Intuit’s latest AI breakthrough isn’t about faster chatbots—it’s about enabling systems that can autonomously complete intricate business tasks. At the upcoming VB Transform 2026 conference, the company will unveil how it rebuilt its AI infrastructure from the ground up to support these advanced agentic workflows.
Customer priorities have shifted dramatically in recent years. Where once simple, rapid responses sufficed, today’s users demand AI agents capable of performing multifaceted operations like financial forecasting, tax filing automation, or inventory management—tasks that legacy architectures simply cannot support. Intuit recognized this gap and responded with a comprehensive overhaul of its technical framework.
From monolithic agents to modular, skill-based systems
The company’s previous approach relied on large, multi-purpose AI agents designed to handle broad capabilities. This strategy proved inefficient for complex workflows, prompting Intuit to transition to a more granular architecture. The new system breaks down agents into specialized components, separating decision-making logic from execution tools—a strategy Nhung Ho, VP of AI at Intuit, describes as "decomposing the brain from the hands."
"We moved away from a system where massive agents performed numerous functions toward one where workflows, skills, and tools operate at the foundational level," Ho explained. This transformation required rebuilding every layer of the company’s AI stack, including the orchestrator, planner, and underlying decision-making frameworks. Crucially, human experts now play an integral role within these automated workflows, ensuring accuracy and adaptability.
Building an abstraction layer for flexibility and scalability
A key innovation in Intuit’s new architecture is the introduction of an abstraction layer that decouples orchestration from specific AI model providers. This design allows the company to integrate the most suitable models—whether from leading third-party providers or its own proprietary tools—without overhauling its entire system.
At VB Transform 2026, Ho will detail how this abstraction layer enables Intuit to:
- - Dynamically select AI models based on task requirements rather than locking into a single provider
- - Scale workflows efficiently by reusing modular components across different products
- - Maintain agility by updating individual elements without disrupting the entire system
This approach mirrors strategies adopted by other industry leaders, as evidenced by the conference’s diverse speaker lineup. Sessions will explore similar agentic AI implementations at Target, Instacart, Asana, Rivian, and Atlassian, highlighting a growing industry trend toward specialized, modular AI architectures.
The future of agentic AI in enterprise workflows
Intuit’s infrastructure redesign reflects a broader industry movement toward AI systems that can autonomously manage complex business processes. By shifting from generalized agents to skill-based architectures—and embedding human oversight where needed—the company has created a scalable foundation for next-generation AI applications.
The implications extend beyond financial services. As companies across sectors seek to automate multi-step workflows, architectures like Intuit’s could become the standard for enterprise AI. Whether optimizing supply chains, personalizing customer interactions, or automating back-office operations, the ability to build flexible, provider-agnostic AI systems will likely define the next phase of enterprise technology adoption.
AI summary
Müşteri beklentilerindeki değişimle birlikte Intuit, AI altyapısını baştan inşa etti. Bu dönüşümün ardındaki stratejiler ve sonuçlarını keşfedin.
