How Intuit is designing embedded AI agents in QuickBooks to serve SMBs

Running a small and medium-sized business (SMB) means making a hundred decisions before lunch and dealing with the fallout if one goes sideways. The pressure isn’t just about doing more with less; it’s about doing it faster, with precision, and at scale. 

Traditional automation has promised relief for years, but most SMB-focused tools still leave too much on the table. While they work well for rule-based, repetitive tasks, they tend to break down when faced with edge cases, unexpected inputs, or workflows that fall outside predefined structures. In addition, these tools tend to silo key data away from where real decisions happen. 

That’s where Intuit has stepped in by bringing its embedded AI agents into QuickBooks. These agents assist businesses with important day-to-day functions like payments, accounting, finance, and customer support.

Whether it’s the Payments Agent, which identifies non-traditional invoicing patterns and adapts reminder schedules accordingly, the Accounting Agent, which avoids misclassifying outlier transactions by flagging them for review, or the Customer Agent managing leads and follow-ups without the daily chase, the focus is the same: less overhead, more momentum. 

I sat down with Ashok Srivastava, Intuit’s Chief Data Officer, to explore how Intuit’s AI agents are engineered to maintain a balance between automation and human oversight, while adapting to the needs of scaling businesses.

AI agents now power every corner of Intuit’s product suite

The new AI agents are developed specifically for QuickBooks, but similar capabilities have been extended across other areas of Intuit’s business.

“The new agents we just announced are currently available on QuickBooks, but were built using the power of the Intuit platform across multiple products,” says Srivastava. “The underlying agentic architecture and our integrated data platform on which these agents are built are shared across the entire Intuit business.” 

In early 2023, Intuit launched the Intuit Generative AI Operating System (GenOS), on which all of its agentic and AI technologies are built and operated. The firm has already deployed AI agents in TurboTax to automate the annual conversion of US and state tax code updates. 

“We’re thinking holistically about Intuit customers and applying the same technology to other products in domain-specific ways,” notes Srivastava. “Our goal is to build specialist agents tailored to each context to do the work for our customers with their approval.”

Intuit’s transformation into an AI-driven expert platform began six years ago, with a focus on combining machine intelligence and human expertise to simplify customer workflows, improve financial outcomes, and enable confident decision-making.

A decade of building and investment in platform infrastructure has enabled Intuit to integrate key services like accounting, payments, taxes, and marketing into unified, AI-native experiences grounded in clean, structured data for its customers. 

As a result, AI now underpins every product — TurboTax, Credit Karma, QuickBooks, and Mailchimp – in the Intuit ecosystem.

The foundation of Intuit’s agentic AI design hinged on a defining question…

Intuit’s AI agents promise to automate QuickBooks’ core business tasks. So, how does the firm ensure that automation doesn’t erode human oversight, especially in critical financial decisions?

“This is one of the most important design questions we faced,” notes Srivastava. “We’re not trying to remove humans from critical decisions. Instead, our approach is about reducing the manual effort that bogs people down, while keeping them in control at every critical decision point.”

Each agent is designed to complete workflows proactively but also to “pause” where human review is essential. For example, the Payments Agent predicts late payments and drafts reminders, but the business owner can review and customize them before sending. Similarly, the Finance Agent can generate scenario forecasts or KPI dashboards; however, it’s up to the owner or accountant to decide on the actions to take.

“We believe that AI should amplify human decision-making, not replace it, especially when dealing with people’s hard-earned money,” adds Srivastava. That’s why real-world experts can pick up where the automation leaves off to answer tough questions and provide accuracy and guidance for businesses. 

Intuit’s GenOS acts as the connective tissue between Intuit’s AI agents and the human expertise that powers the platform. Srivastava explains that the tech infrastructure includes built-in safeguards to address agentic AI risks. For decisions requiring closer examination, there’s a live channel that links directly to Intuit’s human experts and developers, providing an added layer of human oversight and validation.

How can these agents support long-term business evolution?

One of the central questions regarding these QuickBooks agents is their potential to support growing SMBs. Do they evolve intelligently as the business scales, or do they rely on manual reconfiguration to meet the changing needs of SMBs? 

Srivastava notes it’s about striking a balance between adaptability and control. 

The agents are designed to learn and adapt over time, becoming more efficient and helpful as customers use them. Each of these AI agents has specialized training in its domain, using agentic technology to understand customers’ needs and automate complex tasks. 

As a business grows with more products, new customers, and changing processes, these agents will adapt alongside it. “Our agentic architecture supports this evolution with planning and reasoning capabilities, memory for historical context, and integrations that enable agents to coordinate more complex workflows as the business becomes more sophisticated,” explains Srivastava. 

Additionally, they learn from the business’s transaction history, user corrections, and exceptions that get flagged over time. 

“At the same time, we ensure business owners remain in control, with the ability to review and approve what the agents do as their processes evolve over time,” he adds.

Engineered for teamwork, not step on each other’s toes: With multiple AI agents working across QuickBooks business functions, how does Intuit ensure coordination between them to avoid redundancy, miscommunication, or conflicting outputs?

The key to keeping them aligned is that each agent is specialized but doesn’t operate in isolation, according to Srivastava. 

They’re built on a shared data architecture and reasoning engine via GenOS, so they work from the same source of truth. For example, the Accounting Agent and Finance Agent share transaction data, categorizations, and historical context so their outputs stay consistent.

Intuit’s system provides planning, reasoning, and orchestration capabilities that let agents coordinate their actions. When one agent completes a workflow, it can automatically inform or trigger related actions in others. 

“So, they’re moving beyond siloed workflows to deliver a truly integrated virtual team experience, just like a real team would coordinate its work,” Srivastava explains.

[Sidebar]: Intuit’s low-key evolution

Intuit is gradually shifting from its reputation as a tax and accounting provider into something broader — a full financial operating system for small and midsize businesses.

This shift has been happening with deliberate moves. In April, Intuit announced two strategic acquisitions: the HR platform, GoCo, and key technology and talent from Deserve, a mobile-first credit card company. Both deals reflect a deeper strategy of expanding the QuickBooks ecosystem to cover more of the SMB back office.

• GoCo: GoCo helps small businesses manage things like onboarding, PTO, and compliance – pain points that QuickBooks Payroll didn’t fully solve. By bringing GoCo in, Intuit can offer SMBs a more complete solution: managing people and processes alongside finances, all from a single dashboard.
• Deserve: Deserve’s infrastructure enables Intuit to explore embedded credit options — from issuing cards to customizing spending controls — all powered by the data businesses already share through QuickBooks. It’s a way to offer smarter lending and cash flow tools without reinventing the credit card wheel.

Together, these acquisitions show a pattern: rather than build from scratch, Intuit is acquiring what it needs to round out its platform, especially for SMBs. It’s a long-game approach to making QuickBooks a hub for business operations.

And it’s not the first time Intuit has done this. Past acquisitions like Mailchimp (for marketing) and Credit Karma (for consumer finance) follow the same logic: fill gaps, connect workflows, and deepen value for existing users.

As the company continues to expand both horizontally (into new domains like HR and lending) and vertically (deeper into customer workflows), it’s carving out a model that’s less about garnering attention and more about delivering expanded functions to SMBs.