Article

Rise of agentic AI: Promise, peril and the future of work

By:
Miles Davis
insight featured image
Agentic AI is unlocking major competitive gains but it's also racing ahead of human control. Miles Davis reveals how it's reshaping business and why upskilling in AI governance is key to staying relevant and in control.
Contents

The corporate world is getting a new colleague. Not a precocious graduate trainee or a well-heeled consultant, but something altogether stranger: agentic AI. 

These aren't the chatbots that politely answer a prompt and disappear into the ether. They're AI agents that can set objectives, plan tasks, adapt to feedback, and execute actions across multiple domains with unnerving persistence. They are, in effect, autonomous business agents – and they're arriving faster than most companies, regulators, or employees are prepared for.  

The implications are profound.

Competitive firms are already embedding agentic AI

In retail, for instance, such systems already ingest real-time streams of demand signals – from weather forecasts to TikTok trends – and autonomously adjust prices or restock shelves across a national chain. 

In manufacturing, inspection agents pore over camera feeds, identify defects, and reroute workflows before human operators even notice. 

In finance, reconciliation agents knit together ledgers and transaction systems, flagging anomalies and generating forecasts, often with a speed and accuracy that would once have required a team of accountants. 

These aren't proof of concepts. They're live, functioning systems embedded in the day-to-day operations of competitive firms.    

Powerful, unpredictable and potentially dangerous 
But here's the rub: autonomy cuts both ways. The same qualities that make agentic systems useful also make them unpredictable:  

  • A pricing agent, left unchecked, could collude algorithmically with competitors 
  • A quality-control agent might reject batches for spurious reasons, triggering costly supply-chain ripples 
  • A finance agent could make reconciliation decisions so fast and so confidently that humans fail to notice a systemic error until it is too late  

Unlike traditional automation, which does what it's told, agentic AI does what it infers is best. The distinction is subtle but vital – and it makes these systems both powerful and potentially dangerous.    

Navigating agentic AI in the workplace

This poses a knotty problem for business leaders. Ignoring agentic AI isn't an option: competitors who embrace it will enjoy lower costs, faster cycles, and sharper customer insight. But embracing it uncritically is reckless. 

Regulators have yet to grasp the consequences of machines that act with purpose rather than instruction. Labour markets are equally unprepared: a generation of professionals could find their bread-and-butter tasks – the reconciliations, inspections, adjustments – absorbed by agents that neither tire nor sleep. The risk is not redundancy, but irrelevance, unless workers learn to shift their value towards framing problems, applying judgement, and governing systems.    

Professional training is an option but it must evolve as quickly as the technology itself. Courses such as our AI & Digital Skills for Business Impact – particularly its module on Applied Generative AI – can provide a targeted way for employees to skill up. Learners are taught not just how to prompt a model, but how to design agentic systems responsibly: how to set boundaries, govern decisions, and navigate the ethical and legal pitfalls. For early-career professionals, this is an opportunity to leapfrog into roles that demand digital fluency. For mid-career managers, it's a chance to remain relevant in a world where routine analysis is increasingly automated. And for change leaders, it provides the frameworks needed to harness autonomy without surrendering control.    

The prudent path

Yet even with education and foresight, the risks remain. Gartner predicts that more than 40% of agentic AI projects will be abandoned within two years – victims of over-promising, under-delivering, or simply being too difficult to govern. This 'agent washing' – branding simple software as autonomy – threatens to discredit the field before it matures. Worse, a poorly designed system deployed at scale could trigger not just financial losses, but reputational and regulatory crises.    

The prudent path lies between naïve enthusiasm and Luddite resistance. Companies should adopt modular architectures that allow agents to act, but within transparent guardrails. Governments must move quickly to update regulation, ensuring accountability in decisions made by autonomous systems. And individuals must re-skill, shifting from repetitive execution to higher-order oversight.

Agentic AI is not a passing fad. It's the logical extension of automation, analytics, and machine learning – capable of transforming how work is organised and how value is created. But whether it becomes a partner in prosperity or a liability will depend less on the brilliance of its algorithms than on the wisdom with which humans choose to apply it.    

For now, the technology’s trajectory is clear: businesses are inviting these new agents into their operations. The question is whether they will also invite the foresight, governance, and humility required to keep them in check. If they don't, they may discover that their most ambitious hire is also their most unpredictable.    

For more insight and guidance, get in touch with Julia Rockcliffe.

Sign up to our AI and Digital Skills Programme discovery session

Learn more about how our Future skills services can help you
Visit our Future skills page
Learn more about how our Future skills services can help you