Scaling AI in financial services: Europe’s test is execution

In this article, she explores what it will take for financial institutions to turn widespread AI adoption into sustainable business value.

European Commission President Ursula von der Leyen described artificial intelligence alongside climate change as the “tipping point of our time” during her State of the Union address on September 16, 2026

The debate on AI in financial services is no longer about whether AI creates value by scaling. The evidence increasingly suggests that it can and it does.

In Singapore, the Development Bank of Singapore (DBS) recently declared more than S$1 billion in measurable value from AI initiatives this year, which is roughly €680 million, an increase from the 2025 figure of S$750 million.

This value was generated by the deployment of AI at scale – hundreds of use cases and nearly two thousand models.[1]

It seems there is certainly value that can be delivered from the use of AI at scale. 

However, we are currently scaling from human-led financial services to AI-enabled and increasingly AI-delegated financial services and that requires a series of operational foundations that are proving increasingly challenging. 

The operational foundations for scaling AI

Research suggests five operational foundations consistently distinguish institutions that successfully scale AI from those that remain trapped in experimentation.

1.  Data quality and accessibility remain the single most important prerequisite.

Financial services organisations possess enormous quantities of data, but much of it remains fragmented across product lines, countries, legal entities and legacy systems.

The effectiveness of artificial intelligence depends less on the volume of data and more on the ability of the AI tool to access high-quality, governed, interoperable data at scale.

It also requires context – what does "exposure," "claim," or "counterparty" mean consistently across systems built over decades for humans and applications, not for AI and agents. Most firms don’t have an “agent-ready” layer in their data infrastructure and that is increasingly exposing those who can move beyond simple efficiency and labour-substitution applications of AI to more fundamental use cases.

The Financial Stability Board (FSB) has identified data quality and governance as one of the key vulnerabilities associated with wider AI adoption in finance and noted that third-party concentration, model risk, cyber risks and data governance remain among the most significant constraints on wider adoption. [fsb.org]

2. Cloud and digital infrastructure are becoming strategic enablers

Generative AI and advanced machine learning require scalable computing, storage, model-development environments and access to specialised hardware. The challenge for Europe and the UK is that much of the AI value chain, including cloud infrastructure, semiconductor manufacturing and frontier models, remains concentrated outside of both the UK and the EU.

The ECB has noted that widespread use of AI could increase operational concentration risk and dependency on a limited number of AI and technology providers. The same concern appears prominently in the FSB's assessment of systemic AI risks. 

3. Governance and control frameworks must mature at the same pace as innovation.

Scaling AI requires model governance, validation, explainability, monitoring, human oversight, data lineage and accountability structures that are already familiar to financial institutions through prudential regulation and model risk management.

The industry's comparative advantage is that FS is already one of the most heavily governed sectors in the economy. The challenge is extending existing governance approaches to large language models and more autonomous AI systems.

We can see that AI deployment in European financial services is becoming widespread across customer support, fraud detection, AML, risk assessment and internal processes, reinforcing the need for robust governance frameworks.

4. Skills are rapidly becoming a binding constraint.

Financial institutions need not only data scientists and engineers but also AI-literate risk managers, AI-literate compliance officers, AI-literate auditors, AI-literate lawyers and AI-literate business leaders to allow them firstly to navigate the challenges of AI and secondly to optimise value from it.

Those firms that have been successful in scaling increasingly recognise that AI transformation is as much an organisational change programme as a technology programme.

5. Institutions need executive sponsorship and enterprise operating models.

Evidence from large banks in Europe and North America suggests that decentralised experimentation creates numerous pilots but little scale.

Organisations that treat AI as a strategic transformation programme with central oversight tend to achieve greater value creation and faster deployment.

Importantly, most successful deployments today remain focused on augmentation rather than autonomy. AI is proving highly effective as a co-pilot for employees, analysts, customer service teams, software developers and compliance functions.

Fully autonomous decision-making remains comparatively limited and that’s where the challenge of scaling becomes clear. The Bank of England and FCA report that only a small proportion of AI use cases involve fully autonomous decision-making.

The European challenge of scaling

The UK and European challenge at the moment is not a shortage of AI use cases but more a shortage of scale, capital and infrastructure.

If we examine three key regional models, each of which is very different:

 One of the central questions over the next few years is whether Europe can create the conditions for AI scaling at the same pace as the US and parts of Asia.

1. Capital availability and investment capacity

History tells us that transformative technologies have been scaled in environments with deep pools of risk capital and strong innovation ecosystems.

The United States continues to have significant advantages through its venture capital markets, private equity ecosystem, technology sector and concentration of global AI providers. The largest AI foundation model developers, hyperscale cloud providers and semiconductor innovators are predominantly US-based. This creates a virtuous cycle in which capital, talent and technological capabilities reinforce each other.

The US benefits from an integrated AI ecosystem. Europe and the UK remain highly dependent on importing critical elements of the AI value chain. 

Europe's challenge is somewhat different. European financial institutions are generally well capitalised, but Europe has traditionally been less effective at mobilising large-scale private investment into frontier technologies and scaling technology companies. As a result, many European firms remain consumers of AI technologies developed elsewhere rather than producers of them.

This creates three economic dependencies:

• Dependence on non-European foundation models.
• Dependence on non-European cloud providers.
• Dependence on non-European compute and semiconductor capacity.

The ECB has highlighted the growing concentration risks associated with a small number of AI and technology providers. At the same time, the FSB has similarly identified third-party dependencies and concentration risk as potential systemic concerns.

2. Computing capacity and infrastructure sovereignty

The other critical issue gaining attention in financial services discussions in recent months is access to computing power. 

Former President of the European Central Bank  Mario Draghi highlighted in a recent Financial Times article that Europe hosts only three times as much AI computing capacity as a single data centre site being completed in Malaysia this year. [2] 

European model developers collectively earn less than 2% of the revenue of just two US counterparts and we know that European companies and institutions were recently cut off from access to frontier AI capabilities.

The next phase of AI deployment, particularly generative AI and agentic systems, will require exponentially greater computational capacity. This raises questions around:

• Data-centre availability
• Energy infrastructure
• GPU and semiconductor access
• High-performance computing capacity 

The United States and China have invested heavily in AI infrastructure. Singapore, South Korea and some Gulf jurisdictions are also making substantial strategic investments.

Europe faces a risk that financial institutions may want to scale AI faster than the underlying regional infrastructure permits.

This is not simply an operational issue, it is increasingly a question of industrial policy and strategic autonomy. 

The next phase of AI scaling will probably be determined not only by what individual institutions do, but by whether Europe creates the economic conditions needed to scale.

They are increasingly concerned about access to capital, compute, energy, cloud infrastructure and technology ecosystems.

The US currently leads because it combines innovation, funding and scale.

The question is whether we can move quickly enough to ensure that European financial institutions remain producers of value from AI rather than merely consumers of technologies developed elsewhere.

The real test for financial services is no longer whether we can build AI models or whether they can deliver value. We can and they do. It is whether institutions can combine trusted data, resilient infrastructure, strong governance and skilled people. In the end, successful AI adoption at scale will depend less on the intelligence of the technology and more on the organisation’s readiness, the availability of risk capital and the computing infrastructure to support AI for financial services organisations at scale.

[1] DBS declared a revenue uplift of over S$1 billion this year from AI

[2] Europe’s difficult choices on AI