Leaders from financial, technology and regulatory bodies shared their views on the adoption of AI, including the associated data challenges and ethical questions, during a panel discussion entitled "Mastering your AI Journey" at this year's Interact conference in New York, which was moderated by Madhu Ramu (Head of Product for Software, Enterprise Solutions, S&P Global Market Intelligence).

The panelists considered the technology from a data and operations perspective, the challenges of starting to use this technology, and its fairness and bias issues.

Gen AI is not the only AI game in town, however impressive and attention-getting, said an investment firm panelist. Large language models, the basis for Gen AI, can provide just as great a user experience, they said.

This is now the second year that Gen AI has been widely available, and companies are actually starting to put proofs of concepts into production. They still have to choose the right type of AI. The problems or confusion around Gen AI are not new, and firms have to figure out where to apply the technology, the investment firm panelist stated.

Even for firms that have gone live with Gen AI technology, some are struggling, possibly because the models they used were too large, and they did too little or the bare minimum of data engineering and prompt engineering.

Getting the data right is therefore a big part of getting Gen AI right, which could in turn fulfill the promise "big data" once had for deriving insights from large amounts of both structured and unstructured data. One panelist said Gen AI could be used to find and unlock data relevant to a task or analysis, especially when not all the necessary data has even been located.

When applying Gen AI or some other appropriate form of AI to unstructured data, financial services firms will have to earn trust from their clients that data is being collected, evaluated and processed fairly. For instance, the end result could decide whether someone or their company gets a loan or not. Even in training AI, panelists pointed out, there are concerns about what data is used and whether it introduces bias.

When taking data work like this and bringing it to bear on specific use cases, it can be useful to start a data usage review board asking what data assets are going to be used, whether rights to that data are needed and whether it is useful data or in a useful form. The circumstances of use cases can vary, but if results can be reviewed by people, as with testing an IT system, for example, that builds trust, according to a panelist.

Another concern often raised about applying AI in various fields is that it will replace people's jobs. For instance, financial firms who believe they have an AI solution just as effective as a first-year analyst, might think they could save time and costs by firing or not hiring more first-year analysts. The problem with this, one of the panelists observed, is that the firm will never end up with any mid-level or senior analysts - and AI cannot produce the kind of insights or analysis these analysts can.

S&P Global provides industry-leading data, software and technology platforms and managed services to tackle some of the most difficult challenges in financial markets. We help our customers better understand complicated markets, reduce risk, operate more efficiently and comply with financial regulation.

This article was published by S&P Global Market Intelligence and not by S&P Global Ratings, which is a separately managed division of S&P Global.