Combining AI agents and robots to cut web inquiry response time by two-thirds

Pushing the limits of back-office automation with business logic

As the automotive industry faced sweeping transformation, Toyota Finance found itself compelled to rethink its approach to system investment. With capital increasingly being directed toward strengthening customer-facing operations, investment in back-office systems became relatively constrained. The company's response was to adopt a new automation approach combining AI and robots. Using the UiPath Platform™, a proof of concept (PoC) confirmed that processing time for web inquiry responses could be reduced by approximately two-thirds. A new model of work transformation—built on role-sharing between people, AI, and robots—had begun to take shape.

A shift toward customer-facing investment demands back-office reform

Toyota Finance supports customers across every stage of car ownership with a range of financial services. With sales, finance, and credit cards as its core businesses, the company has been expanding its customer touchpoints through tools like the "TOYOTA Wallet" cashless payment app. Kenichi Hirano, General Manager of the Digital Business Promotion Department, explains: "TOYOTA Wallet is specifically designed to deepen our retention strategy—using the power of finance to stay continuously connected with owners."

Kenichi Hirano, Digital Business Promotion Department Minister, Toyota Finance

However, the company's strategic focus on front-end services—combining auto credit, credit card, and lifestyle offerings to drive continued customer engagement—created pressure on how internal systems are developed and maintained. Hirano reflects: "Financial services require enormous investment in internal systems—even just the credit card business alone. We've invested heavily over the years to automate backend systems for lending and credit assessment. But as we've begun prioritizing customer-facing products like TOYOTA Wallet, investment has shifted away from back-end systems toward customer services. Investment in internal corporate systems has declined relatively."

As large-scale investment in back-end business systems became harder to justify, fundamental system overhauls grew difficult. The company had managed costs through incremental automation initiatives—but challenges were mounting. As Hirano notes: "Automation work that's still not done today has become harder to justify outsourcing due to weak ROI. We've built automation systems in-house using tools like RPA, but we've started hitting the limits of what existing tools can do to improve efficiency." The need was clear: to drive operational improvement by automating remaining processes while still leveraging legacy infrastructure.

While existing RPA tools could handle UI-based automation—controlling screens and interfaces—implementing complex business logic through UI interactions alone proved difficult. The company began evaluating automation tools capable of implementing business logic directly.

Breaking through the limits of generative AI alone with a robot-AI combination

Alongside their efforts to automate business logic, Toyota Finance also intensified focus on AI adoption. Yuki Hirai, Senior Specialist in the Digital Work Promotion Group, Digital Business Promotion Department, recalls: "We saw generative AI spreading fast and worked to build internal literacy. We used tools like Microsoft Copilot to automate individual tasks and improve quality. But that alone didn't translate into broader business automation." From the perspective of driving automation through end user computing (EUC) and RPA, the team began exploring ways to apply generative AI to handle the kind of ambiguous, judgment-based tasks that typically require human intervention.

Yuki Hirai, Digital Work Promotion Group Director, Toyota Finance

"We thought that if we handed off tasks we hadn't been able to formalize to generative AI, it might handle them 'just right," Hirai explains. "Around the spring of 2025, when AI agents were generating significant buzz, we partnered with a systems integrator to begin a PoC using an AI agent platform."

Specifically, the team ran a PoC to apply AI agents to the task of automating email responses to customer inquiries. The approach used retrieval augmented generation (RAG) to register the company's reply templates and manuals with the AI, enabling it to generate responses based on internal knowledge. However, challenges quickly emerged. "The AI would sometimes produce different answers to the same question, or do things we had explicitly prohibited—in some cases, generative AI alone proved difficult to control," Hirai explains. "We couldn't consistently exceed the accuracy threshold set by the operations team, and ultimately concluded that we couldn't hand full control over to generative AI."

Through the PoC process, an important insight emerged: business processes are, at their core, bundles of business logic, and an AI agent platform anchored by rule-based robots—capable of executing predictably—was better suited to the task. "That insight led us to take a closer look at UiPath as a platform that could implement business logic with robots at the center," says Hirai. The team zeroed in on the combination of UiPath AI agent creation tool, Agent Builder in UiPath Studio, and UiPath Robots.

Having decided to step back from letting generative AI compose emails on its own, the team reconsidered the value of AI in a more bounded role. "We realized that even just having the AI classify which template was most appropriate for a given inquiry would meaningfully improve operational efficiency on the floor," says Hirai. When the team shared their situation with UiPath, a hands-on demo combining AI agents and robots was ready the following week. Seeing the demo, Hirai recalls: "Once I saw the AI agents and robots working together, I felt confident that generative AI could genuinely contribute to automation."

In practice, when a new web form inquiry comes in, a robot gathers relevant information from existing systems—including the customer information system—and a response-drafting AI agent then generates a suggested reply. The approach strategically combines robots, AI agents, and human oversight within the business logic flow to maximize operational efficiency. With an 80% accuracy rate set as the PoC criterion, generative AI alone fell short—but by combining it with business logic through UiPath, the team achieved 93% accuracy.

Dramatic PoC results validate the power of AI-robot division of labor

When a PoC was conducted for this measure, the results exceeded expectations. Tomomasa Takahashi, Group Manager of the Digital Work Promotion Group in the Digital Business Promotion Department, said, "A task that used to take 13 minutes was reduced to 4 minutes. We were able to reduce the time to about one-third of what it was before, which surprised the team members on the ground. We feel that the division of labor was effective, with robots automating tasks such as researching customer information systems and checking internal information, and AI agents handling the task of creating documents."

Tomomasa Takahashi, Digital Work Promotion Group Manager, Toyota Finance

The web inquiry system kicked off its PoC in October 2025 and reached full production in January 2026—the result of iterative discussions with UiPath to shift from an AI-first automation approach to one combining rule-based robots. Takahashi reflects on the achievement: "App development can sometimes take six months to a year. We value the speed at which we went from kickoff to production implementation in just three months. It was made possible not just by our team and UiPath, but by the cooperation of the operations departments as well. The fact that web inquiry response times were cut by two-thirds has been a major win for operations, and it's become a catalyst for them to view UiPath-powered automation in a positive light."

The experience of combining robots and AI agents to automate business logic also surfaced important insights. Hirano notes: "We had assumed that back-office work contained a lot of ambiguous human judgment that would be hard to automate. But through the PoC, we discovered that much of it could actually be formalized into business logic." Hirai adds: "One of the key takeaways from this initiative was realizing that even tasks we'd written off as too human-dependent could be highly automated through logical structuring."

Accelerating business transformation with in-house development across 2,000 employees

The combined AI and robot automation approach has borne fruit beyond web inquiries. Hirano explains: "We also ran a PoC on automating expense reimbursement. Using UiPath's AI-OCR tool, Document Understanding, we digitized invoices and automated the subsequent processing with robots. Initially, the finance department had planned to build a separate system—but by using the UiPath Platform, we were able to prototype quickly. The automation goes beyond simple document reading: it also checks and reconciles data against internal master records, improving the efficiency of verification work." He describes this as evidence of the platform's broad applicability.

In reality, front-line operations such as call handling, and back-office work such as loan assessment and accounting, carry costs running into the tens of billions of yen. By applying the UiPath Platform's AI and robot combination to processes that previously couldn't be automated, Hirano looks ahead: "If we can cut back-office workload in half, that would be transformative." Takahashi echoes this: "There are still plenty of high-variety, low-volume tasks in our internal systems, as well as paper-based procedures that haven't been automated. We'll continue building on our successes as we advance AI adoption."

In an environment where investment in back-office automation and system development is difficult to justify, sustained in-house development remains critical. As Hirano puts it: "I feel strongly that enabling business users who understand our operations to master UiPath is essential to automation. To continuously improve internal-facing services on our own terms, we have no choice but to leverage AI and robots through platforms like UiPath."

Toyota Finance has optimized its use of AI and robots by tailoring their application to the company's own workflows—driven by the necessity of in-house development. Hirano reflects: "With UiPath acting as an intermediary layer on top of our legacy infrastructure, we can now see a clear path to automating our operations. What people used to handle manually is now being organized as logic through UiPath implementation. As we push forward on AI adoption, we are simultaneously working to slim down our legacy systems—with the ultimate goal of rebuilding our operations without being held back by existing systems. UiPath is valuable for organizing and automating work while leveraging existing systems, and it will continue to be essential as we build AI-native workflows."

Underlying all of this is a cultural commitment to ownership. "Outsourcing automation is one option, but we aim to automate our own work ourselves. Our vision is one where our people become the protagonists of digital transformation using tools like UiPath. We want our approximately 2,000 employees to master UiPath as a weapon for shifting their energy toward the work that truly matters—and generating new value from it" (Hirano). With UiPath as the foundation, the company's initiative to combine AI and robots in the right roles at the right times is steadily propelling its digital transformation forward.

Financial operations cannot be controlled by the roll of the dice. A combination of robots reliably handling necessary tasks and generative AI for decision making and text generation was the most suitable approach.

Kenichi Hirano, Head of Digital Business Promotion Department, Toyota Finance