exclamatory icon
Warning: Browser Not Supported
You are using a browser version that we do not support. Please use one of these supported browsers to ensure the best experience on this site:
Chrome
Download
Edge
Download
Safari
Download
Firefox
Download

Article: Defining tech in 2025 — AI agents

Writtten by Mercer's Jason Averbook, Ravin Jesuthasan, Adriana O'Kain and Jess Von Bank

Artificial intelligence (AI) keeps evolving — and with agentic AI, the digital agents joining our ranks will soon take automation to the next level. Despite slow take-up in HR and mixed reports about the return on investment (ROI), global executives see AI as the top business priority and the biggest potential value driver in 2025. Agentic AI is transforming company systems, processes and the workforce, and HR leaders who don’t prepare will be left behind.

To clarify, agentic AI isn’t exactly new. It already shapes our lives and can take many forms, from simple thermostats to air traffic control systems. But the next generation of agents running on large language models (LLMs) can function in a way that seems almost human.

The shift from automation to agency

Agentic AI is a complex, autonomous system that uses a mix of technologies to achieve goals without relying on constant human input. It learns, perceives, reasons, plans and even makes decisions by combining:

  • The efficiency of robotic process automation (RPA)
  • The natural language processing (NLP) of chatbots
  • The trial-and-error of machine learning (ML)
  • The processing power of LLMs
  • The data-driven logic of predictive AI
  • The novel outputs of generative AI (Gen AI)

The complementary features of these technologies are critical to delivering the benefits of agentic AI:

Technology Core function Benefits Challenges HR use cases
Robotic Process Automation (RPA) Handles fixed, repetitive tasks More accurate and efficient vs. humans alone Requires present rules and strict processes Administrative tasks such as data entry and scheduling
Traditional Chatbot Interacts with users via scripted dialogue Reduces the cost of high-volume, low-value interactions Limited to specific inputs and responses Routine employee interactions, such as policy inquiries
Natural Language Processing (NLP) Allows programs to understand human language Makes technology more useful and accessible Bias, errors and upredictability Training HR tools on company data
Large Language Model (LLM) Analyzes and creates text Processes data at high speed and scale Bias, errors and unpredictability Various AI-powered tools
Predictive AI Analyzes data to find patterns and make predictions Supports objective, informed decisions Accuracy depends on a range of factors Workforce and benefits planning
Generative AI (Gen AI) Creates new content in response to user prompts Unlocks efficiency, knowledge and creativity at scale Needs constant input to function; poses multiple risks Employee engagement and HR content creation
Agentic AI Acts on its own to achieve goals Supports business objectives with minimal input Could make unexpected or risky decisions Combining and optimizing all of the above to execute on processes

 

Modern agentic AI is a giant leap forward. Other systems and bots can be reactive and input-dependent, like new hires awaiting their next assignment. AI agents are proactive, developing strategies, optimizing workflows, finding solutions and executing tasks to meet business objectives — all with minimal supervision.

The next level of human-machine teaming

The true power of agentic AI isn’t just in automation; it's in augmentation. While other types of AI are excellent tools for specific tasks, agentic AI is uniquely suited to augment the workforce more broadly. Agents can mimic human reasoning, interactions and decision-making to effectively work alongside human employees, rather than just supporting them. When we blend human creativity with AI’s analytical capabilities, we create a workforce that’s greater than the sum of its parts.

"Nearly half (47%) of executives believe rethinking talent strategies in light of increased AI use will deliver ROI in 2025."
— 2025 Executive Outlook, Mercer

As companies focus on the tools and capabilities they truly need, vendors and in-house teams are building AI agents that specialize in different functions. These agents might collaborate via multi-agent models, or even human-machine teams that work toward a shared goal.

Here are some common types of agents, and how they could work with humans:

  • Conversational agents support a range of self-service tasks, from workers’ time-off requests to managers checking employee availability.
  • Functional agents can take on certain personas or roles, like a recruiting agent to screen candidates or explain role requirements.
  • Supervisory agents might delegate tasks to other agents, assess workforce metrics in real time, and create reports for human managers or request a human-in-the-loop when needed.
  • Utility agents support tasks at the direction of supervisory agents, like completing enrollment forms or sending routine notifications.
  • All agents could, in theory, request human input when needed to mitigate risk — though it’s unlikely that every agent will be designed this way

Navigating this fresh division of labor requires work design: Deconstructing jobs into tasks; identifying which tasks are best performed by humans, technology or a combination of the two; and reconstructing new ways of working. The rise of agents allows for human input that’s less constant and more thoughtful, and thus fundamentally changes the nature of work. Many employees will pivot from task execution and prompt engineering to applying uniquely human skills such as empathy, creativity and ethical AI governance.

Enterprise tech and the human workforce are not the same

As AI continues to function more like humans, discussing AI in human terms can help explain and understand it. NVIDIA CEO Jensen Huang went so far as to say that IT departments will become HR for AI agents. It’s an interesting comparison, but one that is inherently wrong; this false equivalency between tools and talent diminishes humans and the role of work in humanity.

"84% of HR leaders predict the HR function will become more automated and tech-enabled."
— 2024 Voice of the CHRO, Mercer

Let’s be clear: The human and digital parts of our workforce are not the same. IT will certainly handle the more technical, tactical side of agent adoption and management. But in a climate of cyber risk, cost optimization and skills shortages, the rise of agentic AI poses new challenges and opportunities for HR to consider.

Change management. Even positive change can bring stress and uncertainty, which in turn can harm trust, engagement and productivity. As organizations consider AI investments and new workflows, it’s important to make a detailed plan — and to clearly communicate to the workforce how these changes could affect them.

Digital transformation. Becoming digital goes beyond choosing the right agents. For organizations to see ROI from their technology spend, employees need to fully embrace these tools and co-create new ways of working. HR can stimulate this change by helping leaders foster a culture of continuous learning and experimentation, continuously designing and redesigning work to optimize human-machine capabilities.

Employee experience (EX). The most successful implementations of agentic AI don’t just streamline processes — they optimize experiences. Agents adapt to user behavior and other conditions in real time, without the need for human intervention. When AI handles the routine, humans can focus on relationships, creativity and strategic thinking.

Ethical AI governance. Connecting agents to company data, systems and operations could also put organizations at risk. While agents can self-optimize and act on their own, this autonomy leaves room for harmful decisions and unintended consequences. Creating safeguards and keeping humans in the loop will be essential.

Skills-powered organizations. Agentic AI is changing the nature of work, and it accelerates the shift among employers from jobs-based work models to skills-powered organizations. By understanding how tasks are being substituted or augmented and the implications for the skills of the workforce, employers can better adapt to changing market demands.

Strategic workforce planning.  As skills become the currency of work, strategic workforce planning shifts from being an infrequent, headcount-based process to a dynamic, work- and skills-powered capability that is tightly integrated with the business planning process.

"Designing talent processes around skills is one of HR leaders’ top three priorities in 2025."
— 2024-2025 Global Talent Trends, Mercer

While HR and IT clearly aren’t merging anytime soon, managing human-machine teams will require both functions to work together and learn from each other. IT might focus more on orchestrating the optimal combination of technologies; HR can help business leaders redesign work around AI’s capabilities while ensuring the seamless development and deployment of the new skills required.

These joint efforts might also include a system of record for agents. Much like human capital management (HCM) software or a human resources information system (HRIS), an agent system of record can help organizations manage in-house and third-party agents, from onboarding to monitoring to compliance. This becomes more important as more agents get added to the company roster.

Charting a path to thrive with agentic AI

This is the Year of Agentic AI — and with HR functions on track to show early achievements, the future of work is brighter than ever. Executives think AI and automation will have the biggest impact on cost optimization this year, so conversations around cost and efficiency can help secure buy-in for agents. But the greatest return comes from empowering people to do more impactful work.

As business leaders double down on AI this year, maxing out ROI requires a full-scale digital transformation. It’s about realigning every part of the business around AI: core systems and processes, innovation and strategy, insights and reporting, analytics and decision-making, team structures and tech stacks, governance and compliance, investments and risk management.

Many organizations have begun their AI journey, and they’re ready to keep scaling. If you want to better understand or even deploy AI agents, here’s what to consider:

  1. Lead with the problem you already have: The secret to HR Tech success applies to agentic AI, too. Implementing tech is one thing, but deploying for outcomes is another. Business strategy leads technology strategy, which then leads decisions around tools and use cases. Avoid shiny object syndrome by leading with the problem you're trying to solve.
  2. Technology doesn't change the way you work, you do: Any technology will fall short of intended value if we fail to design for change enablement, data fitness, strong governance and continuous improvement. This requires a rethink around work and role design, measuring productivity, and optimizing the digital workforce experience.
  3. Cost vs. value: We’re all still learning how agentic AI will scale, including how it's priced and who will absorb the costs. If a vendor offers agentic AI capabilities, prepare for cost models to evolve over time. And lean into workforce analytics, skills-powered work models, and a tech-enabled HR function to drive short-term impact and long-term value.
  4. Rethink everything. Agents are only as good as the data, systems and workflows they connect to. Engaging in broader efforts like work design and digital transformation can help maximize the return on your AI investments.

Ready to master the future of work with agentic AI? Get in touch and let’s talk deployment.