Use Cases Only Possible with Agent Builder

Japan Forum
1

Hello
I have been using UiPath Studio so far, but I am considering trying out UiPath Agent Builder (hereafter referred to as Agent Builder) in Studio Web.
When I look for use cases for Agent Builder online, I find the following two examples:

  • Assignment of internal personnel: By flagging information about “past troubles between specific members,” it seems like this could also be managed with UiPath Studio without using Agent Builder.
  • Checking for inconsistencies in invoices and expenses: Detecting errors and fraudulent data in exceptional patterns also seems feasible with UiPath Studio.
    If anyone knows of use cases that can only be accomplished with Agent Builder, please share.

• Assignment of internal personnel: Determine and suggest the optimal personnel by considering not only operational status and skills but also concerns about human relationships that are difficult to formalize, such as “there had been troubles between specific members in the past.”
• Invoice and expense inconsistency checks: AI detects exceptional pattern errors and fraudulent data, and creates an escalation screen to prompt the person in charge for a final check.

2

Hi @gorby,

  • Your observation is valid. Both examples can be implemented using Studio, with suitable rules, data and AI integrations. I wouldn’t describe them as use cases that only Agent Builder can deliver.

  • The key difference is how the next action is decided. In a conventional Studio workflow, you define the execution paths. With Agent Builder, you configure a goal, instructions, context and permitted tools, allowing the agent to choose its next step based on the information it receives.

  • For personnel assignment: If availability, skills and known conflicts are structured fields, rules or an optimisation algorithm may be sufficient. An agent could help interpret project requirements and documented feedback, explain trade-offs and propose a shortlist for a manager to review. It cannot reliably infer personal relationships without relevant evidence.

  • For invoice checks: Duplicate invoices, amount mismatches and approval limits are well suited to deterministic checks. An agent becomes useful when investigating an exception across supplier emails, purchase orders and policy documents, then recommending a next step with supporting evidence. A suspicious finding should remain a flag for investigation, not a confirmed fraud decision.

  • A practical example to try: An invoice dispute assistant that reviews the supplier’s complaint, retrieves invoice and delivery details, identifies missing evidence and recommends whether to request clarification or escalate. The investigation path can change depending on what it finds.

  • Agent Builder’s practical advantage is the integrated development experience: prompts, tools, context, human escalations and evaluations. You could assemble similar capabilities through Studio and external services, but that would involve additional implementation work.

  • One clarification: the escalation screen is configured through an Action app in Action Center. The agent raises the escalation and supplies relevant details; it does not automatically invent the approval screen at runtime.

  • My recommendation is to start with one exception-heavy investigation step. Let the agent review context and recommend an action, keep calculations and system updates in Studio, and retain human approval where required.