Can we chat?
08 July 2025
Paul Chiappetta, managing director, Digital Innovation, Capital Markets, and Christian Rasmussen, senior advisor at Broadridge, take a look at how generative AI can facilitate better treasury management
Image: stock.adobe.com/Nelson
Treasury group: 鈥淲hat is the current client concentration of all third-party SFT liabilities? I need this broken down from highest to lowest.鈥�
Repo trader: 鈥淚 need to get back to you.鈥�
Is that a familiar scenario? For the treasury group at firms that are active in the repo market, this kind of exchange is all too common 鈥� and often frustrating.
On the other side, repo traders are frequently interrupted by important yet poorly timed questions, which can be especially disruptive during periods of market volatility.
Now balance the repo desk鈥檚 need to optimise trade execution against the treasury group鈥檚 need for fast and constant feedback on leverage, liquidity, and funding ratios. That information directly impacts the firm鈥檚 balance sheet and helps it to manage risk. And the market volatility that is keeping repo traders busy may have triggered an early warning indicator, raising questions about market liquidity that the firm鈥檚 treasury group need immediate answers to.
Opportunities to drive efficiency with generative and agentic AI
In the future, artificial intelligence (AI) will offer an alternative to these inefficient interactions by facilitating data access.
AI is not the capital market鈥檚 first technology disruptor. Innovations such as trading algorithms, straight-through processing (STP), and distributed ledger technology (DLT) have transformed global markets. Technology has made it possible for new financial products 鈥� such as cryptocurrency 鈥� to be developed.
Today, capital market participants are adopting new AI techniques such as reinforcement learning, deep learning, and natural language processing. These techniques help leverage insights for forecasting, trading and other operational efficiency and revenue-generating opportunities. For example, AI algorithms allow market participants to process unstructured data, which can help improve their trading strategies and risk management efforts. AI is freeing individuals from mundane tasks, such as responding to routine, often time-sensitive, questions.
Also emerging are agentic AI architectures, built around specialised, autonomous agents that can self-direct, collaborate, coordinate tasks, and independently execute workflows, transforming manual business processes into automated, intelligent operations.
Like many financial markets, repo 鈥� primarily an over-the-counter (OTC) market 鈥� has聽taken a cautious approach to embracing digital transformation and integrating AI into its operations.聽This could be indicative聽of firms' reluctance to put their scarce resources to work until they see greater market acceptance via the existing infrastructure.聽
Regulatory catalysts: T+1 and mandatory clearing
The repo market is in the process of moving away from voice trading and manual processing.
Two relatively recent regulatory changes, T+1 and mandated central counterparty clearing, have motivated market participants to make significant investment in technology for their repo trading and post-trade settlement systems.
In May 2024, a new T+1 settlement cycle deadline went into effect. Broker-dealers now have the trade day plus one additional day to settle transactions rather than the trade day plus two additional days. With this tighter settlement deadline, firms had no choice but to begin using STP in their repo transactions.
An upcoming change that will test the repo market鈥檚 technology infrastructure is the move to mandated central counterparty clearing as of 30 June 2027. Firms will need to ensure their technology infrastructure is capable of scaling and integrating with new counterparties and central clearing agencies. With sponsored repo trade volume anticipated to grow, even dealers currently clearing trades through the Fixed Income Clearing Corporation (FICC) will need to review 鈥� and possibly upgrade 鈥� their existing tech stack.
As of March 2025, the FICC is already clearing 46 per cent of covered Treasury repo. That compares to a reported 30 per cent in 2024.
Building a better chatbot
Previously, chatbots and digital assistants relied on conversational AI. However, the 2022 launch of Open AI's聽chat generative pre-trained transformer (ChatGPT) changed how the world looks at 鈥� and uses 鈥� chatbots.
ChatGPT uses generative, rather than conversational, AI. This is an important distinction as generative AI is better equipped to process complex data. And unlike conversational AI, it can create new content. Generative AI can also be trained to automate repetitive tasks. The result? Chatbots built with generative AI can perform research, generate reports and retrieve and organise data.
With its more sophisticated capabilities, generative AI is well suited to create tools that increase operational efficiencies for financial services firms鈥� institutional businesses. An example is Broadridge鈥檚 generative AI application for the corporate bond market, BondGPT, first launched in the LTX e-trading platform in 2023. Broadridge鈥檚 OpsGPT is another example, which helps manage operations across the post-trade lifecycle. Recent agentic AI enhancements to OpsGPT include automation of settlement fails analysis, optimised inventory management, and streamlined email communication workflows.
Generative AI and Agentic AI for treasury and repo insights
When it comes to solving the inefficient dynamics between the repo desk and treasury group, generative AI could be used to create efficiencies and save time. Chatbots and dashboards that proactively update information needed by the treasury group to better understand various points of liquidity and liquidity trends could minimise the need for back and forth and facilitate more seamless workflows.
For example, a treasury group asks for information that includes backdated data. To calculate the answer, a trader may need to step away from the repo desk for several hours during the trading day. In contrast, AI can provide this information to the treasury group faster and without interrupting the trade flow.
In addition to generative AI use cases, agentic AI can be applied to benefit treasury management. For example, it can be used for limit monitoring and alerting: if a limit threshold is breached, an AI agent can automatically send a warning email to notify relevant stakeholders.
For repo traders, spending less time during the trading day providing the treasury group with real-time input is not the only benefit of generative AI. Repo traders rely heavily on proprietary networks and third-party data providers that offer varying degrees of accuracy and timeliness. These fragmented data sources and the time-consuming data collecting process make it challenging for traders to make informed decisions on risk and ratios. AI鈥檚 ability to gather and aggregate data accurately and quickly would enhance operational efficiencies and provide repo traders with decision-making support.
The timing鈥檚 right
New market developments and regulatory changes make this an ideal time for repo market participants to look for ways to integrate AI to solve daily operational challenges. We are excited about the potential that generative AI and agentic AI offer our repo market clients 鈥� traders and their treasury group colleagues 鈥� as well as those in the broader securities finance arena and look forward to helping them unlock its potential.
Repo trader: 鈥淚 need to get back to you.鈥�
Is that a familiar scenario? For the treasury group at firms that are active in the repo market, this kind of exchange is all too common 鈥� and often frustrating.
On the other side, repo traders are frequently interrupted by important yet poorly timed questions, which can be especially disruptive during periods of market volatility.
Now balance the repo desk鈥檚 need to optimise trade execution against the treasury group鈥檚 need for fast and constant feedback on leverage, liquidity, and funding ratios. That information directly impacts the firm鈥檚 balance sheet and helps it to manage risk. And the market volatility that is keeping repo traders busy may have triggered an early warning indicator, raising questions about market liquidity that the firm鈥檚 treasury group need immediate answers to.
Opportunities to drive efficiency with generative and agentic AI
In the future, artificial intelligence (AI) will offer an alternative to these inefficient interactions by facilitating data access.
AI is not the capital market鈥檚 first technology disruptor. Innovations such as trading algorithms, straight-through processing (STP), and distributed ledger technology (DLT) have transformed global markets. Technology has made it possible for new financial products 鈥� such as cryptocurrency 鈥� to be developed.
Today, capital market participants are adopting new AI techniques such as reinforcement learning, deep learning, and natural language processing. These techniques help leverage insights for forecasting, trading and other operational efficiency and revenue-generating opportunities. For example, AI algorithms allow market participants to process unstructured data, which can help improve their trading strategies and risk management efforts. AI is freeing individuals from mundane tasks, such as responding to routine, often time-sensitive, questions.
Also emerging are agentic AI architectures, built around specialised, autonomous agents that can self-direct, collaborate, coordinate tasks, and independently execute workflows, transforming manual business processes into automated, intelligent operations.
Like many financial markets, repo 鈥� primarily an over-the-counter (OTC) market 鈥� has聽taken a cautious approach to embracing digital transformation and integrating AI into its operations.聽This could be indicative聽of firms' reluctance to put their scarce resources to work until they see greater market acceptance via the existing infrastructure.聽
Regulatory catalysts: T+1 and mandatory clearing
The repo market is in the process of moving away from voice trading and manual processing.
Two relatively recent regulatory changes, T+1 and mandated central counterparty clearing, have motivated market participants to make significant investment in technology for their repo trading and post-trade settlement systems.
In May 2024, a new T+1 settlement cycle deadline went into effect. Broker-dealers now have the trade day plus one additional day to settle transactions rather than the trade day plus two additional days. With this tighter settlement deadline, firms had no choice but to begin using STP in their repo transactions.
An upcoming change that will test the repo market鈥檚 technology infrastructure is the move to mandated central counterparty clearing as of 30 June 2027. Firms will need to ensure their technology infrastructure is capable of scaling and integrating with new counterparties and central clearing agencies. With sponsored repo trade volume anticipated to grow, even dealers currently clearing trades through the Fixed Income Clearing Corporation (FICC) will need to review 鈥� and possibly upgrade 鈥� their existing tech stack.
As of March 2025, the FICC is already clearing 46 per cent of covered Treasury repo. That compares to a reported 30 per cent in 2024.
Building a better chatbot
Previously, chatbots and digital assistants relied on conversational AI. However, the 2022 launch of Open AI's聽chat generative pre-trained transformer (ChatGPT) changed how the world looks at 鈥� and uses 鈥� chatbots.
ChatGPT uses generative, rather than conversational, AI. This is an important distinction as generative AI is better equipped to process complex data. And unlike conversational AI, it can create new content. Generative AI can also be trained to automate repetitive tasks. The result? Chatbots built with generative AI can perform research, generate reports and retrieve and organise data.
With its more sophisticated capabilities, generative AI is well suited to create tools that increase operational efficiencies for financial services firms鈥� institutional businesses. An example is Broadridge鈥檚 generative AI application for the corporate bond market, BondGPT, first launched in the LTX e-trading platform in 2023. Broadridge鈥檚 OpsGPT is another example, which helps manage operations across the post-trade lifecycle. Recent agentic AI enhancements to OpsGPT include automation of settlement fails analysis, optimised inventory management, and streamlined email communication workflows.
Generative AI and Agentic AI for treasury and repo insights
When it comes to solving the inefficient dynamics between the repo desk and treasury group, generative AI could be used to create efficiencies and save time. Chatbots and dashboards that proactively update information needed by the treasury group to better understand various points of liquidity and liquidity trends could minimise the need for back and forth and facilitate more seamless workflows.
For example, a treasury group asks for information that includes backdated data. To calculate the answer, a trader may need to step away from the repo desk for several hours during the trading day. In contrast, AI can provide this information to the treasury group faster and without interrupting the trade flow.
In addition to generative AI use cases, agentic AI can be applied to benefit treasury management. For example, it can be used for limit monitoring and alerting: if a limit threshold is breached, an AI agent can automatically send a warning email to notify relevant stakeholders.
For repo traders, spending less time during the trading day providing the treasury group with real-time input is not the only benefit of generative AI. Repo traders rely heavily on proprietary networks and third-party data providers that offer varying degrees of accuracy and timeliness. These fragmented data sources and the time-consuming data collecting process make it challenging for traders to make informed decisions on risk and ratios. AI鈥檚 ability to gather and aggregate data accurately and quickly would enhance operational efficiencies and provide repo traders with decision-making support.
The timing鈥檚 right
New market developments and regulatory changes make this an ideal time for repo market participants to look for ways to integrate AI to solve daily operational challenges. We are excited about the potential that generative AI and agentic AI offer our repo market clients 鈥� traders and their treasury group colleagues 鈥� as well as those in the broader securities finance arena and look forward to helping them unlock its potential.
NO FEE, NO RISK
100% ON RETURNS If you invest in only one securities finance news source this year, make sure it is your free subscription to 兔子先生Finance Times
100% ON RETURNS If you invest in only one securities finance news source this year, make sure it is your free subscription to 兔子先生Finance Times
