We have some bad news: Even early adopters of AI may already be behind the curve.

Although AI has been the big-ticket strategy over recent years, with firms capitalising on the promise of benefits in productivity and efficiency, those who thought they were future-proofing might find the next wave of technology overtakes all of their efforts to date.

That next wave takes the form of agentic AI: tools that build on the natural language processing power of genAI to create flexible frameworks attuned for learning and adaptability, and considered by some as the final step before the world sees artificial
general intelligence.

This isn’t just another upgrade – it’s a paradigm shift, featuring AI able to think, decide, and act autonomously, with AI agents forming part of wider agentic AI systems, acting as autonomous, decision-making layers on top of language models that can observe and collect information to generate tailored actions, but without the need for constant user input, or even oversight.

As of last year, less than 1% of enterprise software applications included some form of agentic AI, but tech consultancy Gartner has predicted that by 2028 that number will rise to 33%, allowing for 15% of day-to-day work decisions to be made autonomously.

One sign that this isn’t just hype? Industry heavyweights are already moving to deploy agentic AI at scale.

Some of the world’s biggest businesses have made their move, including Accenture, with the launch of an AI Refinery, McKinsey & Co with the QuantumBlack AI consultancy, not to mention Deloitte’s Agentforce accelerators, all of which are aiming to transform the business technology landscape.

Yet another offering comes from global consultancy PwC, with a recently announced collaboration alongside Microsoft that will see the launch of an Agent Factory, aiming to bring AI-powered agents directly into workflows.

To find out how the AI rulebook is already being rewritten, DIGIT spoke with Stewart Wilson, PwC’s Microsoft Leader and technology consulting partner for Scotland, who highlights that partnerships like that between PwC and Microsoft are key not just for unleashing the power of AI, but also channelling it into meaningful impact.

“Microsoft’s grip on the technology and everything they bring is unquestionable,” said Wilson.

“When you combine that with the industry knowledge and experience of PwC, it’s very, very powerful.”

It’s true that Microsoft is something of a trailblazer in agentic AI, having launched its AutoGen library back in 2023 as a flexible, open-source framework for developing AI agents, but while Big Tech can build the technology, firms like Microsoft often benefit from business partnerships to turn disruptive technology into real enterprise transformation.

“Transforming a business isn’t just about technology,” agreed Wilson.

“It’s about changing the way people work, it’s about changing business processes, it’s about changing culture. It’s about all of those things underpinned by technology.”

To that end, organisations looking to take the next step on their AI journey should think twice before simply rolling out autonomous agents in the hope of achieving elusive returns on productivity, or saving money by replacing staff with virtual coworkers.

Strategy, Skills, and Sustainability

Like any software-based tool, the benefits of agentic AI depend entirely on how it’s being used.

Data from Langchain shows that the top use cases for AI agents right now include research and summarisation (58%), streamlining productivity tasks (53%), customer service duties (46%), and code generation (35%), suggesting that AI agents are being used in basically the same way as earlier AI technologies, like customer-service chatbots or productivity tools.

So how can firms take this to the next level?

“Organisations need and benefit from external help,” said Wilson. “But they have to do that with a plan for sustainability in place. Fast forward a year, and probably the agents will wither on the vine if you haven’t thought about what you put around them.”

That means ensuring operating models are sound, strategies are clear across the business, and there are ongoing routes to value. To help firms achieve this, for its part PwC is ensuring that its Agent Factory will deliver reusable assets to tackle common client challenges, helping speed up ROI.

However, perhaps the most important aspect of a successful transformation is ensuring staff are upskilled to take on roles being reshaped by emerging tech, where research suggests many organisations are falling short.

“It’s got to be human-led to be effective. You’ve got to plan around skills and knowledge transfer,” said Wilson.

“You’ve got to keep relevant by doing the certifications, and make sure your workforce is doing the certifications so that they’re up to speed and have the valid skills and experience.”

Putting people at the heart of an AI strategy might seem counterintuitive, but consider that last year, more than two-thirds (68%) of business leaders reported struggling to attract the right talent to manage their AI solutions, or that 66% of leaders now say they won’t approve new hires without AI skills.

Another study from Ipsos found that 53% of UK office workers want to learn new skills, but only about a third (34%) said they have had the opportunity to learn about using AI at work.

As AI becomes more embedded across industries, the key to success will be ensuring human talent keeps pace, a factor that will separate those with effective strategies from those left behind, but it’s in down in the data where the real value lies.

Bridging the Data Gap

UK Government statistics show that 99% of British businesses now handle digitised data, with more than a fifth (21%) analysing this data to generate new insights and knowledge.

Meanwhile, other research has found that almost three-quarters of firms still struggle to use their data to deliver significant business value.

Agentic AI, coupled with specifically tailored agents, could make this a lot easier.

“Unlocking better decision-making is a massive benefit of agentic AI,” said Wilson.

“One of the biggest issues that I see is a lot of organisations have evolved over timewith a different technical landscape and assets, so all this data is held in so many different spaces.

“The role agentic AI can play, sitting at the heart of data contained in disparate sources, is huge. There’s a lot more we can do around strategic decision-making, actually unlocking faster, quicker, more insightful business decisions.”

However, more important than moving to quickly roll out a half-baked agentic AI strategy just to find some diamond in the data mud is doing it with care, and that has to begin, argued Wilson, with those in leadership.

AI From the Top Down

“Change enabled by AI has to be led from the top.”

“At PwC, all partners are encouraged and expected to use CoPilot to streamline and improve how we work to support our clients and staff. If we’re not using AI, I can’t talk about the benefits it’s delivering to me, and so how can I be telling staff to be embracing AI?”

Data show that more executives are taking this careful approach now that the initial furore around the technology has died down, with 61% of business leaders reporting that their interest in responsible AI has increased over the past year. Those leading the charge are trying to make sure firms stick to this commitment.

“We put a lot of currency on trust,” said Wilson.

“Every organisation, every enterprise has a responsibility to deploy and manage AI responsibly.”

This ethical approach is crucial, as agentic AI is set to become the most powerful tool in business’s arsenal – bringing equally profound implications.

A research paper from Google DeepMind suggests these advanced AI assistants could “radically alter the nature of work, education and creative pursuits as well as how we communicate, coordinate and negotiate with one another, ultimately influencing who we want to be and to become.”

All that to say, the stakes couldn’t really be much higher.

AI for Impact, Not Just Innovation

The idea of having a tool that needs little maintenance or oversight, and can take on complex tasks from diverse data sources to deliver the same results as a human mind, but without the need for breaks or pay, is obviously enticing.

A Capgemini survey from late last year found that 82% of businesses plan to integrate AI agents within the next three years, an ambitious target given the scale of the challenge involved, and one which firms will have to move quickly to meet.

However, pushing too far, too fast could prove damaging for long-term prospects.

“The opportunity does take time to be unlocked. It’s got to be done with care and diligence, and it’s got to be architecturally sound,” said Wilson.

“Make sure it’s infused with your architecture, your business model, and your operating model. Then it’ll be transformative, then it will yield benefits and returns.”

Recommended reading

• AI Divide Could Risk UK Agentic AI Ambitions
• Agentic AI on the Rise Despite Workers’ Concerns
• IT Executives Are Primed for Agentic AI

Think back to the early 1990s, when businesses first grappled with the rise of the internet.

Many firms bolted on isolated digital strategies, disconnected from core operations. But the success stories came from those who reimagined their entire business models, embedding digital at the foundation rather than treating it as an add-on.

Getting the most out of agentic AI will require a similar outlook, something that larger companies have already realised, as the partnership between firms like PwC and Microsoft shows.

But this should be a gradual process, taken with care, and not pushed out just so organisations can claim they are keeping up, something which Wilson highlighted as an issue in the current landscape.

“It shouldn’t be AI for AI’s sake. It’s about infusing and embedding agentic AI in the transformation journey so you’re not just looking at AI in isolation,” said Wilson.

“With all the work we’re deploying, agentic AI has to lead to an outcome, and not just that the technology is delivered, that it delivers something.”

That’s why any new AI strategy should be underpinned by measurable outcomes.

Fewer than half of chief information officers and technology executives (48%) think that their current digital initiatives are either meeting or exceeding business outcome targets, according to Gartner, suggesting that historical ways of measuring outcomes, including ROI, may no longer be fit for purpose.

Yes, it’s possible to calculate how much time AI is saving workers on repetitive tasks, with noticeable positive effects in specific areas like customer management, fraud detection, compliance, and finance, but the biggest impact of agentic models are yet to be witnessed.

According to Wilson, those revolve around speed to insight – the ability of agentic AI to act as a ‘coordinating brain’ to quickly analyse data, extract signals from all the noise, and make decisions based on that analysis.

“Speed to insight is one of the biggest revolutionary changes. I think that’s what can really help accelerate organisations’ ability to gain insight and make decisions quickly, therefore reducing the time to value,” said Wilson.

“Improved forecasting, or invoice clearing, or PO matching, or customer service, that’s the easy stuff. I absolutely think the next big leap will be in that area of speed to insight. But have I seen it yet? No.”

At the pace technology is currently moving, it probably won’t be too long coming, but to make the most of it when it arrives, businesses need to begin reexamining their AI strategies now, ensuring these are founded on the principles of safety, security and responsibility, with a blueprint to deliver measurable outcomes.

Fail to do that, and there might be a lot of information, but very little insight.