Medium-sized and large UK businesses surveyed by SAP are increasing investment in artificial intelligence despite widespread difficulties with data quality, organisational oversight and the measurement of commercial returns.
The second Value of AI in the UK report, produced with Oxford Economics, estimates average AI expenditure among the UK organisations surveyed at £20.6 million in 2026, an increase of 27% from the previous year.
Respondents expect their investment to increase by a further 45% over the next two years. They also anticipate that average returns on AI investment will rise from 22% in 2026 to 39% by 2028, although these are survey-based estimates and forecasts rather than independently audited financial outcomes.
The research focuses on organisations employing at least 500 people. Its reported average expenditure is therefore not representative of the typical business across the wider UK economy, which includes large numbers of smaller employers.
Despite greater spending, only 11% of surveyed UK businesses describe their approach to AI investment as strategic and organisation-wide, compared with 7% in the previous year’s research.
A further 41% describe their approach as piecemeal, while 40% prioritise investment at departmental level. These arrangements can allow individual teams to pursue useful applications but create difficulties when an organisation attempts to consolidate systems, measure expenditure or establish consistent oversight.
Data quality is a particularly widespread problem. Seventy per cent of respondents say the availability or reliability of information is preventing them from obtaining greater value from AI. Two-thirds, 67%, identify integrated data systems as the most important contributor to readiness.
These findings reflect the dependence of AI applications on the underlying information and processes available to them. A new model may be capable of analysing documents or automating activities, but inconsistent records and disconnected applications can limit its effectiveness.
The governance challenge is becoming more demanding as businesses experiment with AI agents capable of undertaking connected tasks through enterprise software.
Seven in ten surveyed UK organisations are piloting agent-based applications. However, 49% say the deployment of AI agents is advancing faster than the organisation’s ability to standardise and govern them.
Only 9% consider their AI governance skills fully ready, while 61% are still developing formal processes and oversight. The findings suggest that operational controls are struggling to keep pace with the speed at which new capabilities are introduced.
Agentic systems can require more extensive safeguards than conventional AI tools because they may interact with business records, carry out instructions and initiate actions across connected applications. Decisions about permissions, testing, monitoring and accountability consequently become part of the deployment process.
Workforce development is another substantial area of investment. Eighty-three per cent of surveyed businesses are upskilling or reskilling employees to work alongside AI, while 79% believe maximising its value requires organisational change extending beyond technical instruction.
Facilities management business Mitie provides an example of a specific operational deployment. The company reported that its AI assistant handled more than 29,000 HR interactions during three months and helped reduce calls to its People Support function by 36%.
The result illustrates how automation can change the distribution of routine work, although it does not establish that comparable systems will deliver the same outcome in other organisations.
Respondents expect AI to support 50% of tasks by 2028, compared with an estimated 31% currently. Those figures describe work assisted by AI rather than a forecast that half of jobs will disappear.
The financial expectations are similarly conditional. Additional spending on software, computing, integration, data preparation and employee training can increase costs before productivity improvements translate into measurable profitability.
The research was conducted online by Oxford Economics between March and May 2026, involving 2,600 director-level and more senior respondents across 13 countries, including the UK. The sample was evenly divided between organisations employing 500–1,500 people and those employing at least 1,500.
Respondents supplied financial estimates in US dollars, with the published sterling figures converted using an exchange rate of 0.75 on 23 September 2026.
The investment forecasts indicate considerable confidence in AI’s potential. Whether those expectations are realised will depend on organisations establishing stronger data foundations, consistent governance and credible methods for measuring the resulting business value.
You must be logged in to post a comment.