Banks are investing heavily in disruptive technologies to boost operational efficiencies including within the post-trade settlement process. The technology financial executives are turning to is called artificial intelligence.
Artificial Intelligence is a group of technologies that are designed to emulate human cognitive functions and traits. AI can perform language processing, understand complex data sets, and draw conclusions from that analysis.
Machine Learning is a type of AI that analyzes vast amounts of data, looking for complex patterns, differences, and gaps in the process. Over time, the program improves its ability to predict and classify data, and as a result, more comprehensive
data-driven decisions are achievable.
One of the key differences between ML and traditional legacy IT systems is that with legacy systems, new code would be needed to analyze any new data that’s introduced. With AI and ML, the program improves on its own, learning through repetition. As new
data is introduced to the process, little or no additional code is needed.
“At RBC Capital Markets eight years ago, seven out of 10 junior hires in the equity trading unit had a business background. Now, 70 percent of newbies are engineers.” –
Machine Learning And Trade Settlement
The vast majority of trades are settled through straight-through processing. These trades are executed, confirmed, and settled with little or no involvement by middle office or back-office staff.
However, for a number of reasons, trades can fail, meaning they don’t settle properly. Failed trades are rejected and kicked out of the straight-through processing system and into an unconfirmed queue that needs to be settled manually.
Middle-office staff must perform exception processing by identifying the reason for the rejection and fix the trade. Resolving the failed trade includes confirming settlement instructions over the phone between the trading desk and the counterparty. Other
times resolution involves communication between the back office and trading desk staff to resolve discrepancies in the financial details of the trade or resolve limit violations. As a result, exception processing of rejected trades becomes a labor-intensive
The exception processing of rejected trades by middle office and back-office staff comprises of the vast majority of the costs associated with trade settlement.
3 Ways AI And Machine Learning Improve Settlement Clearing
Although there are many applications of AI for trade-settlement clearing, the overall process can be broken down into three steps.
AI technologies can analyze vast amounts of historical trading data identifying any trades that failed. Also, the AI process can be done in a fraction of what it currently takes for back-office personnel to perform the analysis. The result is complex trade
data gets broken down and identified far more quickly and efficiently providing a clearer picture of the scope of the problem.
Once the failed trades have been identified, AI can analyze why the trades were rejected. AI spots anomalies, gaps, or what’s missing in the settlement process. There are many reasons for a failed trade. The financial details could be inaccurate due to trader
error, settlement accounts could be outdated or closed, or a spike in market volatility could have created a limit exception since the trade was booked.
Given the volatility of capital markets, identifying an incorrect trade and remedying the situation sooner, by even a few hours, can amount to significant cost savings. This is particularly true if the trade needs to be of unwound and the market had moved
against the trader’s position.
“It takes a human five to ten minutes to reconcile a failed trade. A bot can do it in a quarter of a second.” – American Banker
As a result, AI can identify failed trades, but also analyze and provide the reason for the fail, allowing a remedy to be implemented.
After identifying failed trades and spotting the gaps in the process, AI through pattern recognition analysis can predict the likelihood of which trades might fail and propose solutions. Machine learning occurs as the algorithms evolve through this repetitive
analysis. As AI learns, its predictive capability improves over time creating a far more efficient settlement process. Over time, machine learning can reduce the percentage of failed trades leading to fewer exception processing situations, and lower operational
AI: The Added Layer Of Security
It’s important to note that AI doesn’t need to operate within a vacuum, meaning it can act as added layer of security in catching failed trades. By incorporating AI into the current settlement process, back-office operational efficiencies can be enhanced.
For example, in a typical back office, end-of-day procedures or tie-out involve generating reports to confirm that trades have been confirmed between counterparties, settled correctly, and that market positions for the bank are in balance.
With AI, the end-of-day checklist can be performed intraday allowing middle office and back personnel to target the exceptions and fix the trades before there’s any impact on the bank’s trading position, P&L, or the client’s account.
AI is capable of detecting complex patterns in data, identify gaps in the settlement process, and as a result predict outcomes far more quickly and effectively than the current process.
AI is still in the infancy stage but is attracting increasing amounts of investment by financial institutions. Both the operational efficiencies and costs savings of implementing AI and machine learning are expected to be significant in the years to come.
“According to a report last year from Goldman Sachs, machine learning, and artificial intelligence will enable $34 billion to $43 billion in annual savings and revenue opportunities within the financial sector by 2025.” – CNBC
In summary, AI, through machine learning, can analyze complex sets of data, identify failed trades, and also provide the reason for the fail, allowing a remedy to be quickly implemented. For financial institutions, the expedited remediation of rejected
trades can prevent costly consequences such as market losses, negatively impacted clients, and trading limit or regulatory violations.