AI Agents Aren’t Taking Jobs, They’re Remaking Them

Agentic AI has been the flavor of the month among some of the biggest companies in the world and has dominated predictions for the New Year. Nvidia CEO Jensen Huang, speaking at CES 2025, proclaimed: “The age of agentic AI is here.” OpenAI CEO Sam Altman thinks 2025 may be the year the first “virtual employees” start entering the workforce. Major players have heavily backed AI agents and the role they can play in changing the nature of the workforce. With the new American administration reportedly briefed on day one about super-intelligent agentic AI breakthroughs, it is clear that AI is atop the agenda, both politically and financially.

Like previous emerging technologies, the pace at which AI is being adopted into the lives and workflows of consumers is slower than the constantly shifting discourse around it. According to a recent Prosper Insights & Analytics survey, more than a third of Gen-Z (36.8%) and more than 3 in 10 Millennials (31%) are already using Generative Artificial Intelligence.

The use cases of generative AI are varied. The Prosper Insights & Analytics survey shows that the most popular current use case of AI is research, with 43.3% of respondents stating they use AI to generate summaries or provide relevant information. Yet far fewer people are leveraging AI to execute tasks, with only 23.2% of people using AI as a personal assistant.

Leaders from across the field argue adoption of AI agents that are capable of processing complex instructions and completing repetitive tasks will have a major impact on the workforce in the coming years. However, some industry experts argue that consumers do not commonly understand agentic AI. Daniel Vassilev, co-founder and CEO of Relevance AI, a San Francisco and Sydney-based startup that has already rolled out agentic AI solutions to a plethora of corporate clients, argues that “it will take time for the nuances and differences between different forms of AI to be commonly understood.”

Vassilev and others in the industry pin this knowledge gap on the “confusion between LLMs, agents, and traditional software,” stemming from a wide proliferation of “AI-enabled” products on the market. The ever-shifting sands of the industry continue to move the goalposts, as solutions that were once labelled as “co-pilot,” and therefore inherently reliant on human intervention, are now being branded as “agentic.”

Vassilev notes that it is “unsurprising” that co-pilot solutions are being rebranded, upgraded, or replaced by agentic alternatives, as “co-pilot solutions are very much a short-term trend” seeing as “any activity we can do on autopilot will be done that way.” The shift towards the truly agentic economy is seen across the ecosystem as one of the next major steps in the development of AI.

So, when truly agentic systems are readily available, to both enterprise and consumer markets, what does this mean for the future of work? Joel Hellermark, founder and CEO of Sana, which uses AI to streamline repetitive tasks for office workers by integrating with internal databases and business software tools, argues that the mass adoption of agentic AI will lead to a “new Cambrian explosion of knowledge” with AI agents “removing bottlenecks in understanding and synthesis”.

“What once required teams of specialists might soon be accomplished by individuals, augmented by their AI agents. In this era, the boundaries of what a single human mind can achieve will no longer be defined by biology, but by the extent of their AI agents' data and compute,” Hellermark predicts.

Experts in the AI space are in agreement that time-saving automation like those both Relevance and Sana have rolled out to enterprise customers, will allow people to focus on other areas of their job. Vassilev sees the agentic future of work as “getting more value out of the existing workforce by taking out the mundanity of their jobs.”

Yet much of the discourse around agentic AI has focused on job replacement, in part led by futuristic visions of AI that have proliferated visual media this century so far. From his experience selling agentic AI solutions to enterprise clients, Vassilev argues against the prevailing wind, “across the clients we work with, we have seen an increased headcount.”

The important question for developers at the moment isn’t around how this technology substantively impacts the future of work, but how it can get there. Experts argue that the future of agentic AI is dependent on the improvement in reasoning ability. Kai Zhou, founder and CEO of NetMind.AI, a data infrastructure company hosting a decentralized computing network providing rentable GPUs worldwide, sees the “coordination conundrum” of managing the actions of increasingly larger, more complex systems as one of the main hurdles to overcome. With this conundrum potentially leading to “inefficiencies, conflicts, and even system-wide failures if not properly managed,” Zhou’s outlook on the future is that “ensuring that multiple autonomous agents work harmoniously towards a common goal, especially in dynamic environments, is a monumental task.”

Another major barrier to developing AI agents is the lack of computing power available. Companies worldwide are trying to keep up with the vast amount of GPUs needed to power more advanced AI models, and Big Tech is no exception. In October, Sam Altman explained that the lack of compute capacity was behind the company’s delay in releasing new products. Companies like NetMind.AI, which directly address the lack of computing power by offering GPUs as a service, are becoming increasingly popular among startups and developers as a way to roll out more sophisticated AI innovations like agentic AI and remain competitive against Big Tech.

If 2024 was the year of AI chatbots, 2025 is the year of AI agents. Yet, there is still a disconnect between the power of this emerging technology and the consumer understanding of it. Bridging this gap, and changing the public perception and understanding of AI, and agentic AI in particular, is a major hurdle to its widespread adoption. It’s a case of re-imagining the AI world of the future, because as Vassilev notes “these changes aren't devoid of humans - quite the opposite. The AI workforce is for the human workforce.”