Is Your Data Ready for AI? Possible pitfalls in Financial Services
By Martin Sexton, Senior Business Analyst at Finworks
The conversation around AI in financial services is moving fast, and for good reason. But as organisations race to layer AI on top of their data systems, the questions that matter most are often the ones being asked last. Having worked closely with financial data infrastructure across complex, multi-jurisdictional organisations, I've seen the difference between AI that genuinely improves decision-making and AI that simply accelerates it. The distinction comes down to what sits beneath it. Here is my perspective on where the real opportunities lie and how the real risks can be managed.
Financial Services Data Hubs – Challenges in AI Use
There are obvious advantages to using Artificial Intelligence (AI) in Financial Services, these include:
- Increased productivity
- Quicker investment decision making
- Improved financial crime and surveillance
However, there are a couple of areas that organisations will need to consider when deploying AI architecture:
- Managing Confidentiality - sensitive data should only be accessible to those who have the appropriate access rights to do so.
- Trading Strategies impact on Market Stability – impacts to algorithmic trading due to changes resulting from the introduction of AI.
Managing Confidentiality
Deploying AI technology, data governance needs to be at the forefront. Within large organisations, spanning several jurisdictions, parts of the organisation may only have rights to access certain data within a repository. Whether adhering to internal or vendor access rights any AI implementation will need to support confidentiality. Sensitive data should only be accessible to tools and users who have rights to do so.
Public domain information may not provide organisations with the appropriate level of accuracy. Hence institutions use data vendors to enhance this. By consolidating data from several sources, only then can a real understanding of risk can be understood to effectively support the investment decision making process. Organisations are required to put in place an access rights layer to support the access rights at several levels, data sets, row and attribute levels and in some instances at an atomic level, by content level expressions. And any AI architecture will need to adhere to this when undertaking data integration.
The organisation needs to consider how best to support this. The application of access rights is a key consideration and there are two approaches:
- Apply permissions to both queries and responses. This is easier to implement as roles can be aligned with database views. Less risk of sensitivity leaks, as roles can be aligned with Database credentials.
- Queries are applied to the entire repository and access rights are only applied to results. Deploying masking of information where appropriate. More complex to manage as the access rules applied at time of the request is processed.
Trading Strategies impact on Market Stability
AI’s impact on the investment decision making process is another consideration. In the recent bulletin “Financial stability in the age of artificial intelligence: the role of algorithmic architecture “, this risk has been explored by the ECB Financial Research team.
Organisations are expected to put controls in place to ensure market stability. Trading strategies must avoid excessive and inappropriate buy, sell or redemption scenarios. AI could either result in more cautious or excessive trading outcomes depending upon the AI algorithm employed. Given these possible worse case scenarios, it seems appropriate to undertake stress testing to fine tune these algorithms.
Conclusions
The benefits of AI in financial services are real, but so are the responsibilities that come with it. Confidentiality, access governance, and market stability are not implementation details to be resolved after deployment. They are the foundation on which trustworthy AI is built. Organisations that treat them as such will be the ones that genuinely improve the certainty and explainability of their decisions rather than simply accelerating them.
Finworks helps financial services organisations build the data infrastructure that makes better decisions possible, not just faster ones.
