Daniel Dines, Co-Founder and CEO, UiPath.
Robots have evolved. From the first black and white movies (Metropolis) almost exactly 100 years ago, robots finally became a reality and started helping to weld rivets in automotive engineering plants around the 1970s. They then evolved into robot-like humanoids, with Japanese companies innovating with proof-of-concept walking robots, but these have yet to make it into our homes. Then, in the post-millennial era, software robots or “bots” arrived as the era of robotic process automation began to flourish and our software systems were given new autonomous capabilities to perform not only their own maintenance functions, but also useful productive business tasks.
Despite this lip service to our brief history of robots, the current state of robots and RPA may still be a mix of software and hardware, but it is first and foremost a balanced equation between software.
Bridging the AI Gap
John Kelleher, UiPath’s Regional Vice President for the UK and Ireland, wants to push our notion of robots beyond many pre-existing notions of automation.
“We all know the impact of artificial intelligence. The problem is that there is a gap between the promise of AI and the reality of the environments deployed today,” Kelleher said. “If we look at McKinsey’s analysis of this area, we see that companies have too narrow a view of how AI is applied to work. In addition to the technology choice issues, there is also a key education aspect to overcome and this must work with data architectures that enable change management to have a real impact on how IT works with the business function. In order to bridge the gap between the promise of AI and the real-world operational environments, we are looking at how companies can build the operating model of the future to truly create a new way of working.”
Today, we might think of AI as ubiquitous in the sense that it could potentially be applied to any aspect of the business. Kelleher argues that AI should not be viewed as a single application deployment point in this sense. Rather, it should be viewed as a fabric that can be applied to specifically defined solutions that span an end-to-end operating model in a modern enterprise.
It seems clear that UiPath’s work in the UK and Ireland reflects and resonates with its work in North America and the rest of the world. As CEO and founder of UiPath, Daniel Dines has more opinions on robotics than most. Speaking of what he calls the “dark ages of AI and RPA” (and he’s talking about 2016), Dines says his team understood early on that the two disciplines are symbiotic.
What is agent-based technology?
“Now that the future of AI and RPA is agentic, let’s take a step back and think about what the term means and why it applies to the next era of automation. Simply put, agentic refers to the ability of an AI system to control and manage the flow of a business process,” Dines explained, speaking to press and analysts in London this month. “Many human actions (thinking, moving, blinking, etc.) happen automatically without us having to expend any cognitive power to perform them – and I mention this because human intelligence is fueled by these automated routines.”
For more details, we can define ourselves as agents. When a human operator in a call center or a business person at any level performs a function, we can call them agents. Nowadays, we more generally speak of software agents, that is, a software (or a complete computer program) that performs a prescribed task for a human user, a machine or another virtually defined entity that is part of a workflow system. So, in this context, a human-agent team is a collaborative environment composed of interacting humans working in concert with AI systems.
This clarification is necessary because Dines paints a picture of a world in which human agents are increasingly beginning to delegate their tasks to AI and RPA. The next question is what aspects of our lives remain our responsibility, as opposed to those we can entrust to computers.
Left Brain – Right Brain
Dines asks us to think more about automation. He explains that if we are thirsty during the night, we simply grab a bottle of water from the fridge and drink. This whole action is automatic, learned and done without thinking, so (as we have already discussed) it is all left-brain automatic thinking. Left-brain thinking is about robotics where people create and maintain automation – this is where the structured, logical, systematic processing that is oriented towards efficiency occurs.
The right brain is entirely dedicated to creative intuitive thinking, where we apply adaptability and manage ambiguity. It is also in the right brain that we find all large-scale action models, autonomous decision-making with the ability to manage adaptive behavior.
It’s only possible to bring this intelligence into RPA if we have an AI model that can understand exceptions as they happen, so that it can learn and handle real-world ambiguities more competently. This is where we can add “agent skills” to RPA and start channeling that intelligence from what UiPath would call self-healing robots into dynamic planning and dynamic learning in businesses.
Our future as an agent
“We can imagine AI in the agentic future evolving to a point where it’s able to handle about 80% of what humans do at work. That’s where AI starts to play a more spontaneous (perhaps unstructured and unexpected) role in enterprises (or even in applications like healthcare and any vertical), UiPath’s Dines said. “You might wonder what it takes to get to 100% offloading to AI agents. I don’t think we’re at that point with AI today, although we can offload some instances and tasks where that can happen (self-driving cars for example), but that’s only in deployment scenarios that have relatively controlled, defined, and safe environments. So we have some work to do, but the future is certainly very exciting.”
The company’s comments come at a time when the company has been making direct updates to its platform. UiPath has now integrated several new features into its platform designed to more deeply integrate generative AI into the UiPath business automation platform. UiPath Autopilot for developers and testers uses generative AI and natural language processing in UiPath Studio to create workflows, generate expressions, and help build automations.
“New features include Text to Workflow: Developers can simply describe an automation idea in natural language and Autopilot will create the initial workflow. Text to Expressions: With Autopilot, developers no longer need to remember the exact syntax and structure of expressions. They can describe what they need in natural language and let AI generate the correct expression. Text to Code: Generate code from natural language descriptions to reduce the time to deploy automation projects,” the company said in a technical statement about the product.
A preview of the UiPath plugin and integration with Copilot for Microsoft 365 provides an integration to enable users to automate end-to-end business processes with their colleagues directly within Microsoft Teams. Customers will have access to a pre-built automation library to run automations that perform common, repetitive tasks, as well as specialized automations for job- or industry-specific tasks. Users can also discover and run automations developed by their company.
Workers and agents in unison
Robotics is booming, automation is expanding its scope, evolving its adaptability, expanding its range of capabilities, and expanding its application in businesses. With this rise, AI-powered software agents are now taking over more of our work tasks. This is not a bad thing, it’s actually a good thing – as a UiPath spokesperson said: “If you can apply RPA and AI to your role and find a way to completely automate your work, then you’re clearly smart and that’s the surest way to get promoted.”
However, whatever the outcome of applying AI agents and RPA automation to businesses, the changing dynamics here must surely leave management consultants rubbing their hands with glee. If we can also automate this (sometimes questionable) role, then AI will have truly come of age.
