishing in under an hour. QuickBooks Online customers get paid an average of five days sooner, and overdue invoices are 10% more likely to be paid in full. Intuit's Business Tax Agent helped 58% of small businesses uncover deductions they had previously missed.

Humans in the Loop, by Design

When Intuit rolled out AI agents to 3 million customers, 85% came back. The company credits that retention to pairing AI with human expertise — not replacing one with the other. Customers can bring in an Intuit support person, their own accountant, or an Intuit bookkeeper mid-conversation, and that person joins with full context of what the agent has already done.

The architecture shift also transformed how feedback works. Ho said Intuit went from roughly 0.3% of customers ever providing explicit feedback to nearly 100% with the new system. That volume of signal drives evaluations, which have become the primary measure of whether the system is performing.

Key takeaway: Intuit's experience delivers a clear warning for any company building AI agent systems at scale: orchestration layers that look clean in design can become brittle under real workloads. Decomposing agents into shared, reusable skills — and keeping humans firmly in the loop — is the difference between a demo and a production system that 85% of users choose to return to.