Summary

• UK organisations using AI reported average efficiency gains of 23%, compared with 34% across the six other markets in Infor's survey.
• Full-scale UK deployment increased from 21.8% to 23.6%, while confidence in managing implementation without disruption fell from 75% to 66%.
• The findings identify data governance, legacy integration, human review and unclear accountability as continuing barriers, without independently establishing that industry-specific software produces better results.

British businesses are reporting smaller efficiency improvements from artificial intelligence than their counterparts in six other major markets, even as more organisations move beyond experiments and introduce AI into everyday operations.

Infor’s second Enterprise AI Adoption Impact Index found that UK respondents using AI reported average efficiency gains of 23% over the previous year, compared with 34% across the other six countries surveyed. The gap comes alongside falling confidence in implementing AI without disrupting established processes.

The study, commissioned by Infor and conducted by YouGov in August, surveyed 2,111 business decision-makers across the UK, United States, France, Germany, Japan, Singapore and Australia. The UK sample comprised 254 respondents.

The findings describe organisations’ reported experiences rather than independently measured changes in economic productivity. Nevertheless, they identify difficulties that persist after companies begin spending money on AI, particularly when existing software, business information and governance arrangements are not prepared for its use.

Adoption continues while confidence declines

Full-scale AI deployment among UK respondents increased from 21.8% in the earlier April survey wave to 23.6% in August. When partial adoption is included, more than 60% reported using AI in some form.

Yet the proportion confident that their organisation could manage an AI rollout without disrupting operations fell from 75% to 66% over the same period.

Comfort with autonomous agents handling activities from routine tasks to critical business processes also declined, from 11.7% to 6.3%.

The contrasting movements suggest that organisations are continuing to implement AI while becoming more cautious about the conditions under which it can be trusted to operate.

Deployment figures alone cannot establish whether systems are sufficiently reliable or whether the cost of implementation is justified by the resulting improvement.

An organisation may introduce AI into customer support, document handling or financial administration while retaining significant manual work around those applications.

Employees may need to inspect generated answers, correct mistakes, transfer information between systems or decide whether an automated recommendation is appropriate.

Infor’s findings suggest that this additional work remains substantial. UK respondents reported that 53% of AI-generated output still needed human review, compared with 49% in the United States, the next highest proportion among the seven markets.

Human review can be an appropriate control for consequential work, particularly where errors could create financial, legal or customer problems. Its operational cost nevertheless needs to be considered when calculating the efficiency of an automated process.

Legacy systems add to the implementation bill

Only 62% of UK respondents agreed that their organisation’s data was sufficiently mature and governed to support reliable AI, the second-lowest result among the surveyed markets.

Data readiness affects whether software can retrieve accurate information, connect records belonging to the same business process and identify the permissions governing their use.

A model may generate convincing responses while drawing on incomplete customer records or outdated operational information, requiring employees to investigate discrepancies before the results can be used.

Integration creates further demands. Almost one-quarter of UK respondents, 24%, said legacy or disconnected systems had increased the cost of preparing for AI, while 22% reported delays to deployment.

Some organisations may need to modernise interfaces, reconcile different data structures or establish new arrangements for controlling access before introducing an AI application.

Those activities can be expensive even where the model itself is available through a comparatively inexpensive subscription.

Businesses seeking to reduce operating costs therefore need to assess the complete process, including technical integration and review, rather than compare the price of an AI tool with the time taken by one employee to complete an isolated task.

A system that accelerates the preparation of a customer record may have limited overall value if the result must subsequently be transferred manually into another application.

Responsibility for AI is not always clear

Infor reports that 14% of UK respondents said no single person was responsible for governing AI use, managing risks or ensuring compliance within their organisation.

Responsibility can instead be distributed across executives, technology departments and committees, each with different priorities and operational authority.

Shared governance is not necessarily inappropriate, but uncertainty can arise when decisions about deployment, permissions and risk acceptance depend on several teams without one clearly accountable owner.

AI systems may require access to commercial information, employee records or customer data, making it necessary to determine who can authorise their use and which activities must remain subject to human approval.

More than one-quarter of UK respondents, 28%, identified compliance and regulatory concerns as an implementation barrier.

Those concerns vary between industries. A manufacturer may focus on production reliability and intellectual property, while financial or healthcare organisations can face additional requirements concerning the information used and the consequences of automated decisions.

The study does not establish that one specific regulation is responsible for Britain’s weaker reported gains. Rather, respondents identified several problems that can affect projects simultaneously.

Data quality, unclear ownership and integration costs may combine to delay implementation, even when a company has identified a commercially useful application.

Smaller businesses face a wider adoption gap

The research also identifies differences between respondents from small and large UK organisations.

Forty per cent of respondents from small businesses reported having no AI deployment, compared with 9% from large businesses.

Larger organisations may be better equipped to fund specialist teams and connect AI services to enterprise applications, although their existing technology estates can be more complicated.

Smaller businesses may have fewer internal resources for evaluating models, assessing information governance or managing changes to established workflows.

External software suppliers can provide packaged capabilities, but the suitability of those products still depends on how closely they match the work being performed.

Infor found that 67% of UK respondents believed generic AI products did not adequately reflect their business or industry’s requirements.

As a supplier of industry-specific enterprise applications, Infor has a commercial interest in that finding and uses it to support its strategy for specialised AI tools.

The survey records respondents’ perceptions of generic software; it does not independently demonstrate that Infor products outperform alternative technologies or produce a defined financial return.

Claims about the effectiveness of industry-specific software would require comparative evidence covering deployment costs, output quality, operational performance and longer-term customer outcomes.

The European findings reveal related constraints

Infor also provided a separate EMEA analysis of 772 business decision-makers in the UK, France and Germany.

Its accompanying infographic reports that 71% believed generic AI did not fit their industry, while 31% identified data security concerns as a principal barrier to implementation.

A lack of internal AI talent and the high cost of AI were each cited by 20%, while 19% reported the absence of a defined internal AI strategy.

The regional findings also show that 44% of AI output still required review by a subject-matter expert, according to respondents. Thirty-seven per cent reported full-scale AI deployment, while 55% expected their investment to increase over the following year.

These figures refer to the combined three-country sample and must not be substituted for the corresponding UK-only results.

In particular, the 71% EMEA figure concerning generic AI suitability differs from the UK’s 67% result, just as the regional 44% measure of output requiring review differs from the 53% reported for Britain.

Keeping those populations distinct is necessary for a meaningful comparison. Combining them without identifying the relevant sample would create a misleading impression of the position in the UK market.

Investment remains stronger than measured results

Despite the implementation problems reported, approximately half of UK respondents expected their organisations to increase AI expenditure over the following 12 months.

Their priorities were practical: 39% identified reducing time spent on routine tasks, 34% cited operational costs and 26% pointed to workflow automation.

Those objectives can be measured, but the appropriate evaluation needs to consider the complete business activity rather than the speed of one automated step.

Reducing preparation time may not lower expenditure if employees spend more time reviewing outputs or maintaining the surrounding systems.

Similarly, automating part of a workflow may produce limited benefit if approval procedures, inaccurate records or incompatible applications continue delaying its completion.

Infor’s results suggest that British organisations are investing while still working through the data, process and governance requirements needed to obtain value from those investments.

The company’s industry-specific product strategy represents one commercial response, but the research does not establish that specialised software is the only solution or that a particular supplier’s approach has been independently validated.

Future survey waves may show whether broader deployment is accompanied by stronger reported efficiency gains and greater confidence in operational control.

For now, the difference between the UK’s 23% reported efficiency improvement and the 34% average elsewhere provides a measure of the gap described by respondents, while the underlying causes and financial consequences require evidence from actual deployments.