Alex Balazs has spent more than two decades inside Intuit, starting as an engineer working on early versions of QuickBooks Online, when moving financial workflows to the internet still felt experimental. Now, as CTO, he is helping lead a more radical shift: turning financial software into systems that can think and act on a user’s behalf. “This combines the speed and scale of AI with human judgment and accountability,” he tells Fast Company.
For decades, financial software has functioned as a ledger, categorizing transactions and generating reports about what has already happened. That model is beginning to break. Advances in AI are pushing the category toward real-time interpretation and action, with software that can execute tasks and manage workflows rather than simply record them.
The shift introduces a core tension. Financial systems demand precision, accountability, and auditability. AI systems operate probabilistically, producing outputs based on likelihood rather than certainty. As the stakes rise, so does the challenge of trusting machines with financial decisions.
Intuit is pushing aggressively into that gap. The company, which controls more than 60% of the SMB accounting software market, is working to turn finance into what it calls a “system of intelligence,” a continuously operating layer that understands financial context and acts on it in real time. Its platform processes roughly 60 billion machine learning predictions per day across a data infrastructure spanning 180 petabytes, serving nearly 100 million consumers and 10 million small and midmarket businesses.
The strategy is already translating into growth. In its most recent quarter, Intuit reported $4.7 billion in revenue, up 17% year over year, with operating income rising 44% on a GAAP basis. The company says its platform facilitates close to $890 billion in money movement and $336 billion in payroll annually.
Under Balazs, Intuit has built what it calls its Generative AI Operating System, or GenOS, designed to coordinate models, data, and workflows into task-specific agents that can execute complex financial operations. Through partnerships with OpenAI and Anthropic, the company is also embedding those capabilities into external AI ecosystems while maintaining control over customer data.
Still, the central question remains: If AI begins to function like an autonomous CFO, who is responsible when something goes wrong?
Speaking with Fast Company, Balazs argues the answer is not full automation, but a new architecture of trust, and a rethinking of how human expertise fits into increasingly autonomous financial systems. This conversation has been edited for length and clarity.
When AI agents are autonomously handling accounting, tax preparation, and cash flow, where do you draw the line between assistance and authority? And should businesses be comfortable handing over that level of decision-making to systems that are, at their core, probabilistic?
The customer is always in ultimate control of critical decision-making and is provided with the needed data to help make those decisions. As we continue to build “done-for-you” experiences for customers on the Intuit platform, we’re creating capabilities and experiences where work is done for the customer on our AI-driven expert platform, with their permission. We’ve always put the power in our customers’ hands. This gives us a durable competitive advantage because it’s what matters most to customers when it comes to financial tasks is instilling complete confidence in their high-stakes financial decisions.
Leveraging proprietary data, domain-specific AI platform capabilities and human intelligence, our system of intelligence uses deterministic domain-specific models built on decades of trusted proprietary data. Intuit Intelligence provides answers grounded in its own proprietary data and will take action on the user’s behalf, through automation and with a handoff to a trusted AI-enabled human expert. This is intelligence rooted in lived financial reality, not generic large language models.
As the industry pushes toward full automation, why keep humans so deeply embedded in financial workflows? Where does that handoff actually happen, and what ensures the human layer remains a real safeguard, not just a symbolic one as AI improves?
We’ve learned that for financial workflows, AI alone is not enough for confidence. Customers have a psychological need for a “data trail” back to the balance sheet. Our QuickBooks Live offering is growing alongside AI because human experts provide a “domain expert check,” showcasing the power of human intelligence. While AI handles the high-volume categorization, humans provide the “final mile” of context to ensure accuracy.
Queries in our system of intelligence aren’t just searches. They hit a “conversational front door” that triggers our Generative AI Operating System (GenOS) to query proprietary data against live transaction data. We address the “confidence gap” through a “show your work” approach, providing a data trail back to the balance sheet and ensuring there are “no dead ends” by handing off complex tasks to live experts (e.g., tax, bookkeeping). One of the surprises we’ve seen in our system of intelligence: We expected accounting questions, but new-to-QuickBooks users are using AI to architect their entire business, even asking about warehouse organization and employee handbooks, for example.
Rather than relying on automation alone, we are utilizing human review, oversight, and feedback to validate high-impact outputs, catch errors, refine model performance, and improve decisions over time.
Intuit has marketed GenOS as the orchestration layer. But as the industry moves toward model-agnostic architectures, with partners like OpenAI and Anthropic, is the real moat shifting away from models to orchestration and data ownership? And if so, what stops that layer from becoming standardized or commoditized as competitors and cloud platforms build similar capabilities?
We built our Generative AI Operating System (GenOS) to solve an enormous challenge: making generative AI broadly available for all product teams to develop solutions that integrate the technology safely and responsibly into applications on our platform. In today’s rapidly evolving tech world, our LLM-agnostic strategy gives Intuit technologists the freedom to choose from a catalog of best-in-class commercial LLMs (15+ LLMs, 70+ versions) and our own proprietary custom-trained Intuit Financial LLMs.
GenOS includes embedded guardrails for security with protections designed to address risks such as prompt injection, data leakage, and harmful outputs, all within a broader responsible AI governance framework. The platform also uses standardized runtime and user-experience layers so teams can build, monitor, and improve AI features consistently, helping deliver more reliable performance and a stable experience at scale across products.
Intuit operates across consumers, SMBs, and now the mid-market, while ERP vendors, fintechs, and cloud providers all push to own the enterprise AI layer. What is the platform’s key differentiator and real moat in this race? And as incumbents embed AI into their stacks and hyperscalers control the infrastructure, what prevents Intuit from getting squeezed in the middle as the market consolidates?
We’re at the beginning of a new era of agent-led growth in financial services that represents a massive tailwind for Intuit in our next chapter. Service-as-software built on data, AI, and human intelligence is delivering solid double-digit revenue growth for Intuit with expanding margins and massive customer impact. This plays to Intuit’s platform advantage—and why we’re built for this moment.
Our AI and human intelligence platform innovation is fueling Intuit’s growth and delivering significant customer benefits. We enable businesses to operate from lead to cash, and help consumers from credit building to wealth building, all in a regulated environment. We aren’t just “using data,” we are grounding queries in 625,000 financial attributes per business and 24,000 bank connections on our platform. And as we scale, the business model strengthens: the more customers we engage, the more insights we gain, which improve recommendations, outcomes, and value for every customer. This creates a powerful network effect that reinforces our competitive advantage.
You’re running 60 billion predictions a day on deeply sensitive financial data, yet even the best models can hallucinate or make errors. How do you reconcile that tension between near-perfect accuracy requirements and inherently imperfect systems? Who is ultimately accountable when an agentic AI-driven financial decision goes wrong?
Our platform deploys multiple advanced technologies that draw on our large and relevant data sets designed to help ensure we’re delivering accurate answers to customers and mitigating the risk of hallucination or other types of inaccurate or inappropriate answers. When our AI provides an answer or gives guidance to a customer, it’s drawing on the deep expertise that Intuit has developed over many years, plus the data that gives us a 360-degree view of the customer. This helps make sure the answer given is relevant and grounded in the customer’s own data.
The company has taken a firm stance on data sovereignty, keeping customer data within Intuit while still embedding capabilities into ecosystems like OpenAI and Anthropic. How do you balance that openness with control? And if models increasingly become the primary interface, is there a risk that the platform layer gets abstracted away despite those safeguards?
Customers are establishing relationships with AI tools such as ChatGPT and Claude, and we want to show up at their point of need. Consumers and businesses using Intuit capabilities within these tools get personalized insights and recommendations powered by the platform to take certain actions. We want to be where our customers are and continue to own the customer relationship and data. Security and privacy are within our platform, and we selectively apply user data at the user’s request to power trusted, accurate responses in ChatGPT and Claude when a user is logged into their Intuit account.
If AI agents take over execution, finance teams inevitably shift from doing the work to supervising it. In your view, what does the future finance organization actually look like? Are we heading toward a world of AI auditors and system overseers, or is there a risk that over-automation erodes financial intuition and literacy in ways we don’t yet fully understand?
AI is already contributing to significant growth in the finance industry. This is especially apparent with data-driven digital brands—approximately 92% of companies that use AI in finance say they’ve either met or exceeded ROI expectations. AI is redefining how teams and organizations run and compete. As the role of AI in finance evolves, there’s a clear shift toward intelligence-driven finance operations. Long-term success, though, will depend on balance. Industry leaders must still find ways to leverage human talent if they want to thrive. At the same time, they’ll need to build internal systems that emphasize accountability and responsibility.
