RPA Adapts to AI Agents for Future Relevance

RPA vendors like UiPath & Automation Anywhere add AI features to compete with generative AI agents, they see AI as complementary, not a threat.

What will emerge as the subsequent significant development in enterprise automation? The tech titans respond that it is agents propelled by generative AI. 

While a universally agreed-upon definition of “agent” remains elusive, the term is currently employed to refer to tools propelled by generative AI capable of executing intricate tasks via software and web-based interactions resembling humans. 

An agent might, for instance, generate an itinerary for a client by inputting the client’s information into airline and hotel chains’ websites. Alternatively, an agent may determine the most cost-effective ride-hailing service to a given location through an automated price comparison across various applications. 

Vendors perceive the potential for success. Purportedly, OpenAI, the developer of ChatGPT, is at work on AI agent systems. Google also showcased various agent-like products in early April during its annual Cloud Next conference. 

Analysts at Boston Consulting Group recently published a report titled “Companies should begin preparing for the widespread adoption of autonomous agents today.” The report cited experts who estimate that autonomous agents will enter the mainstream within three to five years. 

Traditional Automation

Where does this leave RPA, then? 

Enterprises began adopting robotic process automation (RPA) more than ten years ago to reduce costs and enhance their digital transformation initiatives. Similar to an agent, RPA automates workflows. The form, however, is considerably more inflexible, as it relies on “if-then” preset principles that govern processes that can be decomposed into discretized, precisely defined steps.”RPA can replicate human actions, such as clicking, typing, or copying and pasting, to complete tasks more quickly and precisely than humans,” explained Saikat Ray, vice president analyst at Gartner, in an interview with TechCrunch. “However, RPA bots cannot perform dynamic, creative, or complex tasks that demand reasoning or natural language processing capabilities.” 

The inflexibility of RPA renders its development costly and significantly restricts its practicality. 

According to a survey conducted in 2022 by Robocorp, an RPA vendor, 69% of organizations that claim to have adopted RPA encounter malfunctioning automation workflows at least once per week, with the majority requiring hours to repair. Assisting organizations in managing their RPA installations and averting breaches has generated substantial revenue. 

Suppliers of RPA are not naive. They are aware of the obstacles and are confident that generative AI could circumvent a significant portion of them without hastening the dissolution of their platforms. RPA and generative AI-powered agents can coexist harmoniously in the eyes of RPA purveyors; they may even develop into complementary entities in the future. 

Generative Automation by AI

With an approximate customer base exceeding 10,000, UiPath, a prominent contender in the Robotic Process Automation (RPA) industry, recently unveiled novel generative AI functionalities centered on message and document processing. Additionally, these functionalities execute automated actions to accomplish what UiPath CEO Bob Enslin calls “one-click digital transformation.” 

“These functionalities furnish clients with task-specific generative AI models,” Enslin explained to TechCrunch. Our generative AI powers workloads such as email text completion, image detection, language translation, filtering personally identifiable information, and rapidly answering people-related inquiries utilizing knowledge from internal data.” 

Enslin describes Clipboard AI, one of UiPath’s recent forays into generative AI, as an endeavor that “brings the power of automation to anyone who has to copy/paste.” Clipboard AI integrates UiPath’s platform with third-party models from OpenAI, Google, and others. Clipboard AI enables users to highlight data from a form and direct it to another form, app, spreadsheet, or database. It utilizes generative AI to determine the appropriate destinations for the copied data.

Enslin stated, “UiPath recognizes the need to integrate AI and action; this is where value is generated.” “We believe end-to-end processes will benefit most from systems that integrate generative AI and human judgment—human-in-the-loop systems.” 

UiPath’s primary competitor, Automation Anywhere, is also endeavoring to integrate generative AI into its RPA technologies.

Automation Anywhere introduced tools powered by generative AI last year. These tools enable users to generate workflows from natural language, extract data from documents, summarize content, and — most significantly — adapt to application changes that would ordinarily fail in an RPA automation. 

Peter White, SVP of enterprise AI and Automation at Automation Anywhere, told TechCrunch, “[Our generative AI models] are trained with anonymized metadata from over 150 million automation processes across thousands of enterprise applications and are built upon [open] large language models.” “We continue to develop bespoke machine learning models for particular tasks on our platform and are currently constructing bespoke models using our automation datasets on top of foundational generative AI models.” 

Advanced RPA

Ray emphasizes the significance of remaining aware of the constraints of generative AI, specifically biases and hallucinations, as it increasingly drives an array of RPA functionalities. Despite the inherent dangers, he is convinced that generative AI can enhance RPA by revolutionizing how these platforms operate and “generating new opportunities for automation.” 

Ray stated, “Generative AI is a potent technology that can augment the functionalities of Robotic Process Automation (RPA) platforms, allowing for the generation of code, comprehension, generation of natural language, and automation of content creation.” By incorporating generative AI models, RPA platforms can provide more excellent value to their clients, boost their productivity and efficiency, and broaden their use cases and applications.” 

Craig Le Clair, principal analyst at Forrester, believes that as the use cases for RPA platforms expand, they can support autonomous agents and generative AI. Indeed, he foresees RPA platforms transforming into comprehensive automation toolkits that facilitate the deployment of RPA and other generative AI technologies. 

He stated, “RPA platforms are architecturally capable of overseeing thousands of task automation, which bodes well for the centralized management of AI agents.” “Thousands of organizations already familiar with RPA platforms will be receptive to utilizing them for agents infused with generative AI.” “UI integration has contributed to RPA’s expansion by facilitating its seamless integration with existing work patterns; this characteristic will continue to be valuable for developing more intelligent agents in the future.” 

Already beginning to move in this direction, UiPath introduced Context Grounding, a new feature that entered preview earlier this month. As Enslin elucidated, Context Grounding is intended to enhance the precision of generative AI models—both original and third-party—by transforming business data that the models may utilize into an “optimized” format, thereby facilitating indexing and searching. 

“To generate more precise and insightful responses, Context Grounding extracts data from company-specific repositories, such as an internal policies and procedures database or a knowledge base,” explained Enslin. 

Le Clair stated that if anything impedes RPA vendors, there is a constant temptation to lock consumers in. The speaker emphasized the importance of platforms maintaining an impartial stance and providing configurable tools compatible with various existing and forthcoming enterprise systems and workflows. 

Enslin responded that UiPath would continue to be “responsible, flexible, and transparent.” 

He continued, “The future of AI will require a combination of generative AI and specialized AI.” “We want clients to be able to utilize all types of AI confidently.” 

White did not explicitly pledge neutrality. However, he emphasized that customer feedback significantly influences Automation Anywhere’s roadmap. 

“With generative AI, our customers across all industries report that their ability to automate a greater number of use cases has grown exponentially,” he explained. “Organizations can increase productivity and decrease operating expenses by integrating generative AI into intelligent automation technologies such as RPA.” Organizations that must incorporate these technologies will encounter difficulties competing with those that fully embrace generative AI and Automation.