Clearstream and Intelli-Select target Q1 23 for collateral management automation release
25 October 2022
Clearstream’s Marton Szigeti and Intelli-Select’s Bart Coppens speak to Bob Currie in advance of the launch of OSCAR, their AI-based collateral eligibility negotiation and screening tool
Image: stock.adobe.com/jamesteohart
Clearstream and Intelli-Select, a Belgium-based fintech, will deliver enhancements to collateral eligibility negotiation and collateral screening with the launch of a new solution that is targeted for release early in the new year.
This collateral tool, named Own Selection Criteria with Automated Reasoning (OSCAR), utilises Intelli-Select’s software, applying artificial intelligence strategies to evaluate collateral eligibility profiles when working with collateral baskets.
This will accelerate the time required for users to develop and negotiate individual collateral baskets through use of AI techniques, including knowledge representation and reasoning (KRR), machine learning and structured natural language processing (NLP). The AI methodology has been developed by Intelli-Select in collaboration with its academic partners at KU Leuven.
Clearstream, the Deutsche Börse-owned post-trade entity, indicates that OSCAR will be the first collateral management tool available in the market that applies AI techniques in this way to create, negotiate and execute baskets of collateral.
As reported in Securities Finance Times in June, OSCAR is designed to streamline these activities through use of AI, enabling users to define their own intelligent eligibility criteria while screening through an automated process for inconsistencies. Clearstream reports that the tool will also facilitate interoperability with different counterparts, assisting negotiation, reconciliation and collaboration.
Marton Szigeti, Clearstream’s head of banking, funding and financing, notes that one of the fundamental workflow challenges in collateral management is to set in place collateral schedules that a firm can exchange with its counterparties. Historically, this has been a heavily paper-based exercise that has been relatively inefficient operationally and can involve significant operational and legal cost.
To create and manage collateral baskets, users have traditionally needed to compare and translate different collateral profiles manually across multiple triparty agents, commonly utilising data held in spreadsheets and flat files. These collateral profiles need to be regularly updated in accordance with evolving business requirements and regulatory changes that prompt users to refine their use of collateral.
Clearstream has worked with Intelli-Select to create OSCAR as a means to streamline this process for the market. This creates a digital version of a collateral schedule which can be agreed with trade counterparties through a smart online interface. This enables users to create collateral schedules simply — through typing information into the online portal that is then read and digitised by OSCAR using natural language processing — enabling the user to match these collateral eligibility criteria electronically with counterparties and to trade within a matter of minutes, when this process would previously have taken several hours or days.
“This solution extends flexibility to build and amend collateral schedules quickly and simply,” says Szigeti. “If a user wishes to switch an asset out of a series of collateral baskets, this can be executed using OSCAR without pages of physical documentation and without an elongated legal approvals process.”
The testing process is now at an advanced stage and OSCAR is likely to be released fully to users of Clearstream’s collateral services in the early months of 2023. This will be implemented through a phased roll out, with basket set up and negotiation to be made available through an initial release in Q1 2023, followed by a second release phase in Q2 2023 that will add eligibility checking prior to further upgrades to the service.
Intelli-Select co-founder and CEO Bart Coppens indicates that OSCAR is built on the latest research in AI techniques, enabling the creation of sophisticated insights with full explainability of the reasoning process, yet based on intuitive natural language inputs. “The partnership with Clearstream gives us the opportunity to put these advanced techniques to work, solving real-world issues, in this case on collateral — but this is just the start as we have identified and are working on many more [potential applications of this model],” says Coppens.
Knowledge representation
In utilising Intelli-Select solutions, OSCAR applies a number of AI techniques including knowledge representation and reasoning and NLP. In its simplest terms, KRR is a field of AI dedicated to representing information about the world in a form that a computer system can apply to solve complex tasks such as diagnosing a medical condition or communicating a dialogue in a natural language. By applying KRR, an intelligent machine (or ‘intelligent agent’) is able to learn from its knowledge and experiences and to take intelligent actions that build on its existing pool of knowledge, such that there is learning and improvement at each stage in the KRR model’s evolution.
This is supported in the OSCAR product by natural language processing, which enables users to enter and amend their collateral schedules using text-based instructions, typically typed into Clearstream’s online client portal. Through this facility, the solution will extract the user’s instructions from typed English keywords and convert this into machine-readable structured language which aligns with the underlying application data model, yet the text remains close enough to the English language to be used in actual legal agreements.
Providing an intuitive element, the solution will advise what haircut should be applied to a collateral asset, for example, in line with the user’s specified risk profile. This will also screen for inconsistencies and suggest alternatives in line with the user’s collateral preferences.
“In our experience, a collateral schedule delivered through an excel file, a flat file for example, could require many thousands of if-else statements to represent electronically using traditional computing logic,” explains Coppens.
Through applying KRR, the platform is able to identify the user’s collateral eligibility criteria and exclusions through applying automated reasoning, dramatically improving the speed and efficiency of creating and applying collateral schedules. By indicating its collateral preferences through keywords entered into the user interface, OSCAR will suggest eligible assets from the user’s inventory that may be used to build and negotiate collateral baskets.
To make the model scalable, Intelli-Select has applied NLP and natural language generation (NLG) to accommodate natural language input and output, enabling the solution to interface with multiple collateral management service providers that may each employ a distinct data model and apply slightly different formats for data communication and reporting.
A parallel strength of this platform is that it applies explainable AI, providing a full audit trail and chain of explanation regarding why a collateral basket will be eligible or ineligible for a specified counterparty. “The explainability element is important,” notes Coppens. “OSCAR will document the reasoning behind each decision, explaining each step of the KRR reasoning process even when it took many steps to get there.”
Intelli-Select presented initial ideas to Clearstream around two years ago and, as the thinking matured, the OSCAR project rolled into action, advancing through a proof of concept, the creation of a minimum viable product and through extensive testing with its early adopter community to bring the programme to its current state of evolution.
In doing so, Intelli-Select has worked closely with data science and AI specialists at KU Leuven, which has supported a research group specialising in artificial intelligence for more than four decades and has teams with expert knowledge in KRR applications.
From this point, the project stakeholders expect to roll out the project relatively quickly. “Many users are already using Clearstream for settlement and custody,” explains Coppens. “The platform is highly intuitive, dictating that there is minimal cultural adaptation required on the users’ side to integrate OSCAR into its workflow and working procedures. And this is already supported by user documentation, explanatory video and FAQs.”
Clearstream’s Szigeti reinforces this point. “Our intention is to move our clients promptly onto the OSCAR platform, recognising the value this can deliver in providing faster, cheaper and more efficient collateral management,” he says. “For clients wishing to trade with a new counterparty over a triparty platform, and to set in place new collateral schedules, it is important to be able to do this on an intraday basis. They do not wish to be waiting days, or even weeks, to get this done. Until now, it has not been possible to move this quickly in setting up with new counterparties, and establishing new collateral schedules, in a legacy environment.”
While the partnership with Clearstream has focused on collateral selection, Intelli-Select has developed a KRR-driven model that could be applied to a much wider set of potential use cases. One example is the company's Fund Analyser solution, which screens the holdings that an asset manager has within a fund product to validate whether this aligns with an investor’s sustainable investment objectives and environmental, social and governance (ESG) criteria. Through its intelligent capability, this tool offers the ability to suggest fund products that align most effectively with an investor’s specified ESG profile.
This collateral tool, named Own Selection Criteria with Automated Reasoning (OSCAR), utilises Intelli-Select’s software, applying artificial intelligence strategies to evaluate collateral eligibility profiles when working with collateral baskets.
This will accelerate the time required for users to develop and negotiate individual collateral baskets through use of AI techniques, including knowledge representation and reasoning (KRR), machine learning and structured natural language processing (NLP). The AI methodology has been developed by Intelli-Select in collaboration with its academic partners at KU Leuven.
Clearstream, the Deutsche Börse-owned post-trade entity, indicates that OSCAR will be the first collateral management tool available in the market that applies AI techniques in this way to create, negotiate and execute baskets of collateral.
As reported in Securities Finance Times in June, OSCAR is designed to streamline these activities through use of AI, enabling users to define their own intelligent eligibility criteria while screening through an automated process for inconsistencies. Clearstream reports that the tool will also facilitate interoperability with different counterparts, assisting negotiation, reconciliation and collaboration.
Marton Szigeti, Clearstream’s head of banking, funding and financing, notes that one of the fundamental workflow challenges in collateral management is to set in place collateral schedules that a firm can exchange with its counterparties. Historically, this has been a heavily paper-based exercise that has been relatively inefficient operationally and can involve significant operational and legal cost.
To create and manage collateral baskets, users have traditionally needed to compare and translate different collateral profiles manually across multiple triparty agents, commonly utilising data held in spreadsheets and flat files. These collateral profiles need to be regularly updated in accordance with evolving business requirements and regulatory changes that prompt users to refine their use of collateral.
Clearstream has worked with Intelli-Select to create OSCAR as a means to streamline this process for the market. This creates a digital version of a collateral schedule which can be agreed with trade counterparties through a smart online interface. This enables users to create collateral schedules simply — through typing information into the online portal that is then read and digitised by OSCAR using natural language processing — enabling the user to match these collateral eligibility criteria electronically with counterparties and to trade within a matter of minutes, when this process would previously have taken several hours or days.
“This solution extends flexibility to build and amend collateral schedules quickly and simply,” says Szigeti. “If a user wishes to switch an asset out of a series of collateral baskets, this can be executed using OSCAR without pages of physical documentation and without an elongated legal approvals process.”
The testing process is now at an advanced stage and OSCAR is likely to be released fully to users of Clearstream’s collateral services in the early months of 2023. This will be implemented through a phased roll out, with basket set up and negotiation to be made available through an initial release in Q1 2023, followed by a second release phase in Q2 2023 that will add eligibility checking prior to further upgrades to the service.
Intelli-Select co-founder and CEO Bart Coppens indicates that OSCAR is built on the latest research in AI techniques, enabling the creation of sophisticated insights with full explainability of the reasoning process, yet based on intuitive natural language inputs. “The partnership with Clearstream gives us the opportunity to put these advanced techniques to work, solving real-world issues, in this case on collateral — but this is just the start as we have identified and are working on many more [potential applications of this model],” says Coppens.
Knowledge representation
In utilising Intelli-Select solutions, OSCAR applies a number of AI techniques including knowledge representation and reasoning and NLP. In its simplest terms, KRR is a field of AI dedicated to representing information about the world in a form that a computer system can apply to solve complex tasks such as diagnosing a medical condition or communicating a dialogue in a natural language. By applying KRR, an intelligent machine (or ‘intelligent agent’) is able to learn from its knowledge and experiences and to take intelligent actions that build on its existing pool of knowledge, such that there is learning and improvement at each stage in the KRR model’s evolution.
This is supported in the OSCAR product by natural language processing, which enables users to enter and amend their collateral schedules using text-based instructions, typically typed into Clearstream’s online client portal. Through this facility, the solution will extract the user’s instructions from typed English keywords and convert this into machine-readable structured language which aligns with the underlying application data model, yet the text remains close enough to the English language to be used in actual legal agreements.
Providing an intuitive element, the solution will advise what haircut should be applied to a collateral asset, for example, in line with the user’s specified risk profile. This will also screen for inconsistencies and suggest alternatives in line with the user’s collateral preferences.
“In our experience, a collateral schedule delivered through an excel file, a flat file for example, could require many thousands of if-else statements to represent electronically using traditional computing logic,” explains Coppens.
Through applying KRR, the platform is able to identify the user’s collateral eligibility criteria and exclusions through applying automated reasoning, dramatically improving the speed and efficiency of creating and applying collateral schedules. By indicating its collateral preferences through keywords entered into the user interface, OSCAR will suggest eligible assets from the user’s inventory that may be used to build and negotiate collateral baskets.
To make the model scalable, Intelli-Select has applied NLP and natural language generation (NLG) to accommodate natural language input and output, enabling the solution to interface with multiple collateral management service providers that may each employ a distinct data model and apply slightly different formats for data communication and reporting.
A parallel strength of this platform is that it applies explainable AI, providing a full audit trail and chain of explanation regarding why a collateral basket will be eligible or ineligible for a specified counterparty. “The explainability element is important,” notes Coppens. “OSCAR will document the reasoning behind each decision, explaining each step of the KRR reasoning process even when it took many steps to get there.”
Intelli-Select presented initial ideas to Clearstream around two years ago and, as the thinking matured, the OSCAR project rolled into action, advancing through a proof of concept, the creation of a minimum viable product and through extensive testing with its early adopter community to bring the programme to its current state of evolution.
In doing so, Intelli-Select has worked closely with data science and AI specialists at KU Leuven, which has supported a research group specialising in artificial intelligence for more than four decades and has teams with expert knowledge in KRR applications.
From this point, the project stakeholders expect to roll out the project relatively quickly. “Many users are already using Clearstream for settlement and custody,” explains Coppens. “The platform is highly intuitive, dictating that there is minimal cultural adaptation required on the users’ side to integrate OSCAR into its workflow and working procedures. And this is already supported by user documentation, explanatory video and FAQs.”
Clearstream’s Szigeti reinforces this point. “Our intention is to move our clients promptly onto the OSCAR platform, recognising the value this can deliver in providing faster, cheaper and more efficient collateral management,” he says. “For clients wishing to trade with a new counterparty over a triparty platform, and to set in place new collateral schedules, it is important to be able to do this on an intraday basis. They do not wish to be waiting days, or even weeks, to get this done. Until now, it has not been possible to move this quickly in setting up with new counterparties, and establishing new collateral schedules, in a legacy environment.”
While the partnership with Clearstream has focused on collateral selection, Intelli-Select has developed a KRR-driven model that could be applied to a much wider set of potential use cases. One example is the company's Fund Analyser solution, which screens the holdings that an asset manager has within a fund product to validate whether this aligns with an investor’s sustainable investment objectives and environmental, social and governance (ESG) criteria. Through its intelligent capability, this tool offers the ability to suggest fund products that align most effectively with an investor’s specified ESG profile.
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 Securities 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 Securities Finance Times
