When Intuit’s CEO demanded the company deliver its most ambitious AI launch by September 2023, the $200 billion software giant behind QuickBooks, TurboTax, and Mailchimp responded with characteristic speed. The result was Intuit Assist—a chatbot-style assistant grafted onto existing applications to demonstrate the company’s AI credentials.
It was supposed to revolutionize how small businesses managed their finances. Instead, it flopped spectacularly.
“When you take a beautiful, well-designed user interface and you simply plop human-like chat on the side, that doesn’t necessarily make it better,” Alex Balazs, Intuit’s Chief Technology Officer, reflects on the failed launch.
The chatbot consumed valuable screen space while confusing users with a blinking cursor that seemed to mock their uncertainty. Dave Talach, Senior Vice President of the QuickBooks team, describes the aftermath as falling into the “trough of disillusionment.” The pressure was intense—he found himself presenting to Intuit’s Board of Directors, explaining what went wrong and how the team planned to recover.
What followed wasn’t a minor course correction but a grueling nine-month transformation that would fundamentally reshape how the 40-year-old company builds products. This is the inside story of how Intuit emerged with a practical playbook for enterprise AI that other leaders can adapt.
The pivot away from chatbots began with a simple observation of customer behavior. Talach’s team noticed QuickBooks users struggling with a common workflow: manually transcribing invoice information from emails into their accounting software, splitting their computer screens between their email client and QuickBooks.
This “split-screen” behavior sparked a crucial realization—why force humans to act as copy-paste machines when artificial intelligence could automatically extract data from emails and populate invoices? The insight represented a fundamental shift in thinking: instead of inventing new behaviors through chat interfaces, Intuit should focus on eliminating “manual toil” within existing customer workflows.
This bottom-up momentum caught the attention of senior leadership. Balazs and Marianna Tessel, General Manager of the business group, recognized they needed to make a decisive commitment. “We need to make a declaration together,” Tessel told Balazs. The path forward required full commitment to an AI-native future—what Balazs describes as “burning the boats.”
To execute this transformation, management deployed Clarence Huang, a key technology leader, from the core tech team directly into the heart of the QuickBooks business. His mission was to scale a “builder-centric mindset” focused on rapid, customer-focused prototyping.
The transformation also required difficult organizational decisions. To empower smaller, faster teams, Intuit eliminated layers of middle management, letting go of 1,800 employees in 2024 whose roles no longer aligned with new priorities. Simultaneously, the company committed to hiring approximately 1,800 new employees with skills in engineering, product development, and customer-facing roles.
Intuit’s successful pivot required rebuilding three fundamental aspects of the organization: its people and culture, its processes and workflows, and its underlying technology infrastructure.
The first pillar focused on assembling the right talent and empowering them to work in entirely new ways.
Aggressive talent acquisition: Intuit expanded its core AI team from just 30 people in 2017 to several hundred today, accelerating hiring over the past two years by recruiting top-tier AI leaders from companies like Uber, Twitter, and ByteDance, the parent company of TikTok.
Cross-functional team structures: The new operating model centered on small, empowered teams that included members from up to 10 different units—data science, research, product management, design, engineering, and others. Each team focused exclusively on delivering specific AI-powered experiences. To enable this focus, managers practiced ruthless prioritization, eliminating any tasks outside the top three priorities. “That ruthless prioritization was really, really important,” Huang emphasizes.
Dissolved role boundaries: Traditional job descriptions became fluid in what Huang calls a “smearing” of roles. Everyone was expected to engage directly with customers—Huang himself maintained a spreadsheet of 30 customer names he called regularly. This cultural shift produced remarkable results, exemplified by data scientist Byron Tang, who used AI-powered coding tools to single-handedly build a fully functional prototype with a polished user interface. Huang recalls his reaction: “Oh my god, you are the renaissance man. You got it all!”
With the right people in place, Intuit systematically dismantled the processes that typically slow large organizations, replacing them with systems optimized for speed and customer focus.
Prototype-driven development: The company abandoned traditional specification documents in favor of a new philosophy: a prototype is worth 10,000 words. Teams began shipping functional prototypes to customers almost immediately. “We’ll literally show a working, functioning prototype to the customer and we’ll code modifications on the spot,” Huang explains. “The reaction on their faces is just magic.”
Customer-centric design: This rapid feedback approach led to key innovations, including a “Slider of Autonomy”—a concept popularized by AI researcher Andrej Karpathy that gives users control over the level of AI intervention. Intuit discovered that customers feared features that seemed “too magical,” so the company created controls ranging from full automation to manual review, providing a “smooth onramp” to trusting AI agents. For example, in QuickBooks’ accounting agent, users can click a button to automatically post all recommended transactions, or use detailed explanations to review the AI’s reasoning before approving changes.
Bureaucracy elimination: Leadership actively removed organizational friction through specific policies: “no meetings on Tuesdays” for the platform team, banned afternoon meetings for individual contributors in business units, and instituted a formal “friction busting” campaign requiring leaders to resolve inter-team disagreements within seven days. The company also revised restrictive AI rollout rules, increasing the limit for customer experiments from 10 to 1,000 participants.
The entire transformation relied on GenOS, Intuit’s internal AI platform developed under Chief Data Officer Ashok Srivastava’s vision to democratize AI access across the company. Rather than building the platform first and then finding applications, GenOS evolved alongside business needs through what CTO Balazs calls “Fast Follow Harvesting.” As customer-facing teams built AI agents, they identified platform gaps, which a central team then addressed with new features.
A cornerstone of GenOS was the Agent Starter Kit, which enabled 900 internal developers to build hundreds of AI agents within five weeks. The platform also included runtime orchestration and governance frameworks to ensure consistent performance and compliance.
Another critical component was an LLM router—a system that automatically directs requests to different large language models depending on which performs best for specific tasks while providing backup options when primary systems fail. Huang recalls receiving a late-night call from Srivastava asking, “OpenAI is down. Are you guys okay?” Because the team was using GenOS, “it just auto-switched to the fallback LLM in the gateway—it was okay.”
This infrastructure allows Intuit to leverage its core competitive advantage: decades of domain-specific financial data. By fine-tuning AI models on a focused set of financial tools and APIs, Intuit’s agents achieve accuracy that general-purpose models cannot match. “In all of our internal benchmarks, our stuff just works better for in-domain data,” Huang notes.
The transformation produced a suite of AI agents deeply integrated into QuickBooks and increasingly across Intuit’s other products. The QuickBooks Payments Agent proactively suggests adding late fees based on customer payment history, while the Customer Agent transforms QuickBooks into a lightweight customer relationship management system by scanning connected Gmail accounts for leads. The Accounting Agent automates transaction categorization and identifies anomalies.
The business results are substantial: Small businesses using these AI agents get paid five days faster on average, are 10 percent more likely to collect on overdue invoices, and save up to 12 hours monthly on administrative tasks. These “virtual employees,” as Talach describes them, surface their work through tiles in the QuickBooks “business feed,” transforming the dashboard into an active, collaborative workspace.
These capabilities could help Intuit compete more effectively with specialized service providers like HubSpot, a marketing and sales platform, by offering more comprehensive solutions to existing customers. In the company’s most recent quarterly earnings call, CEO Sasan Goodarzi attributed strong results—16 percent growth for the full year—to AI investments. The agent launch was already showing traction: “We’re seeing strong engagement in the millions and repeat usage rates significantly above our expectations.”
Intuit is now applying this approach to larger opportunities, recently announcing agents for mid-market companies with up to $100 million in revenue—a significant expansion from the company’s traditional customer base of businesses with $5 million or less in revenue. The logic is straightforward: larger customers have more complex workflows and therefore greater potential benefit from AI automation.
For enterprise leaders navigating their own AI transformations, Intuit’s experience offers several practical insights. Initial failures aren’t just common—they may be necessary learning experiences. Success requires more than deploying impressive AI capabilities; it demands dismantling existing organizational structures and building new cultures, processes, and platforms that enable established companies to move with startup agility.
The most important lesson centers on starting point: begin with understanding the actual work customers do, not with the technology capabilities you want to showcase. Intuit’s breakthrough came from observing the mundane reality of split-screen invoice transcription, not from pursuing cutting-edge AI research.
The transformation also demonstrates that successful enterprise AI implementation requires simultaneous changes across people, processes, and technology. Attempting to layer AI onto existing structures typically produces the kind of confusing, space-consuming chatbot that initially failed for Intuit. True transformation requires the courage to “burn the boats” and commit fully to new ways of working—a challenging but ultimately rewarding path for organizations ready to embrace the AI-native future.