Join an event that enterprise leaders have been trusted for nearly 20 years. VB Transform brings together people who build real enterprise AI strategies. learn more
Intuit has travelled over the past few years using technology-incorporating Generative AI as part of its services at QuickBooks, Credit Karma, Turbotax and MailChimp.
Today, the company is taking the next step with a range of AI agents that go beyond that to change how small and medium-sized businesses operate. These new agents act as virtual teams that automate workflows and provide real-time business insights. These include payment, accounts and financial capabilities that directly affect business operations. According to Intuit, customers save up to 12 hours a month, and on average they get paid up to 5 days faster thanks to new agents.
“Looking at the trajectory of AI experiences at early Intuit, AI is embedded in the background and we saw a shift in providing information to our customers with Intuit Assist,” Ashok Srivastava, AI Chief and Head of Data at Intuit, told VentureBeat. “What you're looking at now is a complete redesign. Agents are actually working on behalf of the customer and have permission.”
Technical Architecture: From starter kits to production agents
Intuit has been working on the path from assistant to agent AI for a while.
In September 2024, the company detailed its plans to use AI to automate complex tasks. This is a solidly built approach to the company's Generation AI Operating System (GENOS) platform, which is the basis of AI efforts.
Earlier this month, Intuit announced a series of initiatives to further expand its capabilities. The company has developed its own rapid optimization service that optimizes large-scale language model (LLM) queries. We also developed what is called the Intelligent Data Cognitive Layer of Enterprise Data, which allows us to understand the various data sources needed for our enterprise workflows.
Going it a step further, Intuit has developed an agent starter kit based on the company's technical foundation to enable agent AI development.
Agent Portfolio: From cash flow to customer management
Introducing technical foundations, including an agent starter kit, Intuit has built a set of new agents to help business owners get things done.
Intuit's agent suite demonstrates the technical refinement needed to move from predictive AI to autonomous workflow execution. Each agent coordinates predictions, natural language processing (NLP), and autonomous decision-making within a complete business process. They are:
Payment Agent: Autonomously optimize cash flow by predicting late payments, generating invoices, and performing follow-up sequences.
Accounting Agent: Represents the evolution of Intuit from a rule-based system to autonomous bookkeeping. Agents now autonomously handle transaction classification, settlement and workflow completion, providing cleaners and more accurate books.
Financial AgentAutomate strategic analytics that traditionally require dedicated business intelligence (BI) tools and human analysts. Provides key performance indicators (KPI) analysis, scenario planning, and forecasting, and autonomously generate growth recommendations based on how the company is doing against peer benchmarks.
Intuit also builds a customer hub agent that can help you with your customer acquisition tasks. Payroll processing and project management efforts are also part of future release plans.
Beyond Conversational UI: Task-oriented Agent Design
The new agent shows an evolution in how AI is presented to users.
Redesigning the Intuit interface reveals key user experience principles for enterprise agent deployment. Rather than bolting AI features to existing software, we have fundamentally rebuilt the AI QuickBooks user experience.
“Currently, the user interface is oriented around the business tasks that need to be performed,” explained Srivastava. “Real-time insights and recommendations can come directly to the user.”
This task-centric approach contrasts with the chat-based interface that dominates current enterprise AI tools. Instead of asking users to learn prompt strategies or navigate conversation flows, agents work within existing business workflows. The system includes what Intuit calls a “business feed” that brings the actions and recommendations of agents contextually to the surface.
Trust and Verification: Closed Loop Challenge
One of the most technically important aspects of Intuit implementation addresses the critical challenges of autonomous agents deployment: validation and trust. Enterprise AI teams often suffer from black box issues. How do you ensure that AI agents perform correctly when operated autonomously?
“To build trust in artificial intelligence systems, we need to provide our customers with a proof point that what they think is happening is actually happening,” emphasized Srivastava. “That closed loop is very important.”
Intuit's solution builds validation capabilities directly on Genos, allowing the system to provide evidence of agent actions and outcomes. For payment agents, this means indicating to the user that an invoice has been sent, tracking delivery and demonstrating improvements in payment cycles due to agent behavior.
This validation approach provides templates for enterprise teams deploying autonomous agents in high-stakes business processes. Rather than asking users to trust the output of AI, the system provides auditable trails and measurable results.
What does this mean for companies looking to join Agent AI?
Intuit's Evolution provides a concrete roadmap for enterprise teams planning autonomous AI implementations.
Focus on completing the workflow, not on conversation. Rather than building a generic chat interface, you target specific business processes for end-to-end automation.
Build an agent orchestration infrastructure: Invest in a platform that coordinates prediction, language processing, and autonomous execution within a unified workflow, rather than isolated AI tools.
The design verification system will be in advance: Include comprehensive audit trails, results tracking, and user notifications as core features rather than afterthoughts.
Map your workflow before building your technology: Use the Customer Advisory Program to define agent functions based on actual operational challenges.
Planning an interface redesign: Optimizes the UX of agent-driven workflows rather than traditional software navigation patterns.
“As large-scale language models become commoditized, the experience built on them becomes much more important,” Srivastava said.
Source link
