Copyright Forbes
Deploying agentic technology enterprisewide will depend on cultivating coding-savvy talent with domain-specific expertise, says UiPath CEO Daniel Dines. There may be no enterprise job today that can’t be enhanced through the use of automation and AI, and agentic AI represents the next step toward fully realizing that potential. So says Daniel Dines, who founded software company UiPath two decades ago with the aim of helping to reduce the burden placed on humans to perform repetitive administrative tasks. “We are at a unique point in history,” says Dines, who is UiPath’s CEO. “Robotic process automation has long helped companies be more efficient by using software to emulate people as they perform well-defined tasks and following rules to achieve the same outcomes. Today, agentic AI offers a tremendous opportunity to extend the possibilities further across the enterprise.” Agentic automation combines AI agents, robotic process automation (RPA), people, and systems to deliver end-to-end AI transformation of dynamic, probabilistic processes, allowing human workers to focus on higher-value tasks, Dines says. In a recent conversation with Ranjit Bawa, chief strategy and technology officer with Deloitte LLP, Dines expounded on his vision for the future of AI in the workplace, including why it will require the cultivation of a “new breed” of developers within the enterprise. What would you say are some of the key differences between agentic AI and RPA in terms of their applicability within the enterprise? RPA is a powerful tool for automating processes that are well-defined and follow prescribed rules. Agentic AI, on the other hand, uses large language models (LLMs), which are better suited for finding patterns in data than following a sequence of well-defined steps. There is a vast amount of work that cannot be expressed in rules. This is what LLMs are made for: They can provide the domain-specific enterprise context decision-makers need to make data actionable. Agents can complete critical business processes and tasks that were not previously possible to automate because they are now able to act independently and make dynamic decisions. What’s more, they can learn from their experience and get better over time. It’s not uncommon to hear people discuss RPA as if it’s being replaced, but that’s not the case at all. Rather, RPA and agentic AI are complementary partners in productivity—it is the combination that will allow enterprises to reap the full potential of AI by enabling end-to-end transformation of processes. Through agentic automation, companies will use both robots and agents to complete work tasks. It will usher in a new era of work. What might be required in this hybrid world, where agentic AI, agentic automation, and other tools coexist within the enterprise? In many ways, agents are like advanced APIs. I think many business applications will provide a framework to help enterprises build agents that connect and live within those applications. What makes the most sense is for agents to be platform-agnostic, so that they can extract the enterprise-specific information needed by decision-makers, regardless of where it lives. Much of the innovation in this space will focus on frameworks for orchestrating the numerous agents and robots likely to be part of many enterprise workflows in the years ahead. Daniel Dines, CEO, UiPath A low-code framework can help people in multiple roles across the enterprise build agents locally—that capability shouldn’t be limited just to professional developers. Deploying agentic technology enterprisewide and reaching the long tail of complex and differentiated use cases will require a new breed of developers who can capture the business context and translate it into examples and prompts that embody that domain-specific knowledge. You might call them “functional-technical” people. It’s less about traditional programming and more a synthesis of business understanding and coding skills. Creating a good prompt and giving the right information to the LLM that will process it is an essential part of building a good agent. What do you believe are some of the most common barriers to adoption? I’d say the two biggest ones are talent and inertia. First, cultivating those functional-technical people within the enterprise will be a key to scaling and capturing the full scope of agentic AI’s potential. Overcoming inertia will also be important—it was an early bottleneck in the adoption of RPA. It’s not enough simply to build an agent; just as important is all the change management involved in empowering people and convincing them to use it. This is not about replacing jobs—it’s about enhancing them so that humans can focus on higher-value work. That’s a ball that needs to be rolled from the top down. Looking ahead, how do you predict this technology landscape will evolve? I think we’re at a cusp in that we have started to see a limit to the compute capabilities of LLMs. The technology is extremely powerful, but companies are realizing it is not going to scale linearly from now on—we must design smarter ways to use what we have and bring more reliability into the picture. That’s one of the reasons agentic AI is capturing so much attention: Many enterprises now see that to fully reap the benefits of LLMs, we have to find a way to deploy them autonomously as part of enterprise workflows and systems, with security and governance built in along the way. It’s still early innings for agentic technology. Much of the innovation in this space will focus on frameworks for orchestrating the numerous agents and robots likely to be part of many enterprise workflows in the years ahead. What advice would you offer technology and business leaders for getting started with agentic technology? Rather than getting stuck in a cycle of perpetual proofs of concept, consider attacking your biggest problem and go for a big outcome. With a significant success in hand, you can then prove that there are not just opportunities to rethink business processes but also potential productivity enhancements and opportunities to uncover new revenue streams. The sooner you get started, the better your position can be in the journey toward those ends.
