Last month, I attended UiPath's Forward conference, an invitation following my Forbes article on AI agents. The discussions revealed compelling insights into the burgeoning field of agentic AI applications and their implications for enterprise workflows.
UiPath, a prominent player in Robotic Process Automation (RPA), traditionally focused on digitising highly structured business processes. RPA, often involving software rather than physical robots, streamlines operations, boosting accuracy, speed, and cost-efficiency. It excels in tasks typically considered tedious or undesirable for humans, such as converting paper documents into digital records. UiPath possesses substantial intellectual property in the machine learning and document processing crucial for effective RPA implementation.
The core theme of Forward was UiPath's "Chapter 3.0," a strategic shift towards agentic applications alongside its existing RPA capabilities. This is particularly interesting given demonstrations from competitors like Salesforce and ServiceNow, which suggest that AI agents might supplant RPA. However, conversations with UiPath CTO Raghu Malpani convinced me that UiPath's approach is unique and potentially transformative.
UiPath unveiled several enhancements to its RPA platform, central to which is the integration of AI and agentic programming as an extension, not a replacement, of its existing technology. Unlike some platforms focusing on individual or departmental agents, UiPath emphasises enterprise-wide application. This is reflected in three key new capabilities:
Firstly, Agent Builder provides a more structured, guided development experience. While retaining a low-code approach, it offers richer controls, moving from a "blank sheet of paper" to a more intuitive form-based system. This contextual framework simplifies development by providing structure for workflows, integration points, and access controls.
Secondly, Connector Builder tackles the scalability challenge. Secure, cross-platform integration is essential for scaling agents. Connector Builder acts as an API-based integration hub, ensuring governed and manageable service for agentic applications.
Thirdly, Autopilot for Everyone empowers business users to create their own agents. A noteworthy aspect of UiPathâs approach is its recognition of the synergy between personal and enterprise agents. UiPath Autopilot allows users to build personal agents capable of invoking enterprise agents as needed, facilitating a workflow assembly rather than a purely developmental approach.
Recent demonstrations have showcased AI agents managing workflows, with large language models (LLMs) driving decision-making. This is undeniably useful, particularly for ambiguous processes requiring exceptions to established rules. The low-code/no-code nature, aided by LLMs, simplifies agent creation and deployment.
However, LLMs aren't deterministic; they thrive in ambiguity, unlike RPA processes built on determinism. Both types of applications remain vital, especially for large-scale processes. Indeed, an AI agent could invoke an RPA processâor vice versaâcombining the strengths of both for tasks demanding both deterministic execution and less-structured decision-making, such as mortgage approvals.
The reception to UiPath's announcements was enthusiastic. Discussions with users revealed a strong interest in integrating agents to enhance existing deployments. One particularly illuminating conversation concerned a user with an RPA system generating numerous exceptions needing manual intervention. The user believed that many exceptions were easily resolvable by an agent, thereby minimising human involvement. This highlights the collaborative potential of agents and RPA in enhancing speed and quality.
Given the continued viability of both RPA and agentic applications, the possibility of shared services and infrastructure becomes crucial. UiPathâs existing platform offers inherent advantages. Security, governance, workflow management, and monitoring are established features. A unified platform addresses concerns surrounding application sprawl, increased attack surfaces, and maximising code reuseâissues that have been raised regarding other platforms and general-purpose development tools for agentic applications.
Existing UiPath customers stand to benefit significantly from simpler deployment and management. However, potential new customers will need to consider the same questions raised by other platforms with integrated agentic capabilities. Specifically, will integrating a new platform for agent creation prove more cost-effective than utilising general-purpose toolsets from cloud providers like AWS or IBM?
UiPath's pragmatic approach, shared by its users, was refreshing. Rather than presenting agents as a replacement for RPA, UiPath champions their use as an augmentation, enhancing existing functionality. This balanced perspective avoids the pitfalls of viewing agents as either a silver bullet or an existential threat.
For UiPath's existing client base, adoption will likely be seamless. However, to attract new customers, UiPath might need to invest in enhanced market education and more assertive sales strategies. This investment could prove invaluable, as UiPath's vision focuses on delivering short-term value while facilitating long-term process evolution, rather than disruptive revolution.