The following is a guest post from David Hickey, principal with Baker Tilly’s intelligent automation practice. Opinions are the author’s own.
If 2024 was the year of generative AI, 2025 will be the year of agentic AI.
If you haven’t heard of agentic AI, you are not alone. Think of agentic AI as artificial intelligence on steroids, or AI version 3.0. You may have also recently heard the term agent associated with AI. Agent is an apt description – as a super intelligent technology, agentic AI mimics human thinking in its analysis and processes. The technology relies on large language models, natural language models and specialized software for writing and training machine learning algorithms.
The difference between agentic AI, RPA and generative AI
Agentic AI, commonly called agents, autonomously makes decisions and acts towards complex goals with minimal human supervision, combining technologies like large language models and machine learning. In contrast, robotic process automation automates repetitive, rule-based tasks using software bots, requiring human intervention for exceptions.
Generative AI creates new content such as text, images or code based on user prompts, leveraging deep learning to identify patterns in data. While Agentic AI is highly autonomous and decision-focused, RPA is task-oriented and Gen AI excels in generating creative content based on user input.
Agentic AI agents and RPA digital workers can work together to enhance efficiency and decision-making in business processes. Agentic AI can handle complex decision-making tasks and adapt to dynamic environments, while RPA digital workers excel at automating repetitive, rule-based tasks.
For example, an agentic AI agent might analyze data to identify patterns and make strategic decisions, then delegate specific tasks to the RPA digital workers to execute specific tasks. This collaboration allows businesses to leverage the strengths of both technologies, achieving greater automation and more intelligent, responsive operations.
AI agents in action
One way to think of honing agents is to think of rumble strips on the side of a road. Humans provide these guardrails for which the agent is allowed to operate, thus reducing the types of activities the agent is authorized to participate in. When the agent hits a rail, it backs off and informs the user it cannot proceed, alerting a human coworker that it needs assistance. As the human works with the agent, those guard rails expand as the agent learns.
Agents take customer engagement to a whole new level. When spliced with agentic AI, bots can help make decisions beyond a prescriptive matrix or flow chart. Because they are proactive, agents can “think,” reason and adapt to a dynamic environment without needing direction from a human. In fact, agents improve their own thought processes with every iteration of problem solving.
What’s even more interesting is agentic AI can help AI systems set goals so they can work intelligently and more independently. As they continually learn and refine their processes, agents can adopt an organization’s values, brand and perspective.
Agentic AI is spreading across different industries, with early adoption among high-profile companies like IBM, Apple, Siemens, FedEx, Duke Energy, UPS, Tesla, Goldman Sachs and PayPal. The technology is already being used in customer service applications, medical diagnosis and customized healthcare treatment solutions, process automation in business, supply chain optimization in manufacturing, financial and algorithmic trading, smart grids and energy management, self-driving vehicles, climate modeling and environmental protection.
What does agentic AI mean to CFOs?
Agentic AI takes efficiency to the next level as it builds on existing AI platforms with human-like decision-making, relieving employees of monotonous routine tasks, allowing them to focus on more important work.
CFOs will be happy to know that like other forms of AI, agentic is scalable and flexible. For example, organizations can build it into customer-facing applications for a highly customized experience or sophisticated help desk. Or they could embed agentic AI behind the scenes in operations. In many cases, more sophisticated companies will integrate agentic AI in both customer-facing applications and internal operations.
Because they are always learning, self-improving and building on past “experiences,” agents can handle complex, dynamic, changing scenarios. By their very nature, they can help solve evolving real-world problems.
Will agents replace humans or cause workforce displacement?
Similar to earlier versions of AI, we don’t expect AI agents to eliminate human touch. In some complex cases, agent decisions will require human oversight and review of decisions. What we do know is that agentic AI frees up employees’ time so they can do more high-level, strategic and meaningful work. As they roll out their agentic AI strategy, business leaders will also need to be mindful of other ethical concerns beyond job displacement. This includes potential for misuse, unintended consequences, data bias and transparency and data governance.
Not surprisingly, like other emerging technologies, agentic AI requires thoughtful and strategic implementation. This means starting with process identification and determining which specific process or functions are suitable for agentic AI.
Business leaders also need to determine organizational value and impact and find ways to evaluate and measure to ensure the technology is delivering clear benefits. Companies should also be mindful of team composition, and, if necessary, secure external experts to ensure successful implementation. Beyond the technical feasibility, there are other considerations such as data security.
For now, CFOs and other business leaders need to wrap their heads around the concept of “agents” and keep their minds open to how this powerful technology can best serve the needs of their organization. Whether they are ready or not, the new, exciting phase of agentic AI is upon us
