Standard Chartered Bank (SCB) is navigating the complexities of artificial intelligence (AI) adoption by grounding its strategy in well-governed, outcome-focused data, according to its group chief data officer, Mohammed Rahim.
Appointed to the role in December 2024 after 19 years working across various data functions at SCB, Rahim is steering the bank’s data initiatives to directly support its AI ambitions, shifting the narrative internally and externally.
“We’re moving away from the ‘data-driven’ term,” Rahim told Computer Weekly. “We’ve repositioned our narrative because it’s not data for the sake of data. It’s data in pursuit of an outcome, which is to service our clients better.”
This philosophy underpins SCB’s approach to both data management and AI deployment. The bank is currently finalising a formal AI strategy – due to be presented to its board in the middle of this year – that eschews technology for technology’s sake.
“Our strategy around AI is very much not AI for the sake of AI, but AI in pursuit of our business goals,” said Rahim, adding that the goals are centred on positioning SCB as a leading bank for affluent and cross-border clients, with a focus on sustainability.
Operationally, SCB is establishing a central AI platform alongside enhancements to its existing data platform. Rahim’s team, in partnership with their technology colleagues, is responsible for building the necessary guardrails.
“We will have a central team in my organisation that will build the guardrails for AI, such as things like responsible AI, and making sure we’ve got the right platform and technology ecosystem to support our AI strategy,” he said.
‘Hub and spoke’
This central capability supports a “hub and spoke” operating model, designed to push AI capabilities and use case development out to the bank’s various business lines and functions, ensuring alignment with specific needs while maintaining central oversight.
Early steps in operationalising AI include the recent roll-out of SC GPT, an internal large language model akin to ChatGPT, made available to 70,000 employees across 41 markets. Rahim said the tool, which has handled over 150,000 prompts, complements the more targeted development of high-impact AI use cases via the hub-and-spoke structure.
We’ve repositioned our narrative because it’s not data for the sake of data. It’s data in pursuit of an outcome, which is to service our clients better Mohammed Rahim, Standard Chartered Bank
Crucially, the success of these AI initiatives hinges on the quality, relevance and governance of the underlying data. SCB operates a bank-wide data lake, which is currently being modernised.
The focus, said Rahim, is to ensure the data is fit for specific purposes. “It’s making sure we understand what outcomes we want to deliver and making sure those the data points are correct for that outcome,” he added.
When developing AI models, SCB assesses data requirements on a use-case basis, but also increasingly by capability to avoid redundant efforts. For instance, rather than having multiple teams build separate language translation tools, SCB aims to ensure the data is correct for a central translation capability, said Rahim.
Beyond mere correctness, Rahim highlighted the need for data to be representative of the real world. He cited the concept of data drift, where changes to data, while accurate, could potentially skew AI model outputs.
“Before Covid, credit card companies would look at how much you travel to offer you air-mile cards,” said Rahim. “But come Covid, everyone’s air miles dropped to zero, and when credit card companies looked at their algorithms, they saw zero offers being made. The data was correct, but the algorithm hadn’t adjusted for the data. It was no longer representative of the real world.”
To that end, he said SCB is focused on ensuring data, and that algorithms are “continuously correct for the outcome as the world shifts”.
The primary application of SCB’s data is enhancing client experience rather than direct monetisation. Rahim described a pilot project using AI to assist call centre agents. By feeding policy documents into an AI model, agents can quickly query complex scenarios, such as the implications of early loan repayment, providing faster and more accurate answers to customers. “Client experience is at the heart of what we’re trying to do when we focus on our AI strategy,” he said.
Addressing the challenge of data silos in large organisations, Rahim acknowledged the tension between breaking down barriers and complying with increasingly complex regulations, especially when it comes to cross-border data flows. “How do we make sure we are breaking down data silos but not breaching laws? That’s the dilemma,” he said.
SCB’s solution involves modernising its data lake infrastructure to incorporate sophisticated access controls. “We still believe that a central data lake is the right way to go, but it needs to be modernised to cater for different restrictions,” said Rahim.
He described the concept as building “curtains around who can see what types of data”, with controls based not just on roles, but also on geographic location and data residency laws. Currently, SCB’s data lake infrastructure is mainly on-premise, but bank is evaluating the optimal mix of on-premise and cloud deployments to address complex requirements.
SCB has also established a responsible AI council that has developed a framework that integrates perspectives from data privacy, cyber security, architecture governance and risk management. AI models undergo rigorous checks against this framework before deployment, particularly around privacy and potential bias. “We’re very much focused on applying our responsible AI standards,” said Rahim. “We try as much as possible to avoid using very sensitive data in our algorithms because we don’t want unintended biases.”
Looking ahead, SCB is forming generative AI (GenAI) squads to work on high-impact use cases aligned with strategic goals in areas such as operations, finance and compliance. Concurrently, the bank is refreshing its responsible AI framework to address specific challenges posed by GenAI, agentic AI and open-source models, such as hosting locations, potential biases and security risks.
SCB is also actively working to uplift AI literacy across its workforce. This includes initiatives such as “promptathons”, which are workshops to teach employees how to effectively interact with AI models. Equally important is reinforcing individual accountability, ensuring staff understand that while AI can assist, the human remains responsible for the final output. “Whilst AI may rewrite your email for you, you’re still accountable,” said Rahim.
