DTCC defines its AI play

The Depository Trust and Clearing Corporation’s (DTCC’s) Johnna Powell outlines the firm’s artificial intelligence strategy, how new technologies might impact the various post-trade processes it supports and the best practices she believes firms should adopt when integrating AI governance into their existing risk management functions.

What does DTCC’s AI strategy entail?

Johnna Powell, managing director, head of technology, research and innovation, DTCC: DTCC’s AI strategy is not a stand-alone plan. It is designed to enable our corporate strategy at the highest level by delivering on modernization, maximizing the value for our stakeholders, advancing risk management excellence, building on our global network connectivity and supporting the retention and development of our talent.

To develop our AI strategy, we underwent a meticulous process to ensure it tied to those objectives with a focus on four key areas: enhancing the client experience, improving risk mitigation, increasing productivity—which is where we spend about 80% of our time—and advancing AI research and development across the firm. Near-term and long-term goals have been established as we advance our AI strategy.

In the near term, we are focusing on establishing AI as an enterprise capability by increasing collaboration and ownership across all departments within DTCC. Over the past two years, we’ve had a more centralized focus on AI and now we are expanding that approach—with a plan to ensure we place the proper tools in the right hands of the correct stakeholders. We’re developing an enterprise-wide AI sandbox that empowers teams through democratized AI skills and tools, while remaining focused on our commitment to responsible AI advancement. We have also rolled out overall foundational AI training programs and certification, as well as specific AI tool trainings.

In the long term, we aim to deliver AI transformation at scale. To achieve this, not only will generative AI (GenAI) and AI technology need to be involved, but we will also need to break down and revamp business processes. Ultimately, the goal is that AI and GenAI are used pervasively across DTCC, with multiple new products and services, new processes owned across departments and new, disruptive business models that provide substantial efficiency gains for DTCC and our clients.

There’s a delicate balance between pushing innovation and managing systemic risk—something a lot of firms are struggling with. How do you find the right balance?

Johnna Powell: One of the key pillars of DTCC’s AI strategy reflects the need to effectively balance innovation with improvements in risk mitigation. AI is a new technology and there are risks associated with implementing it incorrectly. As a result, we’re approaching innovation by clearly defining use-cases that provide measurable business value. At the same time, we developed a framework that establishes transparent governance and control on how DTCC onboards new technology.

From the beginning of our GenAI journey, we worked closely with our risk and security teams to develop and implement an AI policy that ensures effective governance and operating models across the organization. Next, we established the AI Council, including representation from all departments across DTCC. The goal of the Council is to provide overall governance and appropriate monitoring of the firm’s AI use-cases and technology to ensure they align with our AI policy and do not introduce risk.

Additionally, we established an AI Enablement Team that unofficially reports to the AI Council and mirrors similar DTCC representation. Its role is to initially analyze the significant pipeline of proposed AI use-cases to ensure compliance with the AI policy—including an evaluation of potential risks—such as security or technology risks. It also works to eliminate duplicate initiatives, confirm that the use-cases provide measurable business value and present high-priority use-cases to the AI Council for approval. In addition to this, DTCC continues to engage with regulators and industry bodies.

Which areas of post-trade market infrastructure do you believe can benefit most from AI adoption?

Johnna Powell: Aside from the obvious meta-benefits of improved productivity and efficiency that most co-pilots provide—such as finding and summarizing documents effortlessly—settlement and clearing optimization is also a high-priority area in which AI could offer significant benefits. For example, you could leverage AI predictive analytics to enhance risk mitigation by essentially predicting more accurately and quickly whether a trade might fail. Additionally, there is an opportunity to further streamline corporate actions processes. AI data ingestion models could be leveraged, for example, to ensure a single source of truth.

At the same time, data management and insights—such as deployed data reconciliation—will likely see similar value from the use of AI and GenAI due to current manually intensive processes. Many institutions have challenges with unstructured, mislabelled or poorly organized data. AI can address these challenges, specifically in metadata tagging and data structuring.

Also, the convergence of distributed-ledger technology and AI is an exciting area, as the combination could deliver several benefits for the industry. One important use-case is collateral management, where machine learning models could optimize collateral allocation and utilization across various parties to address margin requirements, reduce systemic exposure and lower risk. In fact, market participant firms are already exploring ways to enhance their collateral management functions and are unlocking trapped liquidity with AI-powered, automated decision-making models to assist with effective liquidity management.

What best practices allow the creation of effective oversight structures, and how can AI governance fit into existing risk management functions?

Johnna Powell: DTCC is already following several frameworks to ensure we are aligned with best practices and recommendations. The National Institute of Standards and Technology (NIST) AI Risk Management Framework, for example, provides guidelines around establishing structured oversight, leveraging four core functions: govern, map, measure and manage. It provides guidance around defining clear AI roles and responsibilities within senior management and across teams, establishing transparent accountability. We’ve introduced the mapping and measuring functions to effectively identify and quantify risks specific to each AI use-case or initiative, including guidelines around explainability, data quality, bias and more.

We’ve also established a Responsible AI Working Group to ensure we have a balanced and responsible approach to AI deployment, including, for example, comprehensive end-to-end testing of the models we’re deploying into our environments, with appropriate reporting capabilities.

It is important to note that while the four core tenets of the NIST AI framework are strong and effective, firms are not required to prove adherence to them—meaning they can self-select where they want to focus. In addition to the NIST framework, the Fintech Open Source Foundation, known as FINOS, recently published an AI governance framework that DTCC was involved in developing, which provides even more granular details in these areas.

A key to all of this, though, is being open-minded to implementing new methodologies within existing frameworks. This is an essential part of this journey. One of the ways DTCC has incorporated these frameworks into our existing processes is by ensuring our technology risk management review processes account for AI technologies. If there’s an AI technology we want to onboard, we leverage an established framework when assessing the vendor and its technology to ensure it meets all DTCC standards. Similarly, our model risk management team has employed a framework to evaluate all AI models, ensuring they have minimal drift and underperformance.

Looking six to 12 months ahead, what are the biggest AI opportunities DTCC sees for enhancing business processes across the industry? Where are the really big opportunities for AI adoption?

Johnna Powell: We’re seeing significant traction in improving effectiveness and efficiency across the entire software development life cycle. There is untapped potential for achieving increased efficiency from new agents, such as developer co-pilots that assist with auto-completing code, including code review agents, upgrade agents and code translation agents for Cobol or Assembler modernization. We’re deeply involved in this area by testing many different agents and seeing incredible improvements.

As a result of several co-pilot efforts, we are currently working to refine our overall software development approach. Coding exercises that previously took a developer 10 hours to complete now take only six, increasing productivity by 40%. At the same time, we are seeing improvements in the development life cycle as we introduce new agents into the workflow. We are seeing benefits in many areas, including those related to risk management and surveillance through real-time anomaly detection.

Finally, we anticipate GenAI to be deeply impactful in regulatory compliance. We are currently exploring tools capable of analyzing policy procedures and regulatory controls to assess existing controls, uncover potential compliance gaps and simultaneously identify new requirements.

These are just a few of the areas we have identified where AI can make a huge difference in terms of value, efficiencies, risk reduction, liquidity improvements and more. The potential of AI is vast and we’re excited about its potential to transform both our business and the industry as a whole.

Sponsored content