Joe Atkinson, Global Chief AI Officer with PwC, has held an array of high-ranking roles over his 33 years at the firm, but none have been quite like the one he occupies today.

“Well, it feels like it changes every 15 seconds, so by the time we get done this conversation, the role will probably have changed,” he says. “But I’m responsible for helping to drive our AI strategy, how we transform the firm and really adopt AI to become much more of a native AI professional services firm.”

At the same time, he also works with PwC clients to “help them do the same thing in their businesses”.

“So, it’s an interesting role of transforming our firm while at the same time serving our clients as they transform themselves,” he says.

Atkinson is sitting in The Currency’s podcast studio having just met with some of PwC Ireland’s clients, hours after he landed in from Philadelphia. His team have been monitoring the way in which companies adapt to AI and, like with most new technologies, there are a select few that are vastly outperforming their peers.

Three quarters of the economic gains being derived by AI have been captured by just a fifth of companies, according to the firm’s most recent AI performance survey. One of the commonalities among those doing well from the technology is that they view it as a driver for growth, rather than purely as a cost-cutting tool.

“I think the interesting thing about that is those organisations that are seeing those outsized gains, not only are they accruing much more of the benefit in that small group of companies, but they are outpacing their peers significantly, as the math would suggest,” he says.

“So you have this issue where not only are they driving greater returns, but they’re driving something like seven times the returns of their AI investments relative to their peers.”

Atkinson explains that this is calculated through an index that combines both the “top line implications of AI growth” as well as the “improvements in efficiency.” He says that there are many businesses seeing one or the other but the standout performers are seeing both.

“We look at it as both a growth opportunity: how do you open up new markets, how do you start to access markets that perhaps you couldn’t have gotten into before. How do you build products and capability that you couldn’t get into before?

“At the same time, every organisation has an obligation to figure out how to do what they do today faster and more effectively, and hopefully in doing that free up the investment capacity that they need to chase the next opportunities.”

The superstar category

To date, businesses have been leaning on AI as a tool to drive further efficiencies within their businesses. Be it the automation of tasks through chatbots or agents, spending on AI tools has skyrocketed since ChatGPT was first introduced in 2022.

But Atkinson says the businesses that are also focused on growth through AI are in the “superstar category”. He says these high-performing companies are paying attention to the fundamentals of their business. They’re looking at areas like governance and investments in their data, he says. They are also using it to help define their investment criteria when deciding where to deploy capital. These elements are the “bread and butter” of how organisations operate today, and he says it is “really important” to get them right in order to see gains from AI.

“That becomes maybe not quite half the equation, but it becomes a gate or, if you will, the binary switch. If you have it, you can move into returns. If you don’t have it, you’re really going to struggle to get those returns,” he says.

The PwC AI chief adds that the top performers are going after “the hard stuff” within their organisation by using AI to alleviate areas where there’s “real friction”.

“There are lots of easy spots where AI can improve a process, they can improve a task, but where we’re seeing those companies that are outperforming, it’s because they’re going after the more complicated things,” he says.

He also adds that they are also looking to partner with players in other industries, giving the example of energy companies, which are “spending a lot more time with data infrastructure providers than we probably would have said a few years back.”

Connecting with customers

There are certain areas that Atkinson has seen considerable AI use in. He gives the example of back-office functions that “enable businesses to work”. He says that there is a clear focus on getting those back-office functions to work more efficiently.

Another area he sees significant adoption in is the connection between customers and companies.

“In those consumer businesses where consumers interact with the companies, I think most of us would say we might have an example or two where companies that are really, really great at the consumer experience, but probably most of us could rattle off another set of companies where we say they’re not great at it. And so, it’s a compelling use case,” he says.

“And the interesting thing about that particular space is, it also addresses an area where the labour pool actually isn’t really all that easy to go after. There’s not a lot of demand for jobs in call centres, for example, so you can start to really think about this as solving a problem on a multilevel layer.”

Growing jobs

One of the most significant concerns around the advent of AI has been the impact the technology will have on jobs. AI has the capability to automate an array of tasks within businesses and indeed many companies have already started to put it to use. To some, there is a fear that AI will lead to widespread unemployment, while proponents of the technology relate it to something like the industrial revolution, where machines replaced jobs but created an array of newer higher-value roles.

PwC has been tracking the impact of AI on employment over the past three years. 

“What’s interesting and maybe a little counterintuitive is we’re seeing job growth, and the value of those jobs growing in the industries that are most exposed to AI,” he says.

“You’d sort of think it would go the opposite way, but actually what’s happening is those organisations are unlocking growth, and as a result, they’re unlocking more opportunity for people to join them, so they’re actually seeing labour growth rise.”

The PwC AI Jobs Barometer tracks over a billion job postings from around the world and Atkinson says that it is clear the nature of the new roles is “changing rapidly”, particularly when it comes to the skills required for the jobs.

“A job posting today for the same theoretical title would probably have different skills listed today than it did a year ago,” he says.

“What’s changed is more of those skills would change in a year than what we’ve seen before. So, if there are five skills listed, three of them might be different now. Next year maybe all five of them will be different.”

A cloudy crystal ball

Predicting what will come next with AI is a very unenviable task. The level of change that has happened within business and the wider world since LLMs were first put into use has been staggering.

Atkinson jokes that his crystal ball is “getting cloudier and cloudier” the faster AI develops.

“But I do think whereas today most of us are talking about chatbots and assisted work where the AI is assisting, we’re moving very rapidly into enabled work being the core capability where the chatbot really gives way to more of an agentic workflow where the agents may be operating in the background, and the individual that’s actually responsible for that function actually isn’t watching and really monitoring those activities at the activity level,” he says.

“They’re monitoring outcomes, and they’re monitoring the guardrails that are put in place around that.”

Indeed, those guardrails will form a crucial part of the next era of governance and sovereignty within businesses. In previous technological breakthroughs like cloud computing, there have been moves like local hosting to protect a company’s data.

With AI, there is already a host of regulated industries that are deploying AI such as finance where it is being used to help comply with anti-money laundering regulations or in the legal space, where Harvey AI has proven to be a massive hit with legal firms.

“Is there enough safety and security? There’s probably work to do. And I think that’s where organisations are looking today to try to figure out what that pathway looks like,” Atkinson says.

“But at the end of the day, the plates are moving underneath everybody’s feet here, and so governance, the risk management side of this, what can the agent do, what do I authorise the agent to do, how do I contain the agent to the way that I want to contain it, obviously that’s been a very hot topic right now. Those all become really important.”

Excitement and anxiety

In his role at PwC, Atkinson has the unique perspective of talking to a vast array of companies and workforces. So, when he goes out there to meet the staff and management of these companies, is he met with people eager to learn and excited about AI’s capabilities, or people who are more anxious and uncertain about what the change will bring?

“Oddly, it’s a mix. It depends very much on what kind of company you’re sitting with, right? 

So, when you sit with the technology companies that are those small providers that are working the niche and even the large-language model, the frontier labs, you see excitement,” he says, adding that a similar enthusiasm is found within professional services.

“Then you see organisations that are more on what I’ll characterise as either the old school or have a more challenging story for those workers. There’s no question that the task profile of most jobs is going to change. The question then becomes: what’s the distribution of the tasks in those jobs and does the job exist now or has the job morphed? Do I need to reskill that individual? Do I have a different headcount in different places, and maybe I have less headcount in certain places?”

Atkinson says one of the ways organisations can build trust while adopting AI is to confront the reality of it. He doesn’t think it’s helpful to tell staff that everything is going to stay the same, nor that everything is going to go away.

“I think the reality is a pathway in between, and organisations have to look at what those circumstances are for them, and then I think they have to communicate with integrity and trust to their people about what they should expect,” he says.

*****

When it comes to new graduates or people entering the workforce, it can be an incredibly challenging time. For Atkinson, he believes that expectations have changed significantly when it comes to entry-level positions. He gives the example of someone joining PwC as an associate, adding that they are now expected to have the skills that were more likely to be found in someone with a couple of years of experience.

“We need them to be able to operate in teams, communicate their ideas, influence and convince people of a thesis, help change minds, help drive action, opportunity, influence others, etc,” he says.

“Those are skills that previously we would develop in people over the first two or three years, and now we need them to enter the door with a lot of those skills.”

Atkinson explains how roles have changed over the years, citing how at one point “deep technologists were your developers” whereas now people are developing applications and tools at scale that were once the preserve of engineers. He adds that there are two important aspects here, the first is that there used to be tech jobs and non-tech jobs. Now he says there are just jobs and “everyone’s got to have a bit of both”.

The second thing, he says, centres on a Mark Twain quote: “News of my death has been greatly exaggerated.”

“The thinking that we won’t need computer engineering, computer sciences, I think we tend to oversimplify that question,” he says.

“You still have complexity in enterprise technology, infrastructure capability. You’re still going to need enterprise-scale applications. And the development of those is going to look very different from the historical development of those, but those disciplines and sciences may be a different demand curve than we all thought, but there’s still a very, very strong demand curve there.”

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