ai

Perspectives

Artificial intelligence in capital markets  

Leveraging AI to optimize post-trade processes  

Evolving capital markets landscape presents an array of challenges in post-trade processing - rising costs and margin pressures, higher risk and compliance requirements and increased regulatory interventions. Learn how adopting cutting-edge AI capabilities with a structured approach can help capital markets firms address these challenges and gain a competitive advantage.

Artificial intelligence in post-trade processing

Post-trade processing has seen frequent changes across asset classes and the regulatory landscape. These changes, coupled with human-driven decisions across processes, introduce inefficiencies and lead to errors. Artificial intelligence can play an important role in reducing post-trade processing inefficiencies by easing decision-making through automated self-learning. AI capabilities, on effective application can significantly reduce the requirement for manual interventions, reduce reconciliation requirements, support straight-through processing, and enhance operations considerably.

Learn more about artificial intelligence in post-trade processing

This section is an infogram

This message and the space it occupies will not be displayed when viewing this page either in Live, Preview, or "View as published" modes

Determine the right solutions with an AI application framework

A clear answer to “how will we implement AI?” can help organizations navigate the AI journey efficiently. They need to have a structured approach towards identifying and prioritizing use cases for AI implementation. The following approach can be a starting point for organizations in their journey to implement AI within post-trade processing.

This section is an infogram

This message and the space it occupies will not be displayed when viewing this page either in Live, Preview, or "View as published" modes

Assess AI affinity

As a first step toward implementation, organizations need to apply a filter across their post-trade processes and understand if there could be a potential AI solution to their challenges. We have identified a few factors and illustrative measures that can guide organizations in assessing their processes for AI fitment.

Error Rate Decision fuzziness Process stability Unstructured or semistructured data Anomaly detection Scenario planning
working-on-ipad

Rationale

• Processes with a high degree of errors increase the organization’s operational and regulatory risk.
• AI models, through supervised and unsupervised learning, can be used to predict future error occurrences via anomaly detection.
• AI algorithms can also help in automated root-cause analysis based on historical and real-time data.


Post-trade example

• Trade mismatch
• Settlement failure
• Allocation error

woman-working-on-laptop

Rationale

• Increased human intervention for decision-making leads to an increase in operational and process inefficiencies.
• Human intervention can help in visual object recognition, speech recognition and production, basic natural language prediction, and comprehension.
• Through analysis, AI can help by identifying underlying patterns and building rule sets to undertake these fuzzy decisions.


Post-trade example

• Regulatory violations of trades

city-skyline

Rationale

• Process change involves changes to underlying systems or workflow. It is difficult for traditional rules-based systems to be modified based on these changes.
• AI can uncover key risk factors of change, predict which changes are about to fail, prevent change failure, and monitor for new threats.


Post-trade example

• Portfolio reconciliation with counterparties for newer asset classes in OTC trades

looking-up-at-skysraper

Rationale

• AI algorithms can help in processing unstructured and semistructured data, which cannot occur through standard statistical analysis techniques.
• AI algorithms help in establishing link between structured data and unstructured data.


Post-trade example

• End-customer issues reported via emails and social media related to margin calls, etc.

graph-data-on-screen

Rationale

• Anomalies (i.e., deviation from expected behavior) require analysis across time series or large volumes of data. Traditionally, analysts and risk managers have analyzed this data manually.
• AI algorithms can help analyze large volumes of data generated across a wide time duration to establish trends, recognize patterns, and highlight deviations, if any.


Post-trade example

• Proactive anomaly detection in standing orders
• Expired and/or invalid standing settlement instructions

people-collaborating-in-office

Rationale

• Scenario planning requires analysis of multiple structured and unstructured variables, determining relationships between them and planning for future outcomes that are in the organization’s best interest.
• AI models can help analyze both structured and unstructured data, identify causation and correlation between hundreds of these variables, and generate complex outcomes that are either likely or unlikely.


Post-trade example

• Collateral optimization
• Liquidity and cash management

Analyze business impact and technology feasibility

Business impact analysis

AI can help capital market firms generate insights to drive efficiencies, automate risk management and compliance across post-trade processes, and create value. We believe that firms can analyze the business impact of the AI use case across the following domains:

Enhance insights and productivity:
Analyze and derive insights from a large amount of data in various formats across sources to increase productivity and efficiencies (e.g., identify trends that regularly lead to trade failure).
 

Increase compliance to regulations:
Safeguard assets to help stay compliant, saving hefty penalties traditionally paid for being noncompliant with regulations across geographies (e.g., identify potential fraudulent transactions).

Reduce exposure to risk:
Reduce exposure to risk by enabling an accelerated response across multiple capabilities like counterparty credit checks to cybersecurity (e.g., margin calls from counterparties based on news alerts).
 

Reduce costs and losses:
Reduce losses and costs across the trading life cycle, from reconciliation to settlement and reporting (e.g., reduction in cost of capital through collateral optimization).


Technology feasibility analysis

To harness the true potential of AI, financial institutions will need to make significant changes and ready themselves for the implementation of AI solutions. Some of the factors that organizations need to consider while assessing technology feasibility include:

Data pipeline:
Access to production-grade data, ability to curate and collate structured and unstructured data, and availability of underlying technology required for data processing are important considerations.

Frameworks and infrastructure:
Analyze the availability of a framework’s license or skill set to develop algorithms using multiple open-source or proprietary frameworks, along with the availability of infrastructure to support these frameworks to assess feasibility.

Infrastructure elasticity:
Availability of on-demand processing power in a cost-effective manner through scalable infrastructure needs to be analyzed to understand feasibility.

Data sufficiency:
Use cases should be assessed for sufficiency of data such that AI analysis can be effectively undertaken.

A guide for AI implementation

  • Expand all
  • Collapse All

Once organizations have identified use cases with high business impact and high technology feasibility, it’s time for implementation. While enterprise AI adoption is accelerating at a significant growth rate, many organizations still find implementation and deployment of this technology to be daunting. Potential challenges include:

Business case definition: Defining clear problems to make effective use of AI solutions is challenging and requires taking into consideration existing and future processes, products, and services.

Diverse data requirements: AI solution efficiency not only depends on the volume of data that flows, but also the quality, which is difficult to deliver, as data is usually spread across multiple systems, sources, and varied formats.

Privacy and security concerns: Protecting access to sensitive information from breach or theft, as AI models require vast amounts of data to be able to make efficient decisions.

Integration complexity: Ability to integrate AI solutions with legacy systems, existing applications, business, and infrastructure.

Bias decision-making: Possibility of potential bias decisions made by AI models driven by the flawed data sets that flow in the model.

Legal challenges: An inaccurate or faulty algorithm coupled with inefficient or limited data can lead to incorrect decisions, exposing the enterprise to legal risks.

Organizations need to clearly articulate the problems, define benefits, identify right data, develop requisite skill sets, and overcome functional silos to enable AI implementation. Outlined below are a few considerations that can help organizations roll out AI within their post-trade landscape.

Assemble a core team: Identify a lean team with proven abilities to get things done in the enterprise and accelerate goals.

Conduct AI workshops: Introduce AI to the organization by conducting an AI familiarization workshop with key stakeholders.

Build technology architecture: Identify and close capability gaps, and develop supporting technology architecture and infrastructure (cloud) based on processing power required.

Identify data: Gather the right data spread across silos of business groups and legacy systems and eliminate inconsistencies to provide accurate data sets.

Incremental deployment: Start small, prioritize desired tactics, and identify select use cases to prove efficiency and value, gather feedback, and expand gradually.

Prove it works quickly and scale up: Use an agile, iterative piloting approach to move from strategy to prototyping as quickly as possible. “Fail-learn fast” and achieve rapid results.

Organizations should outline a detailed plan for scaling up AI solutions by ensuring enough collaboration between technology architects and business process leads. Steps for the journey include:

The post-trade processing space is inherently complex, with varied processes across legacy systems and proliferation of data coupled with constantly changing regulatory frameworks. A structured approach to leverage AI can help organizations optimize these processes at scale with increased efficiencies.

Did you find this useful?