UiPath CTO Details ‘Office Layout’ For Agents, Robots And Humans

Posted by Adrian Bridgwater, Senior Contributor | 8 hours ago | /ai, /cloud, /innovation, AI, Cloud, Innovation, standard | Views: 9


Workers need desks. Somewhere inside the interminable firehose stream of AI messages emanating from the technology industry, this basic truth provides us with a lesson for structuring the use of AI agents alongside robotic process automation and its software “robots” in the modern workplace.

Stepping aside from the “your next coworker is an AI service” hypecycle for a second, we can understand where software automation in its various forms should sit if we consider which intelligence service does what. With an understanding of which “desk” (i.e. specific role and function in an enterprise’s operational workflow) an agent or an RPA robot sits at, we can start to design the so-called office of the future and actually graft these services into production-grade software deployments.

What Is Controlled Agency?

Borrowing a term directly from employment law and management practices, an approach known as “controlled agency” can help us determine who (or what) does what in the workplace.

Back in the real world, controlled agency is a term often applied to healthcare, construction and IT contractors to help mandate their critical, legal and functional responsibilities. It governs what they do, how and where they do it… and it helps lay down a regulation for what they should not do.

In our AI world, controlled agency is becoming a key mechanism for delegating work across different automation and intelligence services in much the same way. There are software robots for filling out forms and performing administrative work (clerical jobs, if you will), there are agentic services that can help with business decision-making (business analysts, or management consultants if we must) and there are bleeding edge AI services working to get us to us to Mars (esoteric researchers). Crucially, while all those intelligence services might work in the same building, they should all get a separate desk and a different office security passkey.

Working to apply this precise principle across his firm’s estate of technologies is Raghu Malpani in his role as chief technology officer at agentic automation company UiPath. With its track record in robotic process automation, the organization may be argued to be in a good position to comment on the intersection of RPA and agentic AI.

With Great Power, Comes Great Auditability

“At UiPath, the notion of controlled agency is our methodology for delivering AI agents that act with clarity, context and compliance. It’s not enough for AI to be powerful; it must be dependable, auditable and aligned with enterprise goals,” stated Malpani. “UiPath approaches controlled agency by embedding AI agents within structured workflows, where deterministic tasks are handled by [software]

robots and only non-deterministic tasks are delegated to agents. This ensures agents are used where adaptive decision-making is actually needed. Agents are designed to be single-minded i.e. focused on narrow, well-scoped objectives. Complex workflows are composed of multiple such agents alongside deterministic automation, preserving clarity and modularity.”

To manage risk, Malpani advises that agents must operate within strict guardrails. This means the use of constrained inputs and outputs, policy enforcement, real-time monitoring and clear escalation paths to humans for exceptions or judgment calls. This ensures every agent operates safely within enterprise boundaries, whether retrieving data, updating systems, or interacting with users. The UiPath platform has been engineered to support this with built-in tools for orchestration, governance and evaluation.

From Prototype To Production

“Additionally, through context grounding, each agent uses the right knowledge at the right time. With dynamic evaluations, agents are tested in real-world scenarios to ensure performance, accuracy and trustworthiness before deployment. A scoring and governance model enables software application developers and systems administrators to measure readiness and enforce standards at scale,” detailed Malpani, in a direct press briefing this month. “This methodology turns generative AI from a prototype into a production-grade solution – one that enhances productivity, reduces risk and integrates with your an organization’s existing workforce and systems.”

Malpani’s more detailed comments on automation intelligence control for business-critical (and indeed life-critical) applications, processes and systems come in line after his firm detailed updates to its platform. The company says it wants to address what it calls the key blockers to deployments in this space because conversational AI and agent-based assistants have “demonstrated isolated value”, when scaling across the enterprise. Key blockers include security and compliance risks, lack of reliability, stalled pilot programs and fear of vendor lock-in.

The Second Act For AI

“With this [year’s] launch, we fully enter our second act,” said Daniel Dines, founder and CEO of UiPath. “We’ve built a platform that unifies AI, RPA and human decision making so companies can deliver smarter, more resilient workflows without added complexity. As models and chips commoditize, the value of AI moves up the stack to orchestration and intelligence.”

UiPath Maestro is the orchestration layer that automates, models and optimizes complex business processes end-to-end with built-in process intelligence and key performance indicator monitoring to enable continuous optimization. Maestro provides the centralized oversight needed to scale AI-powered agents across systems and teams. Developers can prototype agents in UiPath Agent Builder within UiPath Studio, while having the opportunity to customize when needed. This means both technically oriented business professionals and experienced programmers can create automations that can adapt to complex business requirements. Additionally, UiPath IXP (intelligent [e]xtraction and processing) has introduced multi-modal, AI-based classification and extraction for unstructured data. Built for high-complexity use cases like claims adjudication, loan origination and electronic batch records, IXP is said to bring enterprise-grade scale to document processing.

Agentic Orchestration, Competitive Analysis

Possibly the next battleground in AI (goodness knows the tech industry loves a new AI substrate to lay claim to dominance in, just imagine the press release opportunties), the road to agentic orchestration and management is heavily paved, possibly with gold. With its undeniable heritage in automation, UiPath is seen by many as one of the first to push forward in this space, but it’s not the only one.

Logically, there are agentic process automation controls found with Automation Anywhere, a company that also places emphasis on workflow analytics via its process discovery options, which come with a meaty side order of governance features. Microsoft Copilot Studio is all about workflow builds and analytics, Blue Prism is very much in the enterprise-grade RPA space, with reasonable scalability and compliance controls, Moveworks operates in this market and SnapLogic offers agentic functionality through its AgentCreator tool, although this is arguably a more data-centric orchestration play… and IBM’s watsonx Orchestrate makes sure the “IBM has a version of everything” mantra holds true.

As we now start to find a seat for automation intelligence of all kinds in the workplace and workflow, knowing where each service actually sits and creating a more accurately defined job description for these virtual teammates may provide a way to integrate them into production environments more effectively. Agents and robots don’t take up much space in the office canteen, but they still need to be told which cubicle to go back to after lunch.



Forbes

Leave a Reply

Your email address will not be published. Required fields are marked *