AI has become a key part of board-level discussions about growth, competitiveness and resilience. But there’s a serious challenge at the heart of many UK businesses’ approach to the technology.
Although they are pouring a small fortune into AI – an average of nearly £16m this year alone, according to new research from SAP and Oxford Economics – only 7% have a strategic, enterprise-wide plan for AI investment. In other words, much of today’s spending is fragmented, ad-hoc and fundamentally short-termist.
In practice, this means that while AI pilots and implementations are showing some benefits, organisations are struggling to scale those gains across the business, so the full impact on productivity and growth has yet to be realised. This is reflected in the fact that 70% of UK businesses are currently unsure whether AI is delivering on its full potential. “At the moment, most AI projects are tech-led and focused on one business process or department, so they’re not necessarily aligned to the company’s strategic ambitions,” says Sonia Nash, head of business AI at SAP UK & Ireland.
To address the challenge, businesses need a strategy that brings together data, people and governance. This enables experiments and scattered tools to be brought out of the shadows, connected, and scaled to deliver truly transformative results – from improved customer experiences and faster innovation, to the creation of unique products and services.
The first step is to establish and embed cross-functional AI governance, which requires secure C-level sponsorship. Once this is in place, “it’s about defining what your core business objectives are, and if – and where – AI could support them,” Nash explains.
In other words, rather than allowing teams and departments to race ahead with siloed proof of concept (PoC) experiments, organisations need to figure out the ‘why’ of AI – i.e. what business problems it could help to solve as part of a holistic strategy – before supporting investment in new tools or technologies. “AI is a large umbrella term and you hear a lot about the next shiny new thing, which is now agentic AI,” says Nash. “But AI agents are not necessarily the right answer to every problem. Sometimes machine learning is the right answer.”
Effective KPIs focused on employee adoption, cost savings, revenue uplift and other important metrics related to key business goals also need to be established to scale success stories and avoid wasted investment. “You need to monitor not just the technical performance of AI, but also whether it’s having a real business impact,” says Nash. Furthermore, “when PoCs lack predefined KPIs, many teams struggle to articulate a tangible business case for them to executives, which makes it difficult for organisations to commit to enterprise-wide investment.”
Empowering people
Employees are often eager to use AI, but lack the training, confidence and support to do so safely and responsibly. This has caused problems with unsanctioned ‘shadow AI’ within many organisations. Indeed, SAP and Oxford Economics’ research found that 68% of organisations say their staff use unapproved AI tools at least occasionally, and 44% have already experienced data leakage due to shadow AI.
“If someone is trying to experiment with a tool for creating AI agents and plugging that into your data without having undertaken a thorough compliance check, they could be putting the company at great risk,” says Nash. “But many employees don’t realise that because they haven’t had the right training.”
More than half (60%) of organisations admit that their staff have not completed comprehensive AI training, which Nash says is “essential” for safe and productive experimentation with AI. “Mandatory training is going to address the two sides of the coin,” she explains. “First of all, you’re going to make sure that people understand the risks that come with using unapproved tools. And at the same time, you can showcase the approved tools that are at their disposal.”
Along with strong guardrails and guidance about when, where and how AI should be used, the cultural change needed to bring AI out of the shadows also rests on transparent communication about its benefits, both for individuals and meeting the organisation’s long-term goals.
“You can’t just tell people out of the blue, ‘you’re using this new tool now,’” says Nash. “You need to bring them along on the journey, and explain why the company has decided to bring AI into the organisation, how it will reshape some business processes and the benefits this will have.”
Employees also need to feel comfortable asking questions, reporting issues and even making mistakes. Concerns about job security should also be frankly and fully addressed. “There’s still a lot of fear surrounding AI, with many employees worrying it will make some of their tasks obsolete,” says Nash. Leaders, therefore, need to convey that the company’s AI strategy is driven by a desire to “augment and help humans, not replace them.”
The message that AI is here to help needs to come from both ends of the spectrum
Grassroots AI communities can help to foster enthusiasm for new tools by enabling employees to experiment together, share knowledge and provide feedback to the wider organisation. It’s a model that recognises people learn best from peers, not just from top-down mandates.
“The message that AI is here to help needs to come from both ends of the spectrum to ensure you have safe, compliant use of AI, and that it is adopted and used in a way that will return value,” says Nash. “Because even though you can build a case internally for a great tool, if it’s not utilised by your workforce, your project has failed.”
Connected data
Without accurate and connected data, it’s almost impossible to scale AI and unlock transformative results. Businesses across all sectors are currently attempting to build the strong data foundation needed for their AI initiatives. But breaking down silos so that information flows freely across the organisation is still a challenge for many.
AI pilots often succeed in controlled environments because they’re using a curated data set. But if there are issues at the organisation level, it’s hard to scale them successfully. “Poor data quality and accessibility is one of the big barriers to bringing PoCs into a production environment,” says Nash.
However, fixing these issues may not be as challenging as some organisations think. “There are tools on the market today that will help with analysing the state of your data, pinpointing what is missing and helping to fix issues so that you can build a brand new foundation,” Nash explains. “So it’s not as difficult as it used to be.”
Moving disconnected legacy systems, processes and data into the cloud can also help to solve data challenges and enable end-to-end use of AI. “AI is now putting the move to the cloud at the forefront of people’s minds, because it is in the cloud that you can benefit from faster innovation and also break down those data silos.”
One misconception Nash often encounters is that companies should extract all their data from different sources, pour it into a data lake and build AI solutions on top. “You can do some interesting things there, but the problem with that approach is that the data isn’t always up to date and is rarely real-time, which you need for AI to be really efficient and useful.”
Having a unified data fabric is key to get the best out of agentic AI
Data lakes also lack the context AI needs to deliver truly transformative results. Agents – and indeed other types of AI tools – need to fully understand data relationships, such as how a purchase order relates to delivery notes, invoices, and customer queries. “These limitations can be overcome,” says Nash, “but it’s often quite difficult.”
SAP’s Business Data Cloud offers a different approach. It brings together data from all of an organisation’s applications – SAP and non-SAP – into a single curated layer. This creates a foundation for AI that is reliable, responsible and relevant. “Having a unified data fabric is key to get the best out of agentic AI, as agents often operate across end-to-end processes and to do so they need access to data from multiple sources,” says Nash.
Partnering with organisations with deep AI expertise, also plays a vital role in helping companies to navigate such a fast-moving space. “Trying to do everything on your own, potentially with a lack of in-house AI talent or infrastructure, is really difficult,” says Nash. “There are vendors, consultancies and academic institutions that can really help, not just by providing specialised skills or advanced platforms, but also by helping you to move forward with proven methodologies.”
For example, she believes it’s important not to look at AI as a “one-off tool or investment” but rather as “a continuous improvement loop that we call the flywheel effect”, whereby every component feeds into the next. In other words, SAP apps power mission-critical business processes that generate huge amounts of rich, contextual business data. This data feeds SAP’s Business AI, enabling real-time intelligence, which is then embedded back into SAP applications to create smarter workflows, predictive insights and transformative user experiences.
UK businesses expect the return on their AI investments to almost double over the next two years, from 17% in 2025 to 32% in 2027. But achieving this will require many of them to shift from fragmented experimentation to strategic, enterprise-wide deployments of the technology.
In short, the path forward for UK businesses isn’t simply about spending more on AI tools. It’s about spending smarter to draw scattered experiments out of the shadows, ensure staff have the training they need and connect fragmented data, so that AI delivers truly transformative results.
For more information, visit sap.com/uk
AI has become a key part of board-level discussions about growth, competitiveness and resilience. But there’s a serious challenge at the heart of many UK businesses' approach to the technology.
Although they are pouring a small fortune into AI – an average of nearly £16m this year alone, according to new research from SAP and Oxford Economics – only 7% have a strategic, enterprise-wide plan for AI investment. In other words, much of today’s spending is fragmented, ad-hoc and fundamentally short-termist.
In practice, this means that while AI pilots and implementations are showing some benefits, organisations are struggling to scale those gains across the business, so the full impact on productivity and growth has yet to be realised. This is reflected in the fact that 70% of UK businesses are currently unsure whether AI is delivering on its full potential. “At the moment, most AI projects are tech-led and focused on one business process or department, so they’re not necessarily aligned to the company’s strategic ambitions,” says Sonia Nash, head of business AI at SAP UK & Ireland.
