Sadly for journalists writing end-of-year trend articles, it’s rare that the evolution of technology happens in steps that can be neatly summarised. Artificial intelligence, however, might be an exception.
In 2023, the world reacted with awe to the release of ChatGPT, stunned by software that could respond to prompts with sometimes uncanny results. Throughout 2024, organisations experimented with generative AI, putting the technology to the test with pilot projects and proofs-of-concept.
Next year, businesses will work to derive real value from artificial intelligence – and show that the investment was worthwhile.
From proofs-of-concept to AI at scale
Looking back at 2024, few organisations managed to implement AI at scale, says Simon Baxter, principal analyst at technology research firm, TechMarketView. Instead, they were busily working out and refining their AI strategies – even the businesses that were ahead in AI.
“Those organisations that have already done a lot of these proofs-of-concept are the ones who are going to be picking five to 10 of them to really scale and understand what their cost profiles look like,” says Baxter. “These pilots might be really cost-effective at small scale, but if you want to scale it to 80% to 90% of your customer interactions, that cost can go up.”
The good news is that, among the organisations who have scaled up their implementations, it has been “quite clear that the gains you get from the current range of generative AI solutions is very real,” he says. “It’s not hype.”
Businesses, such as cloud software company Nutanix, for example, are targeting a 25% improvement to developer productivity by using genAI for code generation, among other functions, says CEO Rajiv Ramaswami. “AI applications in the enterprise are going to become mainstream,” he says.
Agentic AI
A phrase currently on the lips of every genAI provider is ‘agentic AI’. Providers hope that this evolution of the technology might deliver on the more ambitious promises of AI, as these intelligent agents are pointed at tasks in order to free up employees who can then focus on the more important stuff. It’s early days, but companies such as Microsoft are working on creating ‘orchestrator’ bots that manage the other autonomous agents and ensure they’re all working efficiently.
Next year will be a “defining moment for agentic AI,” says Harshul Asnani, president of the Europe business at IT services company, Tech Mahindra. “It will be driven by advances in accessibility, affordability and integration,” Asnani adds. “AI’s ability to automate decision-making processes will enable faster and more precise responses to dynamic market demands.”
He predicts the capabilities delivered by agentic AI, such as personalised customer engagement in retail and predictive analytics in finance, will, over 2025, see the technology become an “everyday operational necessity”.
The scalability and adaptability of agentic AI will open the door for companies in all industries to more quickly and easily perform tasks including trend analysis, resource allocation and real-time problem-solving, he says.
From user-driven to contextual AI
The move to agentic AI is also going to change how enterprises interact with the technology, says Bob de Caux, head of AI at heavy industries software enterprise, IFS.
Until recently, most AI has depended on input in a question-and-answer format from the user. But one promise of agentic AI is to quickly process vast reams of data and put forward helpful, contextual suggestions automatically.
“We’re seeing a real pivot towards business leaders wanting to make more real-time decisions – or close to real-time decisions. It’s not just about replacing what humans were doing before by automating processes, it’s also about being able to look over huge volumes of data that are coming in quickly and serve suggestions to the user.”
IFS is seeing a lot of interest from customers in using AI for forecasting or simulations, says de Caux. Today, forecasting typically involves bringing many disparate data sources together, using a number of different tools. GenAI can simplify and streamline this process. “They want to be able to take a forward-looking view much faster, without the huge effort, so that they’re only doing their planning every six months to a year,” he says. “That mixture of predictive and generative AI can really help.”
The emergence of AI services
With most AI tools so far, businesses buy the software and deploy it. They are expected to manage all the risk or potential fallout themselves. Organisations must operationalise the tooling and deliver a return on investment. But there are early indicators, says TechMarketView’s Baxter, of a new AI services mindset starting to emerge.
In the legal sector, for example, a large firm might use its industry knowledge and technical chops to build its own bespoke AI platform. Customers could then outsource some of their legal needs – say, contract review, document analysis or proposal writing – to this company, which can use its AI platform to run these services more cheaply and effectively, as well as manage all of the risk.
“This would move the dial from ‘let’s just sell the platform and let the customer deal with it’ to putting a bigger wrapper of services around it, where the provider will keep innovating, deal with bug fixes and guarantee the app is delivering the desired outcome because you’re paying for a service not just a product,” says Baxter.
Only training can solve the shadow-AI problem
Employees want to use AI. But, sometimes, organisations do not allow the use of generative AI – for good reasons. They don’t want, for example, confidential company documents or other information being uploaded into public AI platforms. Businesses might also be on shaky compliance grounds if they allow unsupervised AI use at their organisations. The data exposure in this so-called shadow AI is becoming a real challenge.
In 2025, organisations will have to lead from the front if they want to encourage the responsible and ethical use of AI in their businesses. That means lots and lots more training, says Yao Morin, CTO at global real estate business, JLL.
“If users aren’t using applications properly or do not understand the limitations of AI, there is going to be risk,” Morin says. Enterprises must therefore put great emphasis on training and education. They can’t afford to introduce AI tools and hope for the best.
“A lot of enterprises are now starting to talk about how they train their employees to be AI-ready and to understand AI,” says Morin. “That’s a crucial aspect to make sure AI applications are more secure in 2025.”
Manish Jethwa, the CTO of the UK’s national mapping agency, Ordnance Survey, agrees. Next year will likely see a broader commitment to retraining and upskilling employees to prepare them for the impact of AI, as those proofs-of-concept start to scale.
“But we need to be careful that, in our aim to enhance efficiency, we don’t lose the personality, creativity and emotion we bring as humans in the workplace,” he adds. “This is crucial to building strong relationships with customers and partners but also keep for the kind of collaboration and teamwork that defines a business’s culture.”
