For over a decade, Intuit has quietly been preparing for a generative AI revolution that a large part of the world only began to comprehend in the last couple of years.
“We were already working on generative AI capabilities when the rest of the world started investing heavily in it,” Ashok Srivastava, chief AI and data officer at Intuit, told AIM. “We had seen the benefits and were ready to go.”
Srivastava explained that this early investment culminated in the creation of Intuit’s GenOS, launched in June 2023. GenOS now powers all of Intuit’s generative AI and agentic experiences, abstracting away the complexity of various underlying systems—data platforms, user interfaces, and back-end infrastructure—to allow for large-scale deployment of AI agents.
“We created GenOS to avoid the inefficiency of having thousands of developers building thousands of separate AI applications,” Srivastava said. “That would mean losing the power of the platform we’ve built.”
Rather than build fragmented tools, GenOS enables Intuit to deliver consistent, scalable AI experiences by integrating evaluation, personalisation, security, and fraud prevention directly into the core platform.
The system now powers production-ready AI agents, including accounts receivable and accounts payable agents designed to automate cash flow management tasks. These agents have been alive since September 2023.
Why Agentic AI
Srivastava brings a deep background in AI and machine learning, having worked in neural networks, signal processing, and control systems since his graduate school days. He’s authored over 100 papers in the field and continues to contribute as both a leader and a technologist.
“Did I know it was going to be exactly what we saw in ChatGPT? No. But did I know we needed to invest in AI as a core capability? Absolutely. And we did that.”
While many companies were racing to build AI integrations, Intuit took a different path by developing an operating system from the ground up. This gave engineers and developers a unified framework that could deliver consistent, reliable, and secure generative AI experiences.
Microsoft and Google are pouring billions into their own LLM ecosystems. Srivastava emphasised both platform depth and data richness, for example, Microsoft just onboarded around 11,000+ AI models on its Azure AI Foundry.
But he’s more focused on outcomes than sheer volume. “Do we need 11,000 models to get our customers paid faster? No. Probably, we don’t—because we’ve already done it.”
Intuit’s agentic system can get customers paid 45% faster, an average of five days sooner, thanks to automated transaction matching and invoice review. That kind of tangible outcome, Srivastava says, matters more than the underlying model count.
“If I’m running a small business and I get paid faster, that’s a win. I wouldn’t care if there were one or 11,000 models behind it.”
“We didn’t want hundreds of disconnected tools. We wanted a single system that developers can build on, with security and consistency,” Srivastava said.
GenOS enables seamless integration of over a dozen LLMs—from commercial offerings to open-source models fine-tuned in-house. This flexibility allows developers to choose the best model based on security, risk, and performance, all without compromising user experience.
“We’ve built evaluation services that work across all our agentic experiences. If you don’t do that, some experiences work, and others don’t—it becomes inconsistent. Our operating system ensures everything scales.”
On the data side, Intuit’s access to rich, structured financial and behavioural data allows the company to personalise AI experiences in real-time, delivering cash flow forecasts, intelligent recommendations, and context-aware automation tailored to the customer’s immediate needs.
“Our experiences are tuned to the moment. That gives us a serious advantage.”
Open Source as a Strategic Pillar
Another pillar of Intuit’s strategy is its commitment to open source. Projects like Admiral, NumaProj, and Agroproj, are part of the company’s effort to contribute to the broader community and leverage the best available technologies.
“We care and give back—it’s one of our core values,” Srivastava said. “We want to build tools based on our expertise so others can benefit.”
Intuit has received the End User Award from the Cloud Native Computing Foundation twice—in 2019 and 2022—an honour reserved for companies that contribute and scale open source technologies effectively.
While Srivastava declined to pick a “favourite child” among Intuit’s open-source initiatives, he noted that NumaProj has delivered significant value both internally and externally.
Intuit’s platform also powers a suite of widely-used products including QuickBooks, TurboTax, Mailchimp, and Credit Karma. Each is designed not just to function as a tool but to provide “done-for-you” experiences, where AI does the heavy lifting and the user only intervenes when necessary.
“What keeps me up at night is thinking about how we can actually help small businesses and consumers do better,” said Srivastava. “Ninety-nine percent of US firms are small businesses, and many are under pressure from economic changes. They’re trying to survive, navigate uncertainty, and stay afloat.
At the same time, consumers are facing reduced access to capital. “How do we support both groups in this environment? That’s where AI agents come in — they’re built on platforms we’ve developed over many years to solve these exact problems.”
Srivastava used to run an art gallery in Mountain View and he said that if he had access to QuickBooks’ AI back then, with automated accounts receivable and payable, “I would have absolutely used it.” With this background, he also remains optimistic about the use of AI in the field of art.
“It is just another paintbrush.” Creativity has always found different outlets — music, art, literature, science, and maths.
“Some people may choose to use it to express themselves, others won’t. I have close friends who are established artists, and they see AI as just one more medium, not a replacement. That’s how art has always evolved — new styles emerge, people explore them, and then move on. This is no different.”
The post Why Intuit Decided to Build GenOS Instead of Onboarding 11,000 AI Models appeared first on Analytics India Magazine.
