Heralded as the “next big thing” in artificial itelligence when it first hit the market in 2024, agentic AI has the power to reason, plan, use software tools and execute tasks autonomously to achieve goals. Bronagh Riordan, AI and data partner at EY Ireland, says it is best thought of as the move from “AI that advises” to “AI that acts”.
“A useful way to think about it is the difference between asking someone for advice and asking them to complete a task,” she explains. “Traditional AI can tell you what to do, an AI agent can do it for you. For example, instead of helping you plan a business trip, an AI agent could research options, compare flights, book accommodation, update your calendar and prepare your itinerary. Rather than assisting with individual tasks, it can help manage an entire process from start to finish.”
Mark Connor of UCD College of Business explains how it works. “Agents operate through a continuous cycle of thinking, acting and observing. Typically, a user provides an agent with a goal to achieve. The agent’s large language model (LLM) first reasons about the task, determines the most effective approach and develops a plan consisting of a series of actions. It then carries out these actions using the tools available to it, such as searching for information, interacting with software, or accessing external resources.
“After each action, the agent evaluates the outcome, updates its understanding based on the new information and decides on the next step. This think – act – observe cycle repeats until the goal has been achieved or the agent determines that no further progress can be made.”
READ MORE
This simple adaptive loop enables agents to solve complex tasks iteratively by continuously refining their actions based on new information, he adds. “As a result, agentic AI systems can produce more effective outcomes while requiring significantly less human guidance and intervention than traditional large language models.”
Human oversight is still critically important, however. “For trusted, accurate output, AI agents should work in tandem with a human in the loop,” says Martin Duffy, director, head of AI and emerging technologies with PwC Ireland. “Humans should be checking at various points, to ensure the results produced and decisions made are valid.”
The big question is how the technology might affect the way work is done and who does it. “What we are seeing now is routine process work shrinking, with agents handling the admin, chasing and co-ordination between systems,” says AI Ireland founder and executive director Mark Kelly.
“Then the ‘human’ roles shifting from doing tasks to directing and checking agents, more like managing entry level employees. The advice I would offer is to learn how to delegate to AI well. Clear instructions and good judgment become the core skills.”
Kelly believes it will replace some jobs, mainly roles built on repeatable process work such as basic admin, data entry and routine customer service.
“For most roles, it removes tasks rather than jobs. But fewer tasks per role can mean fewer people needed overall. It’s hard to know for sure but it stays assistive where judgment, relationships and accountability matter. Irish businesses should plan for redeployment and reskilling now, not wait.”
Duffy believes we are likely to see AI agents take on more end-to-end responsibilities, co-ordinating complex business processes with minimal human oversight, working alongside other agents, and acting as autonomous digital co-workers that augment human teams and drive greater productivity.
Interestingly, this isn’t translating into job cuts. “In contrast to worries that AI could cause sharp reductions in the number of jobs available, this year’s PwC AI Jobs Barometer shows that jobs are growing in virtually every type of AI-exposed occupation, including highly automatable ones,” Duffy points out.
“AI is amplifying and democratising expertise, enabling employees to multiply their impact and focus on higher-level responsibilities. With the right foundations, both companies and workers can redefine their roles and industries and emerge leaders in their field, particularly as the full gamut of applications becomes clearer.”
The future is augmentation rather than automation and replacement, Riordan contends. “The organisations that realise the greatest value from agentic AI are likely to be those that use it to augment their workforce, helping employees achieve more, make better decisions and tackle challenges that previously required significant time and effort.
“AI agents can handle execution at scale; however, people remain essential for judgment, creativity, leadership and accountability.”
Agentic AI “has the potential to become a force multiplier for human capability”, says Riordan. “By combining human expertise with increasingly capable AI agents, organisations will be able to solve more complex challenges, respond faster to change and unlock new sources of value and innovation. Ultimately, the organisations that benefit most will be those that use agentic AI to elevate the work people do, not simply automate it.”
This may sound overly optimistic, but Connor points out that agentic AI is far from a free lunch. “Early speculation suggested AI would cause widespread job loss as organisations automated tasks,” he notes. “In practice, agentic AI introduces new financial considerations. Because these systems perform multiple reasoning steps, make repeated tool calls, and process more information, they use significantly more tokens than traditional LLMs, raising costs. As a result, organisations must weigh up investing in AI against human talent.”
The technology can also be a driver of jobs growth, Duffy points out. “In Ireland, jobs requiring specific AI skills are growing over five times faster since 2019 than the total jobs market. The research also shows that in Ireland, roles which include AI-related skills tend to be associated with higher advertised salaries, particularly in more AI-exposed industries such as technology and financial services. These roles typically reflect a combination of in-demand capabilities, with AI skills acting as a key enabler that enhances productivity and complements other high-value expertise.”
