UiPath’s Dines Says AI Needs a Manual and Launches Cartographer to Write It
Daniel Dines, founder and CEO of UiPath (explore this topic), shared at the UiPath's FUSION conference that Large Language Models (LLMs) lack several crucial abilities:
- Consequence: LLMs don't bear the consequences of their actions, unlike many enterprise processes.
- Learning: They do not learn from their actions; they only have memory.
- Exactness: LLMs are probabilistic, whereas many business operations demand accuracy every time.
Dines believes these limitations pose significant challenges for deploying trustworthy autonomous AI. To address this, UiPath has introduced UiPath Cartographer, a tool that creates a "Map of Work" detailing how a company's processes actually run.
The Need for a Manual
Dines argued that without a comprehensive manual, it will be difficult to successfully deploy autonomous AI:
“Without this manual, nobody will really succeed in deploying autonomous AI that we can trust to make decisions and understand how we actually work.”
How Cartographer Works
Cartographer utilizes process charts, process-mining diagrams, documents, and standard operating procedures to generate an initial Map of Work. Coding agents then build automations based on this map, and human reviewers ensure accuracy by correcting suggestions. These corrections are documented in a Decision Ledger, continually updating the map with new exceptions.
Customer Success Stories
SMBC, Medline, and Mayo Clinic have already embraced Cartographer. Kei Yamamoto from SMBC shared that the bank has automated over six million hours since 2017, freeing approximately 3,000 employees for alternative roles.
UiPath Cartographer is now available, though pricing information was not disclosed.