Semantic understanding, not just vectors: How Intuit’s data architecture powers agentic AI with measurable ROI

Intuit – FINANCE Program Giant behind products such as Turbotax and QuickBooks – makes significant steps using a generative AI to increase the offers for small business customers.

In a technological landscaping in a technological landscaping, intuitively, an agent-based AI architecture, which introduces material work for small businesses in a water promises. The company has applied the experience that manages all workflows and provides the work of a kuyu.

Intuit, a generative AI builds AI layer called the operating system (Geno). The company explains some ways used by Gen AI to improve individualization VB Transform 2024. Intuit added in September 2024 Agent AI Work streams, an effort that improves transactions for both the company and users.

According to new intuit data, QuickBooks online clients are paid in an average of five days, 10% will be fully paid with overdue invoices. For small businesses where cash flow is king, these are not only growing progress – potentially saving business savings.

Technical Trinity: The information architecture of the intuitive information is the opportunity to the actual agent ai

It is a complex information architecture specially designed to activate agent-based AI experiences that separates the approach of the intuitation from competitors.

The company sets CDO Ashok Srivastava called the “trio” of information systems:

1. Lake information: Basic depot for all data.
2. Client Information Blow (CDC): A special service layer for AI experience.
3. “Event bus“: A stream information system that provides real-time operations.

“CDC presents a service layer for AI experience, then the information is a kind of warehouse for all such information,” said Srivastava Ventureat. “The agent will interact with information and has a collection of information that you can view to collect information.”

Power Agent goes beyond vector facilities to AI

Intuit architecture applies to many institutions from a typical vector database. Vector databases and embeddings are important to enhance the AI ​​models, the intuit accepts a more unified approach to a real semantic agreement.

“The main issue is also important to ensure that our information is good, logical and a semantic understanding,” Srivastava said.

To achieve this semantic concept, the intuit builds a semantic information on the basic information infrastructure. Semantic data layer, around the data, helps ensure context and meaning outside of raw data itself or vector representations. This allows you to better understand the connections and connections between different information sources and elements of AI agents.

It can increase the capacity of vector-based systems with intuitive, more infected data, build a semantic data layer. This allows you to make more informative and meaningful decisions for AI agents for customers.

Outside the main automation: Agentic AI autonomous how to work

Unlike enterprises that automate the AI ​​under the basic workflow automation or the businesses that apply for customer service chatboots, intuitive, it is built to create an experienced full agent for you “prepared for you”. These are applications that manage complex, multi-step tasks when you only require the final human confirmation.

The agent system for QuickBooks users analyzes an invoice status to prepare individual reminder messages that allow business owners to review and confirm the business owners automatically to automatically study the payment date and invoice status. The ability to personalize the system based on the contacts and payment samples directly contributed to faster payments directly.

Intuit, internal, developing autonomous procurement systems and HR helpers apply the same agency principles that develop.

“We have the ability to have the internal agent procurement process that employees can use for supply and book trips,” Srivastava demonstrated how the company eats its EI dog food.

Designed for a substantial model cycle

Potely, the competitive advantage in connection with the application of other enterprises AI is what the system is built in the system of developed thinking models such as DeepSeek.

“We set up a gene working time to think of wide models,” Ashok said. “We’re not behind the eight balls … We are forecasting. We have established opportunities that the basis exposes.”

This forward thinking, intuitive, means, without architectural repair, the experience of its agency can quickly achieve its new thinking. According to Srivastava, the intuitive engineering groups use these opportunities that allow you to provide a large number of tools and data after information.

To afford to work from the AI ​​hype

Perhaps the most important, the approach of an intuitation shows a clear focus on the work results than the technological show.

“It’s a lot of work and much work to revolutionize the world and all this has to revolutionize all this and there are many fans.” But I think it’s better to show that what’s better helpful to real people. “

The company believes that deep thinking skills will provide more comprehensive “for you” experiences that cover more more customer needs. Each experience combines many atomic practices or discrete transactions that create a complete workflow solution together.

What does this mean for the enterprises accepted by AI

For businesses who want to implement AI effectively, Intuitimate approaches offer several valuable lessons for enterprises:

• Note the results over technology: Instead of showing the AI ​​with measurable development targets for a certain job pain points to the target for their sake.
• Build with future models: Design architecture that can concentrate the skills that emerge without being completely rebuilt.
• Address information Problems first: Before the rush to implement agents, your foundation of the data can ensure the justification of the semantic understanding and the cross-system.
• Create a complete experience: Lack of simple automation to create the end until the end to end the workflows offering full solutions.

Agent AIT, AI, fully focusing on the Izilan Eii features of enterprises that follow the example of an intuitive example, can simply achieve similar concrete work results.

Daily Definitions from Daily Works Daily

If you want to surprise your boss, you covered your VB diary. We provide an internal bucket because they work with companies from regulation shifts to practical places, so you can share ideas for the maximum ROI.

Read we read Privacy policy

Thank you for your subscription. Check more VB bulletins are here.

An error occurred.