Robotic process automation used to be a simple pitch: point a piece of software at a repetitive, rules-based task and let it click through the same steps a human would. That pitch is getting complicated. This week's news cycle shows an industry trying to reconcile two decades of screen-scraping bots with a wave of autonomous, reasoning AI agents — and the strategies vendors are choosing look surprisingly different from each other.

UiPath's FUSION event: mapping the messy reality of work

The biggest developments came out of UiPath's FUSION customer conference in Las Vegas this week. UiPath founder and CEO Daniel Dines introduced Cartographer from the keynote stage, calling it a release that builds what the company calls a "Map of Work," a written record of how a business process runs, including the exceptions and judgment calls employees handle but rarely document. The logic is straightforward: written process documentation rarely accurately captures how processes actually run, and it leaves out the workarounds, the reasoning behind decisions, and undocumented tribal knowledge that keeps operations functioning.

Cartographer doesn't just capture that knowledge passively. The Cartographer Agent absorbs existing documentation and can be tasked to interview subject matter experts directly, watch screen recordings, and ask follow-up questions as it goes, synthesizing the results into a process map an analyst can interact with. Alongside it, UiPath for Coding Agents is now generally available across the full platform, letting developers build, test, and operate workflows using any coding agent — including UiPath's own agents or third-party tools like Claude Code, Cursor, Codex, and Antigravity, all natively fluent in UiPath's folders, jobs, queues, and audit logs.

The company also split its orchestration engine: UiPath Maestro is now two complementary products, Maestro Orchestrate for long-running, complex business processes and Maestro Automate for high-volume tasks. And a new desktop agent, UiPath Delegate, is a governed agent that completes tasks described in plain language across applications and business systems, under the same identity, permissions, and audit controls as the rest of the platform. Notably, UiPath's own research found enterprise leaders cite data quality and readiness (38%), integration with existing workflows (37%), and governance and compliance (33%) as the top barriers to scaling agentic AI — which is essentially the problem Cartographer and the governance updates are aimed at solving.

The bigger industry fight: is RPA being replaced or absorbed?

Zoom out from one vendor's roadmap, and a sharper debate is playing out across the automation industry: whether classic RPA bots survive the agentic AI era as a foundation layer, or get swept aside. Some newer, AI-native platforms argue traditional bots are a liability, pointing out that legacy tools relying on UI selectors require significant ongoing maintenance and specialized developers. Others in the space frame it differently, arguing traditional automation isn't disappearing but is instead becoming the reliable scaffolding on which new AI agents operate.

That tension shows up in market forecasts too. Multiple analyst reports published in just the past few days peg the RPA market's growth in the 25% range annually through the end of the decade, explicitly attributing continued expansion to RPA's fusion with AI, hyperautomation strategies, and cognitive automation rather than to standalone bots.

Why it matters

Taken together, this week's news suggests the RPA category isn't vanishing — it's being rebuilt as infrastructure for something bigger. Vendors betting on "AI-native" replacement and vendors betting on "AI-augmented" incumbency are both racing toward the same destination: autonomous agents that can act on institutional knowledge, not just scripted steps. The winners will likely be decided by which approach enterprises trust enough to hand over their messiest, most exception-riddled workflows.

Sources