Rebooting the global asset management industry

34 Rebooting the global asset management industry The Complexity of AI and GenAI is a Drawback Historically, an amalgam of two opposing sets of factors in asset houses has affected the adoption of new technologies: some moderating the pace of adoption, others accelerating it. The end result depends upon their relative strengths at a given time. Two clusters of factors are identified as moderators conspiring to slow the pace of the adoption of AI and GenAI at this stage. Some are related to these technologies and some to organizational factors. On the tech side (Figure3.2, toppanel), the cluster centers on three sets of factors. The first covers legacy self-contained ITplatforms that have evolved piecemeal over time, in response toa felt need, withdifferent chronological ages,making systems integration very difficult (59%). Being customized, suchplatforms strugglewith integratingdisparate data across different asset types leading tomultiple layers ofmanual reconciliation. GraftingGenAI onto themhas been challenging. The second set centers on difficulties in establishing the veracity of Big Data (54%) and their lack of explainability and transparency (51%), raising the risk of inaccurate responses and ‘hallucinations’ (49%). Thus far, the quality of data from internal and external sources has not earned market confidence (see Insights on the next page). On tokenization, for example, a lot needs to be done to ensure that the tokens correctly reflect the value and liquidity of their assets. The third set in the first cluster centers on the divergent regulation of AI and GenAI across global capital markets that exposes asset managers to all manner of reputational risks (47%). Another source of such risks is ethical and legal issues around the outputs from these technologies (44%). Their self-learning capabilities are a double-edged sword. On one side, they truly learn for themselves in order to deliver investible information and actionable insights; on the other, the resulting flexibility means that these self-learning systems are a black box, raising explainability issues. They also inadvertently increase the attack surface for malicious actors to generate synthetic media, conspiracy theories, cyber threats and racial discrimination. Until they become more transparent to their creators and accountable to their users, progress will be a matter of incremental steps, not giant leaps. Thus far, the quality of data from internal and external sources has not earned market confidence. Figure 3.2 Which factors, if any, have slowed progress in adopting AI and GenAI so far? Source: Citi/CREATE-Research Survey 2025 Tech-related % of respondents Organization-related Legacy systems Concerns about the quality of Big Data Lack of explainability and transparency Data security and IP infringement Inaccurate responses or ‘hallucinations’ Divergent national regulations on AI andGenAI Ethical and legal considerations Excessive hype about AI &GenAI Protecting client interests Legacy thinking Overemphasis on short-term profits Worldwide shortage of tech skills Incentives tied to short-term results Tenure rates of top executives too short 59 54 51 49 49 47 44 26 54 36 33 29 22 17 “The problem is not information overload, it is filter failure. The ownership of good quality proprietary data will be a key differentiator.” “Realizing the upside potential of GenAI is as much about leadership, culture and employees as it is about technology.” Interview Quotes