Building AI for financial software requires a different playbook than consumer AI, and of Intuit The latest QuickBooks release provides an example.
The company announced Intuit Intelligence, a system that leverages AI agents on its QuickBooks platform to handle tasks including sales tax compliance and payroll processing. These new agents add to existing accounting and project management agents (which have also been updated) as well as a unified interface that lets users query data in QuickBooks, third-party systems and uploaded files using natural language.
The new development follows years of investment and improvement at Intuit Genusallowing the company to leverage AI capabilities that are lacking Improving latency and accuracy.
But the real news isn’t what Intuit has built — it’s how it’s built it and why its design decisions will make AI more usable. The company’s latest AI rollout represents an evolution built on hard-won lessons about what works and what doesn’t when deploying AI in a financial context.
What the company learned is stark: Even when its accounting agents improved the accuracy of transaction classification by an average of 20 percentage points, they still received complaints about errors.
"Use cases we are trying to solve for customers include tax and finance. If you make a mistake in this world, you lose trust with consumers in buckets and we only bring it back in spoonfuls." Joe Preston, Intuit’s VP of product and design, told VentureBeat.
The architecture of trust: questions about real data rather than productive responses
Intuit’s technology strategy focuses on fundamental design decisions. For financial queries and business intelligence, the system queries actual data instead of generating responses through large language models (LLMs).
aLSO Critical: This data is not in one place. Intuit’s technical implementation allows QuickBooks to incorporate data from a number of disparate sources: native Intuit data, OAUTH-connected third-party systems such as Square for Payment, and user-uploaded files such as spreadsheets containing vendor pricing lists or marketing campaign data. This creates a unified data layer that AI agents can reliably query.
"We are actually querying your real data," Preston explained. "This is very different than if you were to simply copy, paste a spreadsheet or PDF and paste it into ChatGPT."
This architectural choice means that the Intuit Intelligence system acts more as an orchestration layer. It is a natural language interface for data manipulation. When a user asks about projected profits or wants to run payroll, the system translates the natural language query into database operations against verified financial data.
This matters because Intuit’s internal research has largely exposed the use of shadow AI. When surveyed, 25% of accountants using QuickBooks admitted they were already copying and pasting data into ChatGPT or Google Gemini for analysis.
Intuit’s approach treats AI as a query translation and orchestration mechanism, not a content generator. This reduces the risk of fraud that has plagued AI deployments in the financial context.
Explainability as a design requirement, not an afterthought
Beyond the technical architecture, Intuit has made clarity a core user experience in its AI agents. This goes beyond simply providing the right answers: it means showing users the reasoning behind automated decisions.
When Intuit’s accounting agent classifies a transaction, it doesn’t just display the result. It shows reasoning. This isn’t copy marketing about explainable AI, it’s actual UI that displays data points and logic.
"This is to close this trust loop and ensure that customers understand why," Alistair Simpson, Intuit’s head of design, told VentureBeat.
This becomes especially important when you consider Intuit’s user research: while half of small businesses find AI helpful, nearly a quarter don’t use AI at all. The description layer serves both populations: building confidence for newcomers, while providing context for validating experienced users.
This design also enforces human control at key decision points. This view is outside the interface. Intuit connects users directly with human experts, embedded in the same workflows, when automation reaches its limits or when users want validation.
Navigating the transition from form to conversation
Another interesting challenge for Intuit involves managing a fundamental change in the user interface. Preston described it as having one foot in the past and one foot in the future.
"It’s not just Intuit, it’s the market as a whole," Preston said. "Today we still have a lot of users filling out forms and going through tables full of data. We’re very invested in leaning and questioning the ways we do this in our products today, where you’re basically just filling out, form after form, or table after table, because we see where the world is going, which is really a different way of interacting with these products."
This creates a product design challenge: How do you serve users who are comfortable with traditional interfaces while gradually introducing conversational and agent capabilities?
Intuit’s approach is to embed AI agents directly into existing workflows. This means not forcing users to adopt entirely new interaction patterns. The payment agent appears with invoicing workflows. An accounting agent augments rather than replaces an existing reconciliation process. This incremental approach lets users experience the benefits of AI without abandoning familiar processes.
What Enterprise AI Builders Can Learn from Intuit’s Perspective
Intuit’s experience deploying AI in a financial context levels several principles that are broadly applicable to enterprise AI initiatives.
Architecture Matters for Trust: In domains where accuracy is important, consider whether you need translation for content generation or data querying. Intuit’s decision to treat AI as an orchestration and natural language interface layer dramatically reduces the risk of deception and avoids using AI as a production system.
The description should be designed, not bolded: Why would an AI make a decision when trust is at stake? This requires deliberate UX design. This may limit the choice of model.
User control preserves confidence while improving accuracy: Intuit’s accounting agent improved classification accuracy by 20 percentage points. Still, it was important to retain end-user capabilities.
A gradual transition from a familiar interface: Don’t force users to leave the form to chat. Embed AI capabilities into workflows first. Experience the benefits before asking customers to change behavior.
Be honest about what is reactive vs. proactive: Current AI agents primarily respond to cues and automated tasks. True functional intelligence that makes unsolicited strategic recommendations remains an evolving capability.
Address workforce concerns with tooling, not just messaging: If AI is intended to augment rather than replace workers, provide workers with AI tools. Show them how to take advantage of technology.
For businesses navigating the adoption of AI, Intuit’s Journey offers a clear guide. A winning approach prioritizes confidence over demonstration of competence. In domains where mistakes have real consequences, this means investing in accuracy, transparency, and human oversight before achieving conversational sophistication or autonomous action.
Simpson successfully frames the challenge: "We didn’t want it to become a bolt-on layer. We wanted users to be in their natural workflows, and for agents doing the work for users, embedded in the workflow."