Join our daily and weekly newsletters for the latest updates and the exclusive content on AI coverage. Learn more
Intuity – The financial software giant behind products like Turbotax and QuickBooks – Makes significant progress using a generative AI to improve its offers for small businesses.
In a technological landscape flooded with IA promises, Intuit has built an AI architecture based on agents that offers tangible commercial results for small businesses. The company has deployed what it calls “made for you” experiences that manage workflows independently and offer a quantifiable commercial impact.
Intuit built his own AI layer, which he calls an AI (Genos) operating system. The company has detailed some of the ways to use Gen AI to improve personalization at VB Transform 2024. In September 2024, Intuit added Agentic Workflows, an effort that has improved operations for the company and its users.
According to new intuity data, online QuickBooks customers are paid on average five days faster, with invoices late 10% more likely to be paid in full. For small businesses where cash flows are King, they are not only progressive improvements – they are potentially economic innovations.
The Technical Trinity: How Data Architecture of Intuit allows a real agentic AI
What separates the approach of intuits from competitors is its sophisticated data architecture designed specifically to allow AI experiences based on agents.
The company has built what CDO Ashok Srivastava calls “a trinity” of data systems:
- Lake: The fundamental benchmark for all data.
- CLACK DATA Cloud (CDC): A specialized service layer for AI experiences.
- “”Event bus“: A streaming data system allowing real -time operations.
“CDC provides a layer of service for AI experiences, so data lake is a bit of the benchmark for all this data,” Srivastava told Venturebeat. “The agent will interact with the data, and he has a set of data he could examine to extract information.”
Go beyond
Intuit architecture diverges from the approach of the typical vector database that many companies are implemented in a hurry. Although databases and vector interests are important to feed AI models, Intuits recognize that real semantic understanding requires a more holistic approach.
“When the key problem continues to be, it is essentially to guarantee that we have a good logical and semantic understanding of the data,” said Srivastava.
To reach this semantic understanding, Intuits builds a semantic data layer in addition to its basic data infrastructure. The semantic data layer helps to provide a context and a meaning around the data, beyond the raw data itself or its vector representations. It allows IA agents to better understand the relationships and connections between different data sources and elements.
By building this semantic data layer, Intuits is able to increase the capacities of its vector -based systems with a deeper and more contextual understanding of the data. This allows AI agents to make more informed and significant decisions for customers.
Beyond basic automation: how the AI agent finishes whole business processes independently
Unlike companies implementing AI for the automation of the basic workflow or customer service chatbots, Intuits focused on creating fully agentic “Done for You” experiences. These are applications that manage complex tasks and in several stages while requiring final human approval.
For users of QuickBooks, the agentic system analyzes customer payments history and the status of the invoice to automatically write personalized recall messages, allowing business owners to simply examine and approve before sending. The capacity of the system to personalize according to the relational context and the payment models has directly contributed to faster measurable payments.
Intuit applies internal internal agent principles, developing autonomous supply systems and HR assistants.
“We have the possibility of having an internal agent purchase process that employees can use to buy supplies and reserve trips,” said Srivastava, demonstrating how the company eats its own AI dog food.
Designed for the era of the reasoning model
What potentially gives intuity a competitive advantage over the other implementations of the corporate AI, is the way in which the system was designed by providing for the emergence of advanced reasoning models like Deepseek.
“We built Gen Runtime in anticipation of the upcoming reasoning models,” Ashok revealed. “We are not behind the eight balls … We are ahead of. We have built the capacity by assuming that reasoning would exist. »»
This avant-garde conception means that Intuits can quickly integrate new reasoning capacities into their agency experiences as they emerge, without requiring architectural revision. According to Srivastava, Intuits engineering teams already use these capacities to allow agents to reason through a large number of tools and data in a way that was not possible before.
Go from AI media threw to commercial impact
Perhaps the most important, the approach to intuits clearly shows the emphasis on commercial results rather than on technological staging.
“There is a lot of work and a lot of fanfare that takes place these days on the AI itself, that it will revolutionize the world, and all that, which, I think, is good,” said Srivastava. “But I think what is much better is to show that it really helps real people better.”
The company estimates that more in -depth reasoning capacities will allow even more complete “made for you” experiences that cover more customer needs with greater depth. Each experience combines several atomic experiences or discreet operations which together create a complete workflow solution.
What it means for companies adopting AI
For companies that seek to effectively implement AI, the Intuit approach offers several valuable lessons for companies:
- Focus on technology results: Rather than presenting AI for its own target commercial pain points with measurable improvement goals.
- Build with future models in mind: Design architecture which can incorporate emerging reasoning capacities without requiring complete reconstruction.
- First take up data challenges: Before rushing to implement agents, make sure that your data foundation can take charge of semantic understanding and cross reasoning.
- Create complete experiences: Look beyond the simple automation to create workflows from start to finish “fact for you” which provide complete solutions.
While the AI agent continues to ripen, companies that follow the example of intuits by focusing on complete solutions rather than isolated AI functionalities can find themselves to obtain similar concrete trade results rather than simply generating technological buzz.
