UiPath Inc. is putting meat on the bones of the plans it announced last October to enter the agentic artificial intelligence market with today’s rollout of what it calls a comprehensive platform for building and orchestrating teams of AI agents.

The platform is crucial to the robotic process automation company as it struggles to return to its historically high growth rates. Customers’ attention has lately turned to agents that can perform complex tasks with minimal supervision, potentially replacing RPA in many contexts.

UIPath says its years of experience automating digital processes make it well-qualified to address the factors that inhibit organizations from using agents, including security and compliance risks, lack of reliability, scalability limits and fear of vendor lock-in.

“The process-centric focus limits the number of players that can orchestrate end-to-end processes,” said Mark Geene (pictured), senior vice president and general manager of AI products and platform at UiPath. “What we do with Maestro, our orchestration platform, is dynamically orchestrate across agents, robots and humans to execute a long-running workflow.”

Maestro is described as automating, modeling and optimizing complex business processes with built-in process intelligence and real-time key performance indicator monitoring to enable continuous optimization of fleets of agents.

It uses process intelligence, a feature of UiPath’s RPA offering that “understands how a process is running, where the bottlenecks are and how the process is routed,” Geene said. “We find inefficiencies and make recommendations. It’s not a static workflow, but can adapt at runtime.” UiPath provides a set of predefined KPIs and customers can create their own.

Built-in governance

Maestro ensures AI agents operate within clearly defined guardrails of security, predictability and performance with governance, real-time vulnerability assessments, and data access controls built in. Developers can build prototype agents in UiPath Studio using supplied low-code tools and the Python programming language.

“Rules are defined at the agent level as well as the orchestration level,” Geene said. “I give the agent a goal, a context to guide it and policies to work with. We have built-in escalations, which means if the agent isn’t certain what to do or its confidence is below a certain level it can reach out to a person.”

UiPath said its platform also integrates with and supports agents built with third-party agent frameworks from LangChain Inc., Anthropic PBC and Microsoft Corp. It can also automatically create orchestrations from processes defined in Business Process Model and Notation, a standardized graphical notation for business processes..

For document processing, a new Intelligent Xtraction & Processing function performs multi-modal, AI-based classification and extraction of unstructured data for uses like claims adjudication, loan origination and electronic batch records.

UI Agent for computer use, which is in private preview, is a natural language-driven agent that understands user intent, plans multistep tasks and executes actions across interfaces autonomously.

Geene said there is a learning curve to working with agents for developers who are familiar with traditional RPA. “Traditional automation is step-by-step,” he said. “You want to give an agent more autonomy by not giving it a structured workflow. We’ve built a whole set of courses and templates that guide this new design approach and help you decide which type of worker is best for a particular step.”

UiPath’s agentic automation platform is delivered from the company’s cloud and requires no user infrastructure. Pricing is based on the frequency of workload execution and the number of tokens required.

Photo: SiliconANGLE