# UiPath’s new tool could unlock a much bigger wave of automated business processes > Source: > Published: 2026-09-24 13:06:27+00:00 A decade ago, UiPath built its business on watching people work, recording keystrokes and clicks to automate tasks employees already did by hand. This week, amid an industry consumed by autonomous agents, the company’s newest product suggests that same instinct still has a place. UiPath used its Fusion conference in Las Vegas to launch Cartographer, a tool for building what the company calls a Map of Work: a living, governed record of the informal, undocumented knowledge that keeps enterprise processes running, the exceptions, workarounds, and judgment calls that never make it into an official workflow document. Existing automation tools capture the click-by-click steps of a task. Cartographer works at the process level instead, documenting how work actually gets done versus how it’s designed to run. From the main stage, CEO [Daniel Dines](https://www.uipath.com/blog/daniel-dines) described this information as a missing manual enterprises need before trusting agents with real work. “A process map tells you how work is designed to run,” he said. “UiPath Cartographer is the key to uncovering the operational knowledge of how that work actually runs.” Even with Cartographer, the first drafts of a Map of Work “will be incomplete for sure,” Dines admitted on stage. That missing knowledge has kept enterprises’ most complex, exception-heavy processes out of automation’s reach, said [Mark Geene](https://www.linkedin.com/in/markgeene/), UiPath’s group senior vice president of agentic solutions. He characterized Cartographer as the next stage in a five-year push to extend the automation platform into messier enterprise areas. Capturing that knowledge, if it works, opens the door to automating processes once too undocumented and judgment-dependent to hand over to machines at all. How well it will work is a different question. Analysts and customers on the ground described a tool that’s promising on paper, but whose real test comes later, in whether enterprises do the disciplined work of capturing and maintaining what Cartographer helps them uncover. That capture process has a real limit, said [Amy Loomis](https://my.idc.com/getdoc.jsp?containerId=PRF005268), a research director at IDC, one that depends on whether people actually share what they know. “That step will only be as strong as the tribal knowledge of teams and the ability to capture it accurately,” she said. Pairing deterministic guardrails with the inherently unpredictable behavior of large language models (LLMs) differentiates UiPath’s approach, Loomis said. She pointed to the spectrum of options UiPath will offer: prebuilt maps ready out of the box at one end, a curated Process Atlas in the middle, and Cartographer itself handling processes requiring full customization. A human “cartographer” role, an existing business analyst repositioned to own a Map of Work, is the other new piece, she said. Getting a full picture of a process has been a persistent problem, one that often doesn’t surface until well after a bot is live, said [Andrea Simpson](https://www.linkedin.com/in/andrea-simpson-11802773/), IT manager of automation at USI Insurance Services. “So much knowledge is in a person’s head, and it can be hard to get it out of our subject matter experts,” she said. Knowledge management — a problem enterprises have been [dealing with since the cold war](https://www.cio.com/article/254938/enterprise-software-knowledge-management-km-how-to-beat-the-baby-boomer-retirement-blues.html) — is the pain point UiPath aims to ease. But there are limits. Some knowledge will likely stay out of reach, Simpson said. USI is an early adopter of Maestro, UiPath’s orchestration tool, to automate its marketing processes, tying together producer meetings, carrier submissions, and client communication. But the biggest hole is what happens off the record: a phone call, a conversation at someone’s desk. “We wouldn’t want that automated,” she said. “We want to keep that human.” That knowledge never touches a company system until a producer chooses to report it. Perfection won’t be required to get started. “The goal isn’t to understand all the rules and processes from all the years you’ve been doing a process,” said UiPath’s Geene. “The goal is to capture enough information to get agents executing the process. The key is to get live early.” Typically, once an LLM is built and deployed, “their brains freeze,” said [Doug Marcey](https://coronishealth.com/leadership/doug-marcey/), CTO of Coronis Health, a healthcare revenue-cycle-management company that has used UiPath for about a year and a half. Cartographer’s promise, he said, is closer to on-the-job training for agents, gaining institutional knowledge the way a new hire gradually would, with each new exception routed through a “Decision Ledger” for a human owner to approve before feeding back into the map. That trail matters as much as the automation itself in a regulated industry. “The difficulty in healthcare is that compliance is all about vibes,” Marcey said. “The Decision Ledger would be very helpful for us.” Coronis’s denial-management work shows what captured judgment can do. Agents draw on claims data, orchestration coordinates the workflow, and generative AI drafts appeal letters, a combination that’s made it financially viable to contest denials on smaller claims that previously would have been written off, Marcey said. Another UiPath healthcare customer was losing money on tens of millions of denied claims because generating an appeal could take 45 to 75 minutes, so it often wouldn’t happen despite a 50%-to-60% overturn rate on filed appeals, Geene said. With agents built on a Cartographer-style Map of Work, that process now takes four to five minutes. UiPath’s human cartographer role suggests the tool is meant to reshape jobs, not eliminate them. “This isn’t about replacing people,” said IDC’s Loomis. At USI, much of a business analyst’s time today goes into repetitive work, translating what a stakeholder describes into documentation a developer can build from, work Simpson is eager to hand off. “It’s going to save so much time,” she said. “It’ll grow my team’s knowledge and improve the outcome of our processes.” How it all shakes out will come down to people, as with any tech-driven transformation. “The hard problem is the change management at the organizational level,” said Marcey of Coronis Health. And the accountability will remain with humans as well. “We can’t go to clients and say, ‘Hey, the bots screwed up,’” Marcey said.