AI Investments only deliver profits for banks when built on strong data foundation
AI has revolutionised every sector, especially banking and finance.
AI has revolutionised every sector, especially banking and finance. Across Nigeria and elsewhere, banks are ramping up investments in efforts to integrate AI infrastructure into their core operations.
Unfortunately, most of these investments are not translating into profitability and may never do so. The problem is not that AI investment is a bad idea; certainly not. As a matter of fact, AI initiatives have been shown to considerably improve efficiency ratios for banks, according to PwC. The problem lies in the weak data systems/foundation that underpins these AI projects, as well as poor execution.
According to the World Economic Forum, banks are expected to invest $97 billion by 2027 in building out their AI infrastructure. The goal is to fully harness AI’s immense potential for hyper-personalised banking, automation of internal operations, fraud prevention, customer service enhancement, and improved credit scoring, among others. Each of these integrations presents immense benefits for banks. Some of these benefits include lower operating costs, faster processes, faster decision-making, error reduction, happier customers, and ultimately more profit for the banks.
However, investing billions into AI banking infrastructure without first securing a strong foundational data layer is tantamount to wasting money. This is because AI in banking will only be useful and guarantee your investment returns when built on a strong digital core. In its recently published whitepaper ‘The digital-first bank’s guide to AI in 2026’, core-banking technology provider Oradian noted that 95% of all AI projects by banks fail because the banks did not prioritise the integral role of data. It is the foundation on which AI applications function optimally. As the whitepaper rightly noted, “every AI initiative stands on the shoulders of data. The quality and accessibility of data can make or break an AI project. If your institution cannot reliably extract and unify its own data, it isn’t ready for AI.”
Therefore, before investing in any AI project, banks must first assess their readiness to implement the technology in such a way that yields value. After all, the goal of AI investment is not just to innovate, but to increase profitability for the bank.
One of the first things to do is to have a consolidated view of all your customers’ data. You should also be able to access detailed transactional data on each customer dating back at least 24 months. In addition to this (very important), your team should be able to query production-grade data off-core without the risk of downtime. Also, you must have very good data quality checks in place. Finally, make provision for incorporating external data, albeit in a controlled way.
It’s important to note that many banks (especially Nigerian banks) are currently operating on legacy data systems that cannot be interacted with, thus resulting in siloed or inconsistent outputs when queried. This lack of a unified data view makes it impossible for AI to reason, meaning that any AI infrastructure plugged into such a poor data system will basically be useless.
So far, we have established that AI’s success in banking depends on data readiness and digital core strength. But these are not the only necessary conditions for success. The reality is that another major reason why AI banking initiatives fail is because they are often fragmented/poorly executed and lack organisational alignment. A fragmented execution happens when AI projects are initiated in isolation from the business lines where they are intended to be embedded. The result is usually successful demos that never get integrated or become useful.
Therefore, it is imperative to have a clear-cut strategy and ensure that your systems are aligned before embarking on your AI project. This helps you to avoid wasting resources on a fragmented adoption that seldom delivers measurable value for the bank.
It’s a bad idea to attempt implementing all your AI initiatives at once, especially if you expect success. Not all AI projects deserve immediate attention. Simply explained, prioritisation here has to do with outlining, assessing, and ranking all AI use cases based on their need level and impact on your business, as well as the availability of data and the technical feasibility of actualising them. Scoring each AI use case against these yardsticks helps you to decide which AI project deserves to be tackled first.
“The goal here is to mainly focus on initiatives that will have the most favourable impact on the business, are feasible to accomplish, and are supported by clean, accessible datasets,” said Rodney Trivangalo, Vice President of Marketing at Oradian. According to him, doing this helps banks to deprioritise low-impact ideas whilst minimising the likelihood of AI initiatives failing.
This one is a no-brainer. You cannot afford to violate local laws and regulatory stipulations in the course of implementing the AI strategy. Therefore, it is imperative to ensure that your innovations meet all compliance requirements. You must not violate customers’ privacy in any way, and your AI lending decisions must be fair and not discriminate against certain groups.
In the same vein, banks with cross-border operations should also take extra care to study and align with the regulatory requirements on AI in all the countries where they operate. This is important because different countries have different regulatory frameworks.
Each country may have its own stance on AI, leading to a fragmented regulatory landscape for banks and lenders operating across borders. Regulators will expect banks to govern AI with the same rigour as other risks. This means incorporating AI into your enterprise risk management. Ensure your board or senior management is informed about major AI initiatives and signs off on them, similar to approving a new credit policy,” said Abdul Sulaiman, Oradian’s Regional Head for Africa.
In 2026, no bank can escape from integrating AI into its system. It is now inevitable. However, do not make the mistake of implementing AI for the mere purpose of innovation. Every AI integration must lead to profitable scale for the business. And the only way to do that is to ensure that you follow the processes discussed above.
Oradian’s whitepaper offers more exclusive insights and guides to implementing AI banking infrastructure the right way. Download your copy from their website here.
