Make the leap to light speed with agentic automation

Since its inception, the story of automation has been one of linear progress. From robotic process automation (RPA) to intelligent document processing (IDP), each new innovation has made enterprises more productive, profitable, and agile.

But now, something bigger is happening. A new technology—agentic automation—is disrupting this cycle of incremental improvement, unlocking workflow reinvention the likes of which we’ve never seen.

Rather than focusing on individual tasks, agentic automation leverages robots, AI, and orchestration to tackle entire workflows from start to finish. At the heart of agentic automation are AI agents—autonomous software entities that can perceive their environment, navigate unfamiliar territory, and make decisions on the fly.

The blueprint for capturing agentic automation’s value was the theme of our recent quarterly webcast, UiPath Live: The Path to Agentic Automation. We had the privilege of picking the brains of an all-star panel—AI scientists, automation experts, and an enterprise leader—on what this transformation means for businesses now and in the future.

Most leaders haven’t yet grasped the scale of agentic automation’s potential. But the select few that have are primed and ready to reap the greatest rewards from this game-changing technology.

Three areas where agentic automation is redefining how work gets done

While traditional automation remains vital for structured, rules-based tasks, agentic automation thrives where unpredictability reigns. Here’s how it’s bridging gaps that once seemed unbridgeable.

1. From rules to reasoning

How many of your business processes rely on inputs in varying formats? What about less-than-perfect data?

For most organizations, the answer is far too many.

Historically, automation struggled with data inconsistencies. If information didn’t arrive in a structured, standardized package, employees had to step in—translating data between systems, cleaning up discrepancies, and manually reformatting files. While their time could be better spent elsewhere, these irregularities left them with no other option. That is, until AI agents entered the picture.

AI agents thrive in ambiguity. They don't need perfect data to function. Dr. Edward Challis, Head of AI Strategy at UiPath, explained to UiPath Live hosts Mary Tetlow and Geoff Anderson that “agents offer a really powerful way to address tasks where data is constantly changing.” Instead of requiring employees to manually structure every input, AI agents can take a high-level goal and determine the best way to process messy, incomplete, or inconsistent data.

This skill is particularly valuable in industries where data formats vary widely. Take WEX, a global financial technology provider that processes a massive volume of healthcare claims every day. Structured claims—clean digital submissions with standardized fields—are easily handled by RPA. But plenty of others arrive in messier formats, like handwritten physician notes or blurry forms. In the past, employees had to manually make sense of this chaos before claims could be processed, leading to frustrating delays for customers and higher costs for WEX.

Now, AI agents handle this variability automatically. They extract key details, cross-check them against compliance requirements, and escalate only the most complex cases to human teams.

2. From task automation to workflow reinvention

Varying inputs aren’t the only barrier to widespread enterprise automation. There’s also overwhelming process complexity. Too many conditional "if" statements, too many variations, too many exceptions…at a certain point, trying to automate workflows like these with traditional methods becomes unwieldy. “It’s just too time consuming to define the process that automation would have to run for every scenario,” Dr. Challis noted.

Agentic automation takes a fundamentally different approach. Instead of following a script, agents reason through problems—determining what matters, what needs attention, and what can be processed autonomously.

But they don’t work in isolation. AI agents are at their best when part of an ecosystem that blends human intuition, robotic precision, and agentic adaptability. For instance, an agent might break down a complex workflow into subtasks: RPA bots handle data entry, APIs pull real-time market prices, and humans resolve edge cases.

Dr. Challis likened this process to baking a cake. “When you're executing that recipe, you have a lot of choice around which butter you're going to use, how you get the butter, which bowl you're going to use,” he said. “So, there's that flexibility between agency and a defined routine of how that process is going to be done.”

This balance between structure and adaptability is well-suited to a number of persistent enterprise challenges. Anti-money laundering (AML) systems, for instance, are an essential part of financial institutions’ security frameworks. But they’re far from perfect—traditional tools have a tendency to drown compliance teams in false alerts. In some cases, these can be as high as 90%, forcing teams to sift through lots of noise to identify actual risks.

AI agents excel at filtering out false positives to identify real threats. They do this by analyzing both structured and unstructured data—spotting patterns like small, repeated transfers that signal real risks. As Live guest Craig Le Clair, Vice President and Principal Analyst at Forrester pointed out, agents have been shown to reduce false positives in this process by 60%. This has freed compliance teams to focus on high-priority investigations rather than being bogged down in unnecessary reviews.

2. From coding to collaborating

To realize their full value as collaborative partners, employees need to be able to communicate with AI agents in natural language. Large language models (LLMs) like ChatGPT were supposed to unlock this kind of partnership, but for non-programmers, the reality has been more complicated.

Even though LLMs respond to plain-language prompts, getting them to deliver the right results requires some programming knowledge—namely, a deep understanding of problem decomposition and code validation. Simply put, knowing what to ask for is only half the battle. You also need to know if what you get back is correct.

I'm a programmer, so if an LLM gives me back 100 lines of code, I can read that and figure out if that is actually doing what I want it to be doing. On the other hand, if my non-programmer friend tries to do the same thing, it’s really tough to figure out if the output is actually what she wanted. You need a lot of expertise to even check.

Dr. Sarah E. Chasins, Principal Applied Scientist at UiPath

Agentic automation dissolves this barrier. Employees can describe their goals—even if they’re vague—and let agents translate those directives into action. This has been transformative for WEX, as teams no longer need to translate business requirements into rigid logic.

Instead of having to spend weeks going through every deterministic scenario, I'm able to communicate to my developers the goal that I want. What is this business output that I'm looking for? And then I'm able to actually see these things working together. So it's created this openness between the product and the tech teams that we haven't had in traditional coding in the past.

Emily Krohne, Enterprise Automation Principal at WEX

These capabilities are impressive…but how can I be sure AI agents won’t go rogue?

Let’s address the elephant in the room: with the ability to function across systems and processes, how can I be sure agents do what I want them to do?

It's a valid concern. Agents are non-deterministic, and their inherent unpredictability is part of what makes them so powerful. Implementing agentic automation safely requires the right guardrails to ensure agents operate reliably, securely, and transparently.

Keeping humans in the loop

Agents operate with a certain level of autonomy, but that doesn't mean they should run unchecked. In most enterprise use cases, they’ll serve as decision support tools rather than fully autonomous actors. Dr. Challis made this clear: “For the next few years, agents will do some research and make proposals, but a human will need to review it. Before any big changes are made, we’re going to have a human checkpoint.”

Continuous monitoring

Visibility is everything. To ensure AI agents operate as intended, organizations need real-time monitoring both during design and runtime.

Zach Eslami, Senior Manager of Product Management at UiPath, reinforced this point: “[transparency is] a key aspect of making sure your agent is operating well in a siloed environment as well as the outside world.” Organizations need visibility into how agents make decisions so they can refine their performance over time and ensure they stay aligned with business objectives.

Pairing AI agents with deterministic RPA bots is one of the best ways to maintain control. While agents adapt and make context-driven decisions, RPA bots follow strict, rules-based logic—creating a balance between flexibility and predictability. “We believe that our agents demonstrate a new level of controlled agency because they’re not just interacting with tools and applications,” Eslami explained. “They’re not coming up with plans on their own. They’re able to leverage humans and robots to create a new level of determinism to their output and make sure they’re operating in the manner that our users and our customers expect.”

Partnering with trusted vendors

Trust is the foundation of effective enterprise automation. Not just trust in the technology, but in the partnerships that bring it to life. Krohne mentioned that WEX’s “history with UiPath RPA tools let us scale agents faster.”

When AI agents are introduced on top of an existing automation infrastructure, they don’t feel disruptive. Instead, they become a natural extension of what businesses are already doing—enhancing workflows without overhauling them. Eslami reinforced this point: “ultimately, we see agents being able to build on top of the existing UiPath Platform™. That means they're able to leverage all the amazing automation tools that we have available.”

But laying the groundwork is just the first step. To unlock the full potential of agentic automation, enterprises need a way to build, deploy, and manage AI agents at scale.

UiPath Agent Builder gives teams the tools to design agents that cater to their business needs. With built-in monitoring and governance, enterprises can deploy AI agents with confidence—knowing they’ll work as intended from day one. Join the Agent Builder waitlist today to be among the first to explore the next era of automation.

And, for an even deeper dive into the present and future of agentic automation, check out the full UiPath Live: The Path to Agentic Automation episode, now available on demand.