What is the following big thing in enterprise automation? If you ask the tech giants, they’re agents – powered by generative AI.
There is not any universally accepted definition of “, but today the term is used to explain generative AI-powered tools that may perform complex tasks through human-like interactions across software and web platforms.
For example, an agent might create an itinerary by entering a customer's information on airline and hotel chain web sites. Or an agent could order the most cost effective ride to a location by mechanically comparing prices across apps.
Providers sense opportunities. ChatGPT manufacturer OpenAI is allegedly Dive deep into the event of AI agent systems. And Google unveiled a series of agent-like products at its annual Cloud Next conference in early April.
“Companies should prepare for widespread adoption of autonomous agents today,” analysts at Boston Consulting Group wrote in a recent report report – citing experts who estimate that autonomous agents will develop into mainstream in three to 5 years.
Old school automation
So where does that leave RPA?
Robotic Process Automation (RPA) got here into vogue over a decade ago as firms turned to the technology to spice up their digital transformation efforts while reducing costs. Like an agent, RPA drives workflow automation. However, it’s a way more rigid form based on preset “if-then” rules for processes that may be broken down into strictly defined, discretized steps.
“RPA can mimic human actions reminiscent of clicking, typing, or copying and pasting to perform tasks faster and more accurately than humans,” explained Saikat Ray, VP Analyst at Gartner, in an interview with TechCrunch. “RPA bots, nevertheless, have limitations with regards to tackling complex, creative or dynamic tasks that require natural language processing or reasoning skills.”
This rigidity makes RPA expensive to construct – and significantly limits its applicability.
A 2022 Opinion poll from Robocorp, an RPA provider, finds that 69% of firms that say they’ve adopted RPA struggle with broken automation workflows not less than once every week – and these issues often take hours to resolve. Entire businesses have been made out of helping firms manage their RPA installations and forestall them from breaking.
RPA vendors aren’t naive. They are aware of the challenges – and consider that generative AI could solve lots of them without hastening the demise of their platforms. In the eyes of RPA vendors, RPA and generative AI-powered agents can coexist peacefully – and maybe sooner or later even grow and complement one another.
Generative AI automation
UiPath, one among the larger players within the RPA market with an estimated greater than 10,000 customers, including Uber, Enslin calls it “one-click digital transformation.”
“These capabilities provide customers with generative AI models trained for his or her specific tasks,” Enslin told TechCrunch. “Our generative AI enables workloads reminiscent of text completion for emails, categorization, image recognition, language translation, the power to filter out personally identifiable information (and) quickly answer any questions on people topics based on insights from internal data.”
One of UiPath's more moderen explorations into generative AI is Clipboard AI, which mixes UiPath's platform with third-party models from OpenAI, Google and others to, as Enslin puts it, “bring the facility of automation to anyone who copies.” / Paste.” Clipboard AI allows users to spotlight data from a form and – using generative AI to determine the suitable places for the copied data – point it to a different form, app, table or database .
Photo credit: UiPath
“UiPath recognizes the necessity to bring motion and AI together; This is where value is created,” said Enslin. “We consider one of the best performance comes from those who mix generative AI and human judgment – what we call human-in-the-loop – across end-to-end processes.”
Automation Anywhere, UiPath's important competitor, can also be attempting to integrate generative AI into its RPA technologies.
Last 12 months, Automation Anywhere launched generative AI-powered tools to construct natural language workflows, summarize content, extract data from documents and, perhaps most significantly, adapt to changes in apps that will normally fail RPA automation.
“(Our generative AI models are) developed based on (open) large language models and trained with anonymized metadata from greater than 150 million automation processes across 1000’s of enterprise applications,” said Peter White, SVP of enterprise AI and automation at Automation Anywhere. said TechCrunch. “We proceed to construct custom machine learning models for specific tasks inside our platform and at the moment are also constructing custom models based on basic generative AI models using our automation datasets.”
Next generation RPA
Ray points out that it's vital to concentrate on the restrictions of generative AI – namely biases and hallucinations – because it enables a growing variety of RPA capabilities. But risks aside, he believes generative AI will add value to RPA by changing how these platforms work and “creating latest opportunities for automation.”
“Generative AI is a strong technology that may improve the capabilities of RPA platforms by allowing them to grasp and generate natural language, automate content creation, improve decision making, and even generate code,” Ray said. “By integrating generative AI models, RPA platforms can provide more value to their customers, increase their productivity and efficiency, and expand their use cases and applications.”
Craig Le Clair, principal analyst at Forrester, sees RPA platforms as ripe for expansion into autonomous agents and generative AI as their use cases expand. In fact, he expects RPA platforms to remodel into comprehensive automation toolsets – toolsets that help deliver RPA on top of related generative AI technologies.
“RPA platforms have the architecture to administer 1000’s of task automations, and that bodes well for centralized management of AI agents,” he said. “Thousands of firms are well-versed in RPA platforms and can be able to leverage them for generative AI-powered agents. RPA has grown partly due to its ability to simply fit into existing work patterns through UI integration, and this may proceed to be useful for smarter agents in the longer term.”
UiPath is already taking its first steps on this direction with a brand new feature, Context Grounding, which was previewed earlier this month. As Enslin explained it to me, context grounding is meant to enhance the accuracy of generative AI models—each first-party and third-party—by converting business data that those models might depend on into an “optimized” format that is simpler to index and may be searched.
“Context grounding extracts information from company-specific data sets, reminiscent of a knowledge base or internal policies and procedures, to create more accurate and insightful answers,” Enslin said.
“If there’s one thing holding RPA vendors back, it’s the ever-present temptation to lock in customers,” said Le Clair. He emphasized the necessity for platforms to “remain agnostic” and offer tools that may be configured to work with a spread of current – and future – enterprise systems and workflows.
Enslin promised that UiPath would remain “open, flexible and accountable.”
“The way forward for AI would require a mixture of specialised AI with generative AI,” he continued. “We want customers to give you the option to make use of all sorts of AI safely.”
White wasn't necessarily committed to neutrality. However, he emphasized that Automation Anywhere's roadmap is heavily influenced by customer feedback.
“What we’re hearing from all customers across all industries is that their ability to integrate automation into many more use cases has increased exponentially through generative AI,” he said. “By incorporating generative AI into intelligent automation technologies reminiscent of RPA, we see the potential for firms to scale back their operational costs and increase productivity. Companies that don’t adopt these technologies will struggle to compete with others that use generative AI and automation.”
