In recent decades, leading financial institutions have poured considerable resources into transforming their asset servicing operations. Their focus has spanned everything from technological modernization and bolstering straight through processing (STP) rates to tightening risk and control frameworks.
While legacy systems have managed to drive down transaction costs, the indispensable need for expert, knowledge-driven operations teams remain—and is now becoming more pronounced than ever. The surge in transaction volumes has only amplified the demand for seasoned professionals and heightened operational risk.
Enter artificial intelligence: Over the past few years, AI has ushered in a transformative shift in the outlook for corporate actions operations. Purpose-built models now open new horizons, offering fresh opportunities to reimagine efficiency. A spectrum of innovative use cases leveraging AI is emerging, each with the potential to reshape the landscape of asset servicing.
Types of use cases leveraging AI.
Identification of AI tools and data dependencies.
There are lots of different AI tools and algorithms out there that can tackle specific challenges, but it really comes down to picking the right one for your organization. You want something that can handle all the behind-the scenes data work, make sense of it, crunch the numbers, and suggest the best approach for whatever business problem you’re facing. While normal ML algorithms can be used for typical clustering or classification problems, the interpretation of free text has become very convenient with the usage of LLM/SLM’s. Companies are investing in the development of AI agents designed to operate independently in the background and address specific problems.
Further to running the ML models, the data residing in asset servicing applications need to be cleaned/ denormalized and provided as an input to the AI tool. ‘Training data’ sets play a key role in defining the prediction accuracy of a model. While synthetic data can be created and used for training, there is always a preference to train the models on production data.
Bringing together a smart AI tool that can handle data, host different AI agents with all sorts of techniques—like ML algorithms and LLMs—and connect seamlessly with your asset servicing application makes adopting AI in operations much simpler and more effective.
User experience plays a significant role in the development of any application feature in the context of operations users being exposed to multiple kinds of apps in their daily life. Any AI output needs to be plugged into the core application in such a way that the user experience is not inhibited in any way. Suggestions should be displayed alongside any business exceptions on the relevant page reviewed by the user. Likewise, AI-related workflows should be integrated into the core application processes. Also, in AI use cases, which use the context of the page from where the request is invoked, it is important that the output is delivered to the same page within the business process flow.
Trust in AI output and feedback loop.
ML algorithms rely on historical data for training and testing purposes and provide a confidence score associated with the output. While a high degree of confidence denotes higher accuracy of the output, there are outliers which sometime make the results incorrect. Thus, it is important that the financial services institutions identify the importance of the business transaction and enable review controls on the AI output. Low-importance transactions may be updated via straight-through processing (STP) using AI-generated outputs. For other transaction types, initial user acceptance is required. Additionally, it is essential that feedback on the acceptance or rejection of AI suggestions be provided to the AI system for ongoing retraining and algorithm optimization.
Outlook of asset servicing systems leveraging AI.
Financial institutions increasingly use AI to optimize operations, allowing teams to focus on key tasks instead of routine work and ensuring resources go where they matter most.
Use cases around providing suggestions in the operations workflow (based on historical actions taken) as well as allowing the user to query for answers (Q&A) will become a mandatory feature in all transaction processing applications. Concepts around self-optimization/self-healing are gaining traction and could disrupt the way in which business flows are configured.
As AI becomes an integral part of asset servicing, the evolution of standard operational procedures is not just inevitable, it’s essential.