By Sarva Srinivasan, Global Head of Strategies at NeoXam Americas
Another important part of the EU AI Act has come into play – with Article 50 introducing transparency obligations for certain AI systems, requiring organisations to disclose when users are interacting with AI.
Now much of the discussion around these requirements has tended to focus on compliance. Firms are asking whether their chatbots need disclaimers, whether AI-generated outputs need labelling and what governance processes must be put in place.
While these are, of course, really important questions, they are not the most pressing question for financial institutions to answer right now.
The real issue facing the investment management side of the industry surrounds the transparency of the data feeding AI. Many asset managers, particularly those who have invested heavily, have discovered that they are not ready to deploy AI as their underlying data remains fragmented, inconsistent and nigh on impossible to govern.
Explainability Challenge
This is imperative because transparency is not something that can simply be added at the final stage of an AI workflow. If an investment manager cannot explain where data originated, how it was transformed or whether it can be trusted, it becomes significantly harder to explain how an AI system arrived at a particular conclusion.
In practice, the transparency obligations introduced under the EU AI Act may accelerate a trend that was already gathering momentum. Financial institutions are increasingly prioritising data modernisation projects, consolidating data sources, retiring legacy platforms and investing in more robust data management frameworks before moving aggressively into AI adoption.
In fact, some are even focusing on creating a trusted data foundation first, whether through cloud data platforms, investment data hubs or broader operating model transformation programmes. This is because they recognise that AI is only as effective as the information it can access.
Rising Costs
The industry is also beginning to confront a second reality. Certain financial institutions have begun to explore generative AI. Some have experienced a rather rude awakening when they discovered that the costs of token consumption can escalate rapidly without delivering corresponding value to the business.
As a result, attention is shifting towards more targeted approaches. In some cases, asset managers are linking machine learning capabilities and traditional automation with agentic AI in order to switch generative AI for the activities where it delivers the most benefit.
The rationale for this approach, which aims to make governing decision-making easier, is to show rule makers that transparency and accountability are very much front of mind.
The arrival of Article 50 should therefore be viewed as more than another compliance deadline. It is a reminder that successful AI adoption requires strong foundations.
Financial institutions that invest now in data quality, governance and operational transparency will find themselves best placed to both meet the expectations of watchdogs and to realise the genuine commercial value AI promises.
The firms that succeed in the AI era won’t necessarily be those deploying the most advanced models. They will be the ones that can clearly explain the information, processes and decisions that sit behind them.
Subscribe to our newsletter
