Agility Unveiled: Using Generative AI to Reimagine Capital Markets Operations

Authors: Gaurav Dave & Swaran Patnaik

FIRMS IN THE CAPITAL MARKETS ARE CONFRONTING NEW TROUBLES:

The combination of reduced settlement cycles increased trading velocity, and new working methods put significant stress on firms' legacy technology and operating models. The need for change is especially urgent as equities and fixed-income instruments move toward T+1 settlement in North America in May 2024 and the EU/UK following by 2025, requiring same-day affirmation. Differences in settlement cycles across markets and asset classes will further pressure manual operational processes across multiple trade lifecycle functions, including clearing, settlements, trade execution, margins, and fail processing.

The shortening of settlement cycles has led firms to evaluate legacy systems, reduce batch and manual processing, focus on straight-through processing (STP), and reduce fails. Emerging technology advancements in Generative AI and Large Language Models are transforming post-trade functions, driving productivity, increasing scalability, and enhancing risk management. Firms recognize the power of these technologies in resolving complex settlement failure questions and integrating them into operational workflows.

The application of Generative AI could significantly reduce the operational risk and cost burden of T+1, improving operational efficiency by up to 50% in certain areas. This would also mitigate risk, increase costs, and improve governance. Generative AI's natural language capabilities allow users to access data and insights, predicting the best actions and integrating them into workflows. This technology can unlock value for firms, not only meeting T+1 requirements but also promoting long-term transformation of their operational functions.

With a focus on responsible GenAI implementation via architectural design patterns, safety, and accuracy controls. The proposed strategy is built on compliance, secure interfaces, data privacy, and entitlement-based access. The strategy makes an intelligent utilization of LLM and operates at scale for the industry, letting them develop and transform with GenAI technology. The technique also includes building trust with GenAI, which is critical for effective adoption.

The primary challenges for Capital Markets firms are:

1. SEC mandates T+1 settlement for cash shares, corporate debt, and unit investment trusts in the US and Canada in May 2024, leading to shorter settlement cycles. Institutions will need to adapt their operations, which may increase operational burdens.

2. Achieving Continuously Accelerated Operational Productivity: Automating manual, time-consuming procedures, particularly in shorter settlement cycles, is crucial for improving operational efficiency and lowering costs. Manual intervention in trade and settlements is unsustainable and inefficient. Reduced settlement cycles may lead to increased operational costs unless handled through technology.

3. Retaining Talent: The new workforce requires user-friendly system interfaces that enable teams and talent to be fungible across regions and asset classes, among other criteria.

4. Accelerate product development and expansion. Lifecycle:

a. New Asset Classes: As capital markets grow, firms will explore new asset classes with distinct trading requirements. This increases operational difficulties.

b. Globalization presents problems such as cross-border legislation, currencies, and trade hours, complicating operations.

5. Data Accessibility: Accurate, timely, and trustworthy data is essential for decision-making and compliance. Accessing data from many systems might be challenging due to fragmentation.

THE OPPORTUNITY

Challenges provide opportunities. Implementing T+1 is projected to lower settlement risk by 50% and improve market efficiency. Investors will have faster access to cash, improving liquidity and accelerating market recovery. The transition to T+1 settlement will enable developments in market infrastructure. In 2024, market infrastructure will undergo more innovation and digitalization, including same-day and block-level clearing, tokenized atomic settlement, and bilateral settlement, among others. Executives will prioritize technologies like distributed ledger technology (DLT) and artificial intelligence (AI) to improve post-settlement efficiency.

To maximize efficiency, banks and broker-dealers should prioritize automation over manual intervention on a large scale. To prepare, firms should invest in long-term transformation, such as streamlining trade lifecycles, enabling global real-time processing, driving operational efficiency through automation, and consolidating key business capabilities across asset classes.

AI can improve efficiency and scalability in post-trade processes, allowing teams to do sophisticated research and get insight into important risk indicators. GenAI enables us to create user-friendly, chat-based interfaces that provide instant access to data and insights. It also allows for the fungibility of teams and talent across different locales and asset classes.

Settlements OpsGPT can offer a pilot program for operations users.

OpsGPT is a GenAI- and LLM-powered application that helps operations users, analysts, and management teams manage operations across the post-trade lifecycle, building on the successful design principles of BondGPT, the first GPT application.Although operational functions are still fragmented by asset class and market, the industry has made significant progress towards contemporary, unified processes in key functional areas.The Operations Console provides a unified interface for managing day-to-day tasks, accessing information and systems, and self-service through ad-hoc querying. Integrating these workflows with GenAI and OpsGPT improves productivity and governance in operations.

Take trade settlements as an example. The DTCC estimates that a 2% failure rate in trade settlements can cost corporations up to $3 billion globally.2 A trade failure not only incurs financial costs, such as CSDR penalties and carry charges, but also affects company liquidity, systemic risk, reputation, staff usage, and time.

Settlement OpsGPT can offer a user-friendly interface for accessing data and insights across many systems and workflows.

Settlement OpsGPT can offer faster and seamless data accessibility by training on curated and harmonized data from global multi-asset post trade systems. This streamlines access, connectivity, and understanding of data across multiple asset classes in third-party systems. GPT technology helps analysts collect information on trade failures and correct them in seconds. It increases the speed at which breaks are resolved, decreases the number of fails and breaks, and increases capacity for higher-level tasks across multiple asset classes, geographies, and systems.

OpsGPT's AI SME function can provide real-time "how to" assistance for operational issues, accelerating the identification, prioritization, and resolution of settlement failures. For instance, a user may query, "Where can I find my failed trades?", "How do I find my open positions?", or "Do I need special permissions?" The AI SME supports follow-up conversations, allowing users to ask additional questions about the same topic (e.g., "Tell me more about this functionality").

This is only one example. OpsGPT is a co-pilot that improves productivity for operations teams during shortened settlement cycles. It provides front-to-back trade lifecycle event transparency and empowers users to quickly mitigate and prevent risks over time. This is accomplished by allowing users to personalize their support using predictive analytics and actionable insights.

Innovating responsibly to meet industry needs.

OpsGPT should prioritize safe AI techniques and handle industry-specific concerns. The methodology should prioritize compliance, safe interfaces, data protection, entitlement-driven access, and a smart use of LLM.

1. PROTECTING CLIENT DATA: Settlement OpsGPT might be a comprehensive compliance platform that safeguards sensitive client data, personal information, and proprietary knowledge in the capital markets. It may interact with a variety of LLM providers, including OpenAI and cloud-hosted LLMs, providing flexibility in model selection depending on unique use case needs and functional and non-functional criteria. The platform can provide a consistent and scalable API interface, allowing for easy switching between models.The technology maintains regulatory compliance by securely preserving user inputs and LLM responses in a datastore for inspection and audit. The usage of a standard harmonized data language reduces implementation risk, accelerates client onboarding, and improves decision-making in failure resolution.

2. Preventing LLM illusions: Settlement OpsGPT can address the problem of LLM hallucinations, which can result in misleading or incorrect information in the financial sector. By relying on vetted and curated sources, OpsGPT can guarantee accurate responses. Instead of LLM-generated responses, the system collects exact data from reliable sources and displays it through the LLM chat interface. This strategy not only reduces hallucination hazards, but it also enhances response quality, bringing it in line with industry requirements and ensuring reliability.

3. Addressing Regulatory Obligations: OpsGPT, a financial services AI solution, is intended to fulfill changing regulations and industry-specific laws. It includes a compliance layer in the form of an AI agent that watches the application's output and is trained on current regulations. The application architecture is prepared to accommodate upcoming regulatory changes, enabling for compliance updates. This proactive approach enables financial institutions to confidently deploy OpsGPT, ensuring customer and institutional safety.

4. GARDENING The IP: In competitive capital markets, protecting intellectual property (IP) is critical. OpsGPT focuses protecting clients' intellectual property by obtaining information directly from them and applying protection and obfuscation. They use a privately hosted model to assure compliance and privacy, protecting clients' IP while also giving piece of mind and security.

Accelerating the Safe and Orderly Transformation of Post-Trade Operations

AI can speed post-trade operational transformation by enabling end-to-end trade lifecycle event transparency across disparate trade processing ecosystems. This enables end users to quickly mitigate, decrease, and prevent risks over time. AI solutions can assist professionals in researching and resolving failures rapidly, allowing them to identify operational efficiencies throughout their firm. This competitive edge leads to smarter decisions. A committed firm can supply safe and responsible AI solutions on a large scale, developing a pipeline of client-facing products across Capital Markets' post-trade and front office operations, as well as improving internal productivity and efficiency.