AI improves banking operations but exposes execution gaps

AI improves banking operations but exposes execution gaps

At the Shanghai International AI Finance Summit 2026, Wang Kaijing of SenseTime discussed how AI-driven models enable banks to integrate structured and unstructured data to improve risk management and operational efficiency. Chak Wong, managing director at JP Morgan, explained why many AI initiatives fail, stressing the need for strong leadership, organisational alignment and domain knowledge.

In recent years, banks have integrated artificial intelligence (AI) and big data analytics to transform banking operations, enhancing internal processes, decision-making, risk management and customer service. 

Wang Kaijing, vice president of SenseTime’s fintech unit, explained how AI, including large language models (LLMs), expands access to data and enables  proactive, real-time decision-making, acting as the “wings” that support more strategic use of data.

At the core of this shift is the integration of structured and unstructured data. Traditionally, banks relied on structured datasets, such as transactions and customer records. Today, AI enables the analysis of financial reports, customer interactions, and external data sources in a unified way. This supports more nuanced and context-driven decisions across risk, operations and customer engagement.

In the past, business analysis centred on a small number of traditional indicators, such as deposit balances, loan balances and liquidity ratios. These key performance indicators (KPIs) were typically reviewed monthly or quarterly, limiting the scope of decision-making. Today,  banks track a wide range of performance indicators in real time. Wang highlighted that banks can monitor more than 400 data applications across various business processes and cover 100% of their staff, giving decision-makers broader range of real-time data to guide their actions.

AI solutions have significantly enhanced the way data analysts and engineers interact with data. For instance, Chinese banks using AI-driven tools have reported a 600% improvement in the efficiency of data application development, allowing business teams to develop data applications faster —from 0.35 to 2.68 transactions per day.

Enhancing risk management with data driven insights

Risk management remains a primary focus of AI adoption. AI-powered systems now monitor risk across the full loan lifecycle — pre-loan, in-loan and post-loan — providing continuous and granular oversight. By combining internal and external data sources, including behavioural and market signals, banks can assess risk more holistically and act pre-emptively, improving fraud detection.

Traditionally, rule-based engines assess creditworthiness and detect fraud. However, as Wang noted, these systems often lacked flexibility in assessing risks. AI allows banks to assess risk from various perspectives, including financial data, customer behaviour, and external factors like market trends and public sentiment. 

Banks also manage large volumes of complex text, such as regulatory documents, market research reports, and risk assessments. AI’s ability to handle long-context reasoning and multi-modal data —numbers, text, images and voice —improves both the speed and accuracy of data-driven decision-making.

AI is increasingly  used to support regulatory compliance, including anti-money laundering (AML) and fraud prevention. AI systems generate real-time compliance reports, flagging potential risks and violations before they escalate, helping banks meet regulatory requirements efficiently while  reducing reputational and financial risk.

Organisational gaps limit AI impact

Chak Wong, managing director at JP Morgan and global lead of the Machine Learning Centre of Excellence, explained  why many data and AI initiatives fail in banking. Failures rarely come from algorithms, computing power or talent; they mostly arise from misaligned incentives and organisational structure.

Wong emphasised that senior leadership is crucial. Without clear mandates and alignment of data initiatives with strategic goals,  AI projects struggle to deliver value.  Successful AI projects require patience, thoughtful planning, and incentives that support long-term goals.

Wong also stressed the importance of compatibility. While enthusiasm can start projects, long-term success depends on cultural fit, consistent daily practices, and proper data checks. Simple measures, like validating financial data, are often neglected, but crucial to avoid issues down the line.

Domain knowledge is also critical. Technical expertise alone is insufficient; understanding the business context ensures data is meaningful and correctly interpreted. He cited credit spreads in financial markets as an example, where contextual checks prevent errors. Specialists without broad operational knowledge may misinterpret data.

Wong distinguished forecasting from decision-making. Predicting outcomes is useful, but prediction alone does not guarantee actionable strategies. Leaders must focus on causal thinking—understanding the cause-and-effect behind data—to make informed decisions. 

Realising AI value through leadership and readiness

The future of AI in banking hinges on pairing technology with organisational readiness. Banks that align leadership, incentives and cross-department collaboration while embedding deep business understanding can turn AI insights into actionable decisions. Misaligned incentives or gaps between data teams and business leaders limit impact, even with advanced models. Prioritising causal thinking, discipline and long-term commitment enables institutions to translate AI capabilities into measurable operational and strategic value.

Keywords: AI,  big data, large language models, data analytics, real-time decision-making, structured data, unstructured data, operational efficiency, risk management, customer engagement, key performance indicators, data applications, fraud detection, regulatory compliance, anti-money laundering, causal thinking, organisational alignment, leadership, incentives, domain knowledge, cultural fit, cross-functional collaboration, data checks, machine learning, fintech 

Comments (0)
Cancel