The rapid development of AI, along with its increasing adoption, is placing unprecedented demands on traditional data systems, forcing businesses in the banking, financial services and insurance ( BFSI ) sector to prioritize between security, quality and sustainability.

But while technology and business leaders acknowledge the importance of data quality to achieve success in the field of artificial intelligence ( AI ), they are more focused on data security, leaving gaps in AI performance and long-term return on investment, a new report finds.

Almost half ( 48% ) of IT leaders in the BFSI sector say security is their top concern for AI implementation, reflecting the critical need to guard against internal and external threats, according to The State of Data Infrastructure 2024 report from Hitachi Vantara, the data storage, infrastructure, and hybrid cloud management subsidiary of Hitachi. In fact, 84% admit that losing data to an attack or mistake would be catastrophic.

According to the report, 73% of respondents don’t feel that a robust infrastructure is important for the success of AI projects. This is shocking, Hitachi Vantara says, since infrastructure is the key to providing and collecting high-quality data sources – securely.

“Financial institutions worldwide are accelerating AI adoption, but many are realizing their data infrastructure isn’t ready to support it,” says Joe Ong, vice president and general manager for Asean at Hitachi Vantara. “This global research reflects what we're also hearing in Southeast Asia – that the real barrier to AI success isn't the technology itself, but the ability to manage data securely, accurately, and at scale. Financial organizations must focus on strengthening their data foundations to ensure AI delivers real, sustainable impact.”

The real barrier to AI success isn't the technology itself, but the ability to manage data securely, accurately, and at scale

The study also reveals that:

• In BFSI companies, data is only available when and where it is needed a quarter of the time ( 25% ), and BFSI AI models are accurate only 21% of the time.
• About a third of respondents ( 36% ) are concerned about the risk of a data breach from internal AI, and 38% are concerned about the inability to recover data from ransomware.
• Although ransomware attacks are top-of-mind for BFSI IT leaders, 36% say a data breach caused by AI making a mistake is a top three concern for them, and 32% are concerned an AI-enabled attack could cause a data breach.

“The business model in financial services is inherently tied to trust. Reputational harm is a significant risk, and so in our industry, the interaction between security and accuracy is a critical and complex challenge,” says Mark Katz, chief technology officer, financial services, at Hitachi Vantara. “For instance, if a chatbot inadvertently discloses sensitive information that was included in the training data, that will have serious repercussions. Additionally, the cost of a wrong answer or a hallucination poses a significant risk; if someone were to act on bad data, it raises all sorts of questions about liability.”

The cost of a wrong answer or a hallucination poses a significant risk; if someone were to act on bad data, it raises all sorts of questions about liability

Despite accuracy challenges, AI adoption within BFSI is accelerating. However, many are deploying AI without adequate preparation, with 71% of respondents admitting to testing and iterating on live implementations, while only 4% are using controlled sandbox environments.

While financial services leaders are convinced that data quality is the most important consideration for successfully implementing AI, concerns like security are too urgent to ignore, and ROI is suffering.

The report outlines key considerations for building a more resilient, AI-ready infrastructure to help BFSI organizations prepare for the future. These include:

• Responsible experimentation. Two out of five BFSI leaders ( 42% ) said they were building the necessary skills to implement AI through experimentation. Responsible testing in secure sandboxes can mitigate risks while uncovering AI’s potential.
• Sustainability at every level. From energy-efficient data storage to optimized software, business and IT leaders must integrate sustainable thinking into their infrastructure, applications, models, data practices and strategies from the start.
• Simplify and unify systems. Reduce complexity by managing hybrid environments uniformly, automating security tasks and leveraging unified data platforms for faster insights and streamlined AI training.
• Ensure data resilience and leverage AI for defense. Plan for recovery with redundancy systems, roll-back storage and AI model restoration to mitigate risks from failures or attacks. Use AI to identify risks, enhance recovery and secure data with immutable, encrypted and self-healing storage, countering threats from AI-enabled attackers.

The report is based on a survey of 231 BFSI specialists, C-level executives and IT decision-makers spanning 15 countries across the globe.