Though AI is now nearly ubiquitous across the financial, technology, and fintech sectors, many firms are still struggling to achieve full integration, highlighting a persistent “implementation gap”, according to a study by Cloudera.

The survey report, released in November 2025, polled 155 finance, tech and fintech executives worldwide in August 2025, and found that 97% of respondents had deployed at least one AI or machine learning (ML) use case. This signals that AI has moved from an emerging innovation to a strategic business necessity.

However, widespread deployment has not translated into deep adoption. Nearly half (48%) of the surveyed firms reported that their AI/ML maturity level has moved beyond experimentation and proofs of concept but is still not fully embedded in operations.

Data security and data siloes emerge as top barriers to AI deployment

The study found that data fragmentation is among the biggest barriers to AI deployment across regions and firm sizes.

An overwhelming 97% of the financial services firms polled specifically reported that siloed data across their organization is hindering their ability to build and deploy effective AI models. This suggests that data silos have become the critical fault line between strategic ambition and operational execution.

Large global organizations with more than 50,000 employees are the most affected by this, with 38% citing significant impact and 43% moderate impact. This reflects the inherent complexity of managing data across multiple business lines, as the more separated the functions, the more data silos emerge.

Smaller firms are not exempt from this. Among organizations with fewer than 1,000 employees, 25% reported significant impact from data silos and 40% moderate impact, showing that data silos challenges are affecting AI efforts at every scale.

At the regional level, LATAM showed the highest share of significant impact (45%), while North America was evenly split between significant (32%) and moderate (32%).

Europe reported higher moderate impact (61%), where regional regulatory requirements such as General Data Protection Regulation (GDPR) and open banking regulations have already forced institutions to confront data governance challenges. In the Middle East and Africa (MEA), 68% cited moderate concerns, suggesting newer, modernized systems may help mitigate the issue.

Data security also is a major barrier, reflecting heightened industry awareness of privacy risks and ethical responsibilities. High infrastructure costs add to the challenge, particularly in North America, while evolving AI regulations contribute to an increasingly complex and fragmented compliance landscape.

Top AI/ML use cases

The study found that chatbots and knowledge search leveraging large language models (LLMs) are the leading AI/ML use cases globally, with 70% of the surveyed firms either deploying or actively developing these AI use cases.

Adoption is the highest in North America and APAC (81%), reflecting both customer expectations for 24/7 digital engagement and the efficiency gains of automating frontline interactions.

These results align with broader consumer trends. A recent study by Genesys Cloud Services and Twimbit, which surveyed 1,400 consumers across seven Asian markets in October 2025, found that more than 70% of respondents had used a chatbot or a virtual assistant for customer support in the past 12 months. This indicates that AI-driven support is now a familiar part of the customer experience in the region as Asian customers prioritize fast response and resolution as their top customer expectation (80%) and as efficiency become central to positive experience.

Fraud and anomaly detection is the second most deployed AI/ML use case deployed globally, at 64%, with uptake being the highest in MEA (77%) and APAC (74%). This reflects a growing focus on financial crime prevention as fraud threats escalate across emerging digital economies.

A 2023 study commissioned by Lexis Nexis found that 42% of organizations in the United Arab Emirates (UAE) experienced a year-on-year (YoY) increase in online fraud year-on-year (YoY), incurring an average cost of AED 4.19 (AED 3.62 for retailers and AED 4.99 for financial institutions) for every dirham lost to fraud.

Across Europe, the Middle East and Africa, digital channels accounted for 52% of overall fraud losses in 2023, surpassing physical fraud for the first time.

In APAC, AI has ushered in more sophisticated fraud schemes, including deepfake documents, biometric spoofing, and enhanced impersonation. A 2024 APAC study commissioned by GB Group found that 70% of organizations in APAC saw fraud attempts increase over the prior year, with many reporting a surge in impersonalization of digital presence, account takeover fraud, and money laundering and money mules.

Over a fifth (22%) of APAC organizations said identifying fraudsters at the point of onboarding has become extremely difficult, a figure that rises to 31% respondents in Malaysia and 29% in Australia. Overall, 27% of fraud prevention professionals in the APAC region said identifying and stopping fraud at the point of onboarding is now one of the biggest challenges they face in their job.

Europe leads in full AI implementation

Globally, Europe leads in full AI implementation. While only 26% of firms worldwide have achieved full AI integration, 45% of European organizations have reached this stage, supported by the region’s strong fintech ecosystem and regulatory drivers from the EU AI Act.

In North America, organizations are concentrated just below full integration, with 39% at the stage preceding it and 35% fully integrated. Similar patterns appear in MEA and APAC where 61% and 58% of respondents, respectively, are just one stage short of full integration, and only 13% in each region having reached full integration.

Latin America (LATAM) stands out with 26% fully integrated, 13 points higher than both APAC and MEA, suggesting that the region is leapfrogging certain legacy barriers.

Featured image: Edited by Fintech News Singapore, based on image by freepik via Freepik