Ireland is a country that prides itself on being outward-looking — a small, open economy that built its modern success on successive waves of foreign investment and a willingness to adopt technologies early.  

From pharmaceuticals to data centres to cloud computing, global innovations have often arrived here quickly and embedded deeply. 

And yet, when it comes to AI agents — what many believe to be the next major leap in artificial intelligence — Irish organisations find themselves in an unfamiliar position: highly optimistic, but also hesitant. 

PwC Ireland’s new survey on AI agents reveals a striking paradox.  

On one hand, business leaders are convinced that AI agents could potentially reshape the workplace more profoundly than the internet did.  

On the other hand, most organisations are still experimenting at the edges with small trials while having yet to tackle the bigger challenge of rethinking how AI agents can help transform how their business operates.  

At its core is a blend of big ambition and cautious adoption. The technology is racing ahead and companies are experimenting with it while also dealing with a range of related challenges, such as legacy systems integration, messy data, thin resources and the natural nerves that come with shifting how work actually happens. 

To dig into the findings of the survey, I spoke with two of PwC’s senior experts for the latest episode of The Tech Agenda podcast series.  

Robert Byrne, Technology Data & AI Partner at PwC Ireland, has spent years helping organisations test, deploy and make sense of automation and AI systems. He was joined by Laoise Mullane, Director and AI Adoption Lead in PwC Ireland’s Workforce Advisory practice, who works with leaders and employees on the cultural, organisational and skill shifts that new technologies demand. 

Together, their vantage point spans both sides of the transformation: the technical potential and the organisational reality. 

What makes an AI agent different?

Before assessing adoption, it helps to understand what exactly is changing. For several years, artificial intelligence has been defined by GenAI tools — systems that support answering questions, generating documents or automating predictable tasks. AI agents mark a shift from assistance to action. 

“They are systems that can act independently and perform actions on your behalf,” says Byrne. 

To illustrate, he gives a practical example of a smart heating system that becomes far more intelligent once AI agents are applied. Instead of simply following manual settings, the system can receive different inputs — your preferences, usage patterns, even the weather — and then decide when to turn the heating on or off based on what it knows about your routines. 

That basic structure — reading information, interpreting it, deciding and acting — is exactly what makes AI agents so significant inside organisations. 

If a system can autonomously triage customer emails, classify documents, validate claims, process invoices, run compliance checks or respond to HR queries, the implications ripple across every function. Agents don’t just handle inputs; they have the ability to make choices and take action based on those choices. 

This is why PwC expects agentic AI to move quickly from the edges of the enterprise into the centre. 

The survey shows that Irish organisations understand the potential of AI agents — but they are mostly starting in familiar places: with use cases focusing on efficiency, automation and cost reduction. 

“What we are seeing at the moment… usually in the back office… are cases where organizations want to either improve productivity or reduce cost,” Byrne says. “Help me do something quicker or help me reduce my cost.” 

It’s hardly surprising. Generative AI slipped into organisations as a way to boost personal productivity — summarising, drafting and reshaping text. AI agents push that further, helping people work quicker and streamlining the gaps in between. 

But Byrne argues that if organisations stick to small improvements, they’ll miss the real opportunity. 

“The real business case… isn’t really on a one-by-one basis,” he argues. “How do you use agentic AI to change your operating model, to change whole processes or workflows… and less about individual use cases?” 

And that’s the sticking point. The early wins are nice, but they’re not transformative. The survey shows the pattern clearly: Loads of firms say that agents help with productivity, but far fewer can point to real savings — and only a tiny number have revamped their processes to match the technology. 

How the workforce is feeling the early effects

If Byrne looks at systems, Mullane looks at people — and she sees early, tangible impact. 

“It is having an impact on the workforce. That’s the first thing to say,” she says. “People are experiencing the benefits… productivity, creativity and increased quality.” 

But where public debate often drifts toward job displacement, Mullane’s experience suggests something different. 

“It is less about doing the same with less people,” she explains. “It is more about doing more, with more people.” 

PwC’s global labour market modelling supports this view: Jobs are still growing in sectors exposed to AI, just at different rates and with different skill profiles. The challenge is not a collapse in employment, but a shift in what jobs require. 

And here the survey reveals a deeper, more nuanced barrier. Of all the insights from the discussion, one stands out: Mindsets — not technology — are slowing adoption. 

“We have gone from talking about general AI tools to talking about AI agents very quickly,” Mullane says.  

That acceleration comes with emotional consequences. Employees are uncertain about accuracy. They worry about how their roles will change. They struggle to trust systems that make autonomous decisions. This is something Mullane understands better than most. 

“There’s a lack of trust,” she notes. “But people, I think, want to trust the technology.” 

Trust doesn’t come ready-made. It needs involvement and transparency. 

“People are more likely to adopt technologies… when they have been involved and engaged throughout the transformation process,” she argues. 

Yet the survey shows that organisations rank workforce-related challenges — such as change management and employee adoption — significantly lower than technical issues. 

“I was surprised by that statistic and a bit concerned,” says Mullane. 

In Mullane’s view, organisations may underestimate how much cultural and behavioural work will be required to integrate autonomous systems into daily operations. “People challenges are inherently more difficult to solve than system challenges,” she says, and they cannot be an afterthought. 

Ireland’s ambition is strong — but action is slower

For all the hesitation, the survey reveals a deep well of optimism.  

Irish executives can clearly see the potential of what is coming. Nearly two-thirds of those surveyed believe AI agents will reshape the workplace more dramatically than the internet itself — a striking signal of how seriously leaders now take the technology.  

Close to half say they expect roles within their organisations to transform significantly within the next twelve months. Investment intentions tell a similar story: seven out of ten organisations plan to increase their spending on agentic AI over the coming year. And the vast majority — more than eight in ten — are already experimenting, testing or beginning to adopt the technology in some form. 

But this enthusiasm has yet to translate into widespread deployment. Only a small minority of Irish organisations describe their adoption of agents as broad or embedded. The same is true of value realisation.  

While many report early productivity gains, relatively few are seeing improvements in profitability, and an even smaller cohort say they are generating new revenue streams from agentic AI. In other words, the benefits are visible — but they remain largely incremental. 

“There are great levels of appreciation for the opportunity,” Byrne says. “There are great levels of excitement… But probably what is missing is… the level of pace.” 

He rejects the idea that Ireland lacks vision. The difference between Irish and US organisations, he argues, has far more to do with maturity than ambition. 

“I think we lag in adoption. I don’t think we lag in ambition,” he says. “We understand the importance of establishing and preserving the trust that our colleagues and our customers have placed in us to get this right.” 

For Mullane, the momentum matters more than the gap. The fact that more than 80 per cent of organisations are experimenting or in early stages of adoption signals, to her, that Ireland is moving — even if not yet at full speed. 

“83 per cent… are experimenting or in the initial stages of AI agents,” she says. “They are the numbers I am focusing on.” 

While mindset is a significant factor, the survey also reveals structural barriers. Irish companies cite data issues as their single biggest obstacle to realising value from AI agents — far more so than their US counterparts. Legacy systems create bottlenecks. Poor data quality hinders training. Integrations slow down deployment. 

In many organisations, the digital foundations required for agentic AI are still incomplete. This matters because AI agents do not operate in isolation; they depend on clean, connected, timely data to act effectively. Without that, organisations risk hitting a ceiling long before they scale. 

A matter of transformation

Byrne believes that the next phase will require a shift in perspective. Instead of layering agents onto existing processes, organisations need to rethink the processes themselves. 

“If you are talking about transforming your business… move the focus away from… use case, by use case,” he argues. “Think instead… of whole areas of your business… a whole process, a whole business function.” 

He talks about a government client that’s rebuilding a full citizen-facing process from the ground up where agents will form a central part of the new process e.g. reading the applications, sorting them, chasing the missing details and keeping the conversation going. The aim isn’t to tweak the process— it’s to transform it. 

That kind of jump, from experimenting to genuinely rethinking how work is done, is where the top US organisations already are. 

“When we look at high performers… the boxes that they tick are that they are ambitious… focus on both the front office and the back office… and take a transformative lens,” Byrne says. 

Governance and responsible AI 

Mullane emphasises that transformation cannot happen without ethical and responsible guardrails. 

“When you’re setting out to do this, you really need to understand what outcome you’re looking to achieve,” she says. “We need to be measuring all the time to make sure that these outcomes are realised.” 

For her, responsible AI is not a compliance exercise; it is essential to adoption. 

“What is the governance? What are the compliance mechanisms? Who are the people that are going to be involved?” she asks. 

Clear governance makes it easier for employees — and customers — to trust the technology. 

For leaders weighing the scale and uncertainty of this shift, both experts offer clear advice. 

Mullane emphasises participation: “Get involved because the more that you understand… the more you are able to join the conversation and influence how it’s going to be used.” 

Byrne points to courage: “Be braver. Be braver in terms of what you want to get out of this technology and in your pursuit of it… You will absolutely succeed.”  

The Tech Agenda with Ian Kehoe podcast series is sponsored by PwC.