Is the afterlife of RPA agentic, as UiPath suggests?

HFS has repeatedly declared robotic process engineering (RPA) dead, even before UiPath hit the big time with its funding rounds and IPO. Suffice it to say the suggestion was not to pronounce RPA dead but to highlight that broader innovative technologies delivered as an integrated platform were required to progress to end-to-end process automation.

The market has moved on from those giddy heights, and RPA vendors are looking for new ways to keep the proposition relevant. When UiPath struggled to justify its irrational market capitalization, it moved beyond its RPA centricity and “one bot for every employee” postulation by repositioning around an enterprise automation platform. Yet, even that was not enough to placate investors, resulting in Daniel Dines returning as CEO to try and chart a new direction for the company. HFS caught up with its management at the London UiPath on Tour event.

Back to the future or charting a brave new world?

The event’s focal point was not the details of UiPath’s roadmap but the return of its founder as CEO and a new corporate positioning. The return of Daniel Dines should be seen as a reflection of investors’ unhappiness with the company’s sluggish performance at a time when the market is in overdrive on all things AI. Thus, probably unsurprisingly, UiPath is going all in on agentic process automation as its new North Star.

Daniel defines agentic as “the ability of an AI system to design and control the flow of a business process.” He expects a future where agents do 80% of the work, with humans doing 20%, primarily focused on supervision and exception handling. The broader context is, as executives put it, the emergence of GenAI provides an opportunity to rethink end-to-end process automation. Therefore, for HFS, the strategic questions are: how do we finally progress to end-to-end automation, what role can AI play on that journey, and will RPA remain a relevant proposition?

Exhibit 1: Charting the future of RPA

Source: UiPath, 2024

What is the “why” for automation (and UiPath)?

Many of these AI discussions are squarely playing to the investors’ gallery. By strongly leading with North Star topics such as large action models, autonomous decision-making, and adaptive behavior, as Exhibit 1 depicts, UiPath might get some breathing space while re-assessing its go-to-market strategy and overhauling the automation narrative. But it needs to define its role in this brave new world.

We didn’t hear much about technology, process, and cultural debt. Overcoming that is the critical value proposition of RPA. We listened to a great deal of talk about agentic models but little about how robots and agents collaborate. The crossover between the two concepts will be an increasingly busy segment as many ISVs are zeroing in on workflow orchestration. Here, UiPath will be locking horns with the likes of ServiceNow, Salesforce, Workato, Camunda, and many more.

The more important thing for UiPath is to articulate the “why” of the Golden Circle model. What is your purpose, your motivation, and what do you believe? And here, UiPath has to double-click when driving its new narrative by leading with outcomes rather than technology and capabilities.

The challenges on the journey toward agentic automation

We are very early on the journey toward agentic automation. There is much education and learning required from the early deployments. However, we should try to learn from the journeys of RPA and, consequently, intelligent automation. Here are the issues on our minds:

• Is agentic just “cognitive” in disguise? If you exchange “agentic” with “cognitive,” you can easily argue that we have been here before with intelligent automation. Yet, this time, we are pivoting from intelligent document processing (IDP) and off-the-shelf hyperscaler modules to the wondrous world of GenAI. Ultimately, it is about automating the insights and inputs of workflows. Therefore, we should be careful about suggesting that AI will make workflows redundant and that there will be a new quality of automation through GenAI without clearly articulating how.
• Progress toward end-to-end automation: While focusing on end-to-end automation is welcome, the proof of success remains to be seen. We need a better understanding of how to coordinate disparate agents. We have heard many aspirations that those agents could perform dynamic planning, learning, and execution, but for that to occur, we need new classes of algorithms.
• Governance: In HFS’s view, governance is the most significant challenge on the journey toward agentic automation. An automation CoE is not the answer for democratizing AI and orchestrating agents. The complexity of the data and process issues cannot easily be democratized. Enterprises appreciate innovative ideas, but control over outcomes is essential. Thus, the focus must be on ensuring the outcomes of the activities GenAI delivers.
• What can we learn from AIOps? Lessons from the more mature AIOps space could be essential to finding our bearing on the claims of agentic automation. Fundamentally, there is no self-healing where machines fix issues that were not known before. Typically, it denotes the next-best action recommendations. Similarly, GenAI has not led to the reinvention of runbooks (the equivalent of RPA in many IT processes). The most significant change may be the improved MTTR due to better analysis. Therefore, we shouldn’t expect a “revolution” but rather “evolution” through the augmentation of agents.

The Bottom Line: To be relevant to operations leaders, we must understand how to operationalize agentic automation.

It is fascinating that, just as with the early days of automation, executives point to a seminal study by McKinsey highlighting the opportunities with AI. Yet both those reports also discuss the potential disruptions in great detail. Therefore, the answers to many questions on automation and AI are in the eye of the beholder.

The world of RPA is mainly deterministic, often even static. Thus, any suggestions that agentic automation could be largely non-deterministic outline the fundamental change this would bring to organizations. That’s why we need to better understand how to operationalize these innovations. That includes articulating the “why.” How do we prepare organizations to take advantage of innovations such as agentic process automation? This is not about capabilities but about change and outcomes.