UK firms push AI beyond pilot stage, Datadog finds

Datadog has published research showing that most UK organisations have moved AI beyond the pilot stage. Only 6% of UK respondents said they were still experimenting, compared with 25% in France and Germany.

The findings are based on a survey of 756 IT decision-makers in the UK, France and Germany at organisations with more than 1,000 employees that use AI. Across those markets, AI now runs in an average of 29% of enterprise applications, underlining how quickly the technology has moved into day-to-day operations.

That shift has increased reliance on AI tools in core business functions. Four in 10 European IT leaders said switching off individual AI-driven functions, such as incident detection, infrastructure optimisation or code generation, would cause significant disruption.

Yet operational control has not kept pace with adoption. Only 22% of UK IT leaders said they had complete visibility into AI systems in production, while 16% reported low or no visibility into AI usage and costs.

Control gaps

Security and compliance were the leading AI challenge for UK respondents, cited by 25%. More than a quarter, or 26%, said AI adoption in their organisations was moving faster than governance.

Problems linked to AI failures or unexpected behaviour were also widespread. Datadog found that 97% of organisations had experienced some form of negative impact over the past year, including customer-facing issues for 34% of UK respondents and security incidents for 32%.

The figures point to a gap between enthusiasm for deployment and the systems needed to monitor risk. Only 41% of UK IT leaders said they were confident they could detect an AI problem before it affected a customer, compared with 58% in France and 57% in Germany.

Operational complexity is also rising as AI spreads across business systems. In the UK, 37% of respondents said AI had increased operational complexity, suggesting organisations are now dealing with a broader mix of models, software and governance demands than many anticipated.

Cost pressure

Costs were another area in which UK organisations stood out. Nearly two-fifths, or 38%, said AI spending had exceeded expectations, compared with 15% in France and 19% in Germany.

At the same time, only 23% of UK companies said they actively monitor and optimise AI usage and costs, well below the 38% European average. This suggests many businesses are still struggling to connect AI spending with operational value and oversight.

The survey also found that, despite those concerns, AI is being trusted with more consequential tasks. Every organisation surveyed said it expected to give AI unsupervised control over at least one high-stakes action within the next two years.

In the UK, 29% of respondents said they expected AI to approve and execute financial transactions without human supervision. Another 28% said it would interact directly with customers, while across Europe 26% said AI would make hiring, firing or other personnel decisions without human involvement.

The contrast between limited visibility and expanding responsibility is likely to sharpen scrutiny of how companies govern AI in practice. As businesses move from pilot projects to broad deployment, they face more complex questions around reliability, accountability and oversight, particularly when systems are connected to customer-facing or regulated activities.

Pejman Tabassomi, Field CTO at Datadog, said: "As AI moves from assistant to autonomous operator, the operational challenge changes completely. It is no longer enough to know that an AI system is running - organisations need to understand what it can access, what actions it can take and when its behaviour starts to deviate from what was intended. The answer isn't to slow AI down. It's to build the visibility, security and guardrails that allow organisations to give it more responsibility without giving up control. The organisations that succeed won't necessarily be those deploying the most AI. They'll be the ones that can scale it securely, understand its impact and stay in control as autonomy increases."