AI Can Drive Reconciliations Efficiency In Period of Increased Volume

Artificial intelligence can streamline post-trade reconciliation operations, says analyst Virginie O’Shea, reducing the time that staff spend on basic tasks and allowing them to focus on more complex and challenging items.

“There are no magic bullets for these problems, but in harmonizing front-ends across platforms, or enabling manual processes to be better supported by technology, artificial intelligence and machine learning could play a part,” said O’Shea, who is the founder of Firebrand Research. O’Shea was speaking on a webinar hosted by capital markets technology provider Broadridge on May 13.

Even in normal times, O’Shea said, post-trade reconciliations are a pain point for banks. Onboarding a complex reconciliation can take up to 75 days, as firms must first put together business requirements, analyze data flows, and then map them on to new processes. A patchwork of reconciliation systems means that consistency across platforms is lacking.

The situation has worsened during the coronavirus pandemic, as increased trading volumes spell an associated increase in the number of reconciliations that must be performed. Peter Webb, director of product management at Broadridge, was also on the webinar. He said Broadridge’s large US-based clients have seen their regulatory reconciliations leap to 19-times the norm.

O’Shea added that as trading volumes have increased, it’s created down-the-line stressors.

“The current market volatility has caused an increase in trading volume, which has been amply demonstrated over recent weeks by the equities performance of many banks and brokers, despite the market downturn,” O’Shea said. “Now all of the trading activity has had a significant knock-on effect on operations teams across the globe. Given that this is a global event, no market has been shielded from the effects, and it has in fact been magnified for many teams, especially if they are passing the book across the globe.”

Due to the siloed nature of many large financial institutions, these trades might not be in instruments the teams are used to dealing with. Thus, more external reconciliation processes must be conducted, which means late nights for ops teams and increased operational risk, O’Shea said.

At the same time, remote working and connectivity issues have stretched IT support teams, and a relatively obscure back-office function like reconciliations tends to fall down tech support’s list of priorities.

O’Shea said personnel are working from BCP back-up sites, or at home, straining VPN capacity and broadband connectivity. And this is at a time when people are stressed out by the lockdown, and might not be used to working on particular platforms.

An environment dominated by legacy tech is not conducive to flexibility in a crisis. “There is a high degree of key-person risk within both ops and IT as a result,” O’Shea added. “I would draw your attention to the current shortage in Cobol programmers—we are seeing ads for them across banks out there, and that tells me there is a lot of legacy tech lurking.”

Reconciliations may seem like a somewhat obscure, back-office issue, but regulatory interest in operational resilience generally has sharpened, and the C-suite cannot ignore these activities. The Financial Conduct Authority (FCA), for example, recently consulted on operational resilience improvements in the financial industry. Regulators expect executives to have a view of operational risk across the enterprise, and reconciliations can pose such a risk.

“Switching from trading one instrument to an entirely new type of instrument, often means that a firm must set up new reconciliations on a legacy system that was probably designed to cope with things like Swift messaging reconciliations only,” O’Shea said. “That means forcing data into a shape or process that it may not fit. It adds time, it adds complexity, and if you can’t force things to fit, they get done on spreadsheets and user-designed apps.”

Front Foot

O’Shea said firms need to be “on the front foot”: processes must be able to be passed from one team or individual to another; the C-suite must be able to have a view into op risk across the firm. Automation can help with this.

“Understanding your op risk capabilities is much easier when automation is high. The focus is on … enabling existing staff in operations and IT to focus on the more complex end of the spectrum so that when it comes to the key tasks they perform on a day-to-day basis—enablement, support—they can pass tasks from one person to the other in a seamless manner, and be able to respond to market conditions when they arise,” O’Shea said.

Some banks and vendors have attempted to tackle the automation of reconciliations with AI and machine learning. Brown Brothers Harriman, for one, is re-imagining its AI ecosystem, including using a trained algorithm called Linc. 

BBH has an integrated tech platform for its clients called Infomediary; a component of this platform is the InfoRecon application, which automates the reconciliation of cash and securities data across clients’ service providers. The firm has developed a machine learning model for this automation that pulls in client data and combines statistics with new forms of AI as an input in the reconciliations process.

Broadridge has a reconciliations solution within its product suite called Data Control. The vendor has added AI capabilities to several products in this unit, Webb said during the webinar. These include an additional module that allows users to analyze and deploy new reconciliations into the Data Control reconciliation platform. Users drag and drop files into the tool, which analyzes the data within those files and works out how to best map that data. The technology company has also partnered with Singapore-based vendor Tookitaki, to launch another AI-based module in Data Control. This one is designed to automate the management of breaks, which occur when data is missing or erroneous.

Vendor SmartStream Technologies has TLM Reconciliations, which combines AI with a user interface for reporting so that users can perform trend analysis. The company has also developed AIR, its AI-enabled reconciliations platform that allows users to manage reconciliations on an ad-hoc basis. AIR incorporates machine learning and NLP algorithms, and users can drop their records into it and reconcile discrepancies.

Xceptor has built out machine learning and natural language processing components for its platform. The vendor is helping Deutsche Bank in Indonesia to automate its reconciliations in the bank’s securities services business.

Gresham Technologies is expanding its use of robotic process automation in reconciliations it had not targeted previously, such as in cash and nostro accounts.