4 Reasons Why Artificial Intelligence Hasn’t Revolutionized Post-Trading Yet, but Maybe Will Someday

4 Reasons Why Artificial Intelligence Hasn’t Revolutionized Post-Trading Yet, but Maybe Will Someday

Post-trading is a volume business and a data business, so calls for the use of artificial intelligence (AI) in post-trading operations haven’t been long in coming. Read below exactly why this makes it important to take one step at a time when it comes to identifying potential AI use cases.

There is enormous hype and controversy surrounding artificial intelligence (AI). AI will either vanquish cancer or wipe out humanity, or it will either solve all of our productivity problems or make workers superfluous, so it’s said. And what about AI in the financial sector?

Who Is Already Using Artificial Intelligence in the Financial Industry?

In its 2023/24 Future of Finance study, SIX discovered that only 6% of financial institutions around the world are proceeding on the assumption that they won’t be able to employ AI in their business in a major way in the next three years. Study findings published by the European Securities and Markets Authority (ESMA) in February 2023 show that financial institutions’ thinking about AI is focused on the future and that the links in the financial industry value chain are not all affected by AI to the same extent. ESMA’s study titled “AI in EU Securities Markets” concentrates on post-trading and reveals that neither central clearing counterparties nor central securities depositories are widely using AI at present.

Post-Trading and Artificial Intelligence

But why aren’t they? Is post-trading a special case? Perhaps the answer lies in the fact that post-trading relies on absolutely trustworthy data, even more so than other parts of the financial industry depend on it. A regulated infrastructure cannot conduct its business on the basis of information from unknown or uncorroborated sources. The repercussions for risk management and investor protection would be enormous. There’s a risk of technology helping to multiply existing errors and defects.

Hurdles for AI in Post-Trading

Nevertheless, there is a place for AI even in post-trading, and like with other technological advancements in the past such as cloud computing, blockchains, and data science, a deep understanding of AI must precede any potential successful deployment of the technology. Four hurdles in particular must be overcome:

1. Dispersed Data

The data required to train an AI and ultimately put it to use unfortunately are often dispersed across different systems, exist in diverse formats, or are inaccessible altogether. Any attempts to use that information to predict settlement failures in trading transactions, for example, are doomed to founder if the data were not structured and normalized beforehand.

Once data have been structured and normalized, only then can financial institutions begin to think about where the actual benefits of AI exist for them. In the context of settlement failures in trading transactions, historical data could come into consideration, for example, but so could data on stock-price volatility, trading volume, the actual time of trades, the counterparties involved, etc.

2. Misplaced Focus

When it comes to determining what areas to focus on in the search for potential AI use cases, there’s a tendency to concentrate on high-volume transactions. However, given the nature of AI, a “many a small thing makes a large thing” approach would be more promising under some circumstances.

This also includes identifying what already works well. AI can help issuers, counterparties, trading venues, and clearing houses to make those use cases even better. The slightest optimization of a settlement process step multiplied thousands and thousands of times over harbors huge growth potential.

3. Inexistent Data Strategy

Regardless of the industry or company in question, a data strategy is necessary to put AI to use. This means not just identifying the requisite data and making it accessible, but also understanding the origin of the data, its ownership structure, its governance, and the access rights to it. This is tantamount to a learning process for the entire business enterprise.

Everyone, from the board of directors to the executive management, must be aware of the importance of data and needs to get involved. Organizational changes or reallocations of resources may be necessary to deal with resistance to new and perhaps experimental data models and processes. At the worker level, further training is needed to acquire mastery of the skills required to enable efficient utilization of AI. It is important here to provide possibilities to conduct tests in a safe and secure environment.

4. Faith in Technology

Steps must be taken to ensure that the right data flows into AI models. However, experts who studiously examine and verify what AI is actually outputting are also needed. Human common sense, which self-evidently only people can have, becomes an indispensable asset.

AI: The Machine Is Not Enough

It is also important to consider what roll transparency and ethics are to play in order to avoid bias, exclusion, etc. Trusting in the machine alone will not be sufficient. Financial institutions, not least those that are involved in post-trading, are right in taking time to examine potential AI applications calmly and with due care.