EIM as the Foundation for AI Success in the Financial Sector
By Roberto van den Heuvel, Enterprise Information Management Specialist for the Financial Sector at OneXillium:
In this article, I share observations from the field and explain why EIM and AI are an inseparable pair. In my conversations—including those with banks, insurers, pension funds, and financial intermediaries—I’ve noticed that AI is being used more and more in their day-to-day business processes. This is a positive development, but it’s something worth giving careful thought to.
AI Activities I Observe
For example, I know that insurers are using AI to assess claims more quickly, with algorithms automatically analyzing claims and evaluating them for consistency and potential fraud. Banks are increasingly using AI to monitor transactions and detect suspicious activity more quickly, which is essential in the fight against money laundering and even terrorist financing. Pension advisors, in turn, are exploring how AI can help provide personalized advice to participants regarding their pension accumulation. It is used to identify potential pension gaps by analyzing large amounts of financial and demographic data.
Are you a financial intermediary? You’ll likely be using AI in this context as well, in the form of an AI agent that customers can ask questions of. In any language, these agents provide faster answers than you’d expect from a human—error-free and consistent.
Are there any additional risks?
These developments offer enormous opportunities for more efficient processes, better customer service, and risk management. But they also bring new challenges—challenges directly related to the quality, manageability, and organization of data. What may not help—and thus poses an additional risk—is that financial organizations have grown increasingly larger as a result of mergers and acquisitions. Which data can be trusted when databases contradict each other?
Possible consequences
AI models sometimes produce unexpected or even unreliable results. Just ask the victims of the benefits scandal. A faulty AI model detected fraud, and innocent Dutch citizens were wrongly labeled as suspects. But there are other conceivable consequences. Questions regarding compliance remain difficult to answer if auditors cannot be given clear-cut answers. Customer communications may also lack the personal touch that was originally intended.
Don't Blame Anyone
Nine times out of ten, the problem isn’t with the technology, but with the fundamentals: the quality and organization of data. But what exactly is customer data? It may seem like an odd question, but I want to draw your attention to the fact that (customer) data comes in many more different forms these days. It’s no longer just about paper files—such as insurance documents and policies—that have been digitized. These days, it also includes photos (for claims processing), WhatsApp messages (customer contact), and perhaps even video. Outdated document management systems are often unable to handle this effectively. Moreover, this data appears across many communication channels, which can lead to data fragmentation and the loss of a 360° view of the customer. The downside of all this technology has become decentralization, with the associated risks when it comes to the security of this data.
“You can use the smartest algorithms, but if your data is incorrect or inconsistent, you’ll mainly end up increasing errors, not insights.”
, Enterprise Information Management Specialist, Financial Sector
AI decisions must be transparent and justifiable
Another area I get a lot of questions about is compliance. Regulators such as the DNB and AFM are imposing increasingly strict transparency requirements, especially when AI solutions influence decisions. Suppose you use AI to automatically approve or reject customers’ loan applications. In that case, you must be able to demonstrate exactly which data, analyses, and considerations that decision is based on. Without a centralized and well-managed information system, that’s impossible.
The need to centralize data has never been greater
That’s my personal conclusion. Let me explain: if you feed AI models fragmented data that you can’t rely on with certainty, then AI agents, for example, will start generating incorrect answers. What’s more, the technology has advanced to the point where AI agents can also ask each other for advice. So if the data is incorrect, the consequences of “digital gossip” will be incalculable! Just as quickly as AI models can retrieve accurate data, they can also spread incorrect data.
Future-Proof Financial Services
Enterprise Information Management (EIM) isn't just about organizing data; it's about creating a solid, forward-looking foundation. A sound EIM approach ensures that AI initiatives don't stall but can instead gain momentum.
Specifically, this means that EIM helps to:
- Make data centrally and consistently available (Single Source of Truth).
- To continuously ensure the quality and timeliness of information.
- To provide insight and traceability for audits, compliance, and governance.
- Manage various types of data—both structured and unstructured—effectively
- To safely and reliably support innovations such as AI agents.
With EIM, you not only gain control over your data, but also create opportunities for innovation, efficiency, higher customer satisfaction, and (even) safer handling of privacy-sensitive data.