In this episode of Droit Talks About, host Hopeton Lindo is joined by Leith Dennis, Droit’s Head of Knowledge Engineering for Adept, to unpack the latest from Droit: Explore Mode. We’re diving deep into the technology behind Droit’s recently approved patent, designed to revolutionize how financial institutions interact with complex regulatory obligations.
Learn about the efficiencies of Explore Mode for pre-trade scenario testing and high-throughput post-trade data evaluation. This episode is a must-listen for anyone looking to navigate financial regulation with greater clarity and unlock new opportunities
Listen to the full episode below.
Full transcript below:
Hopeton: Hello and welcome to another episode of Droit Talks About, where we help make sense of regulatory change and what it actually means for your compliance processes and procedures. I’m your host, Hopeton Lindo, and today we’re doing something a little different. We’re turning the lens inward to share something new we’ve been working on here at Droit. Something we believe could make a meaningful difference for financial institutions navigating complex compliance workflows. Droit recently received patent approval for a core piece of our technology that changes how firms can interact with our regulatory logic and test how obligations shift under different scenarios. We’re calling this Explore Mode. And to help us unpack all of this and how it can help financial institutions with their regulatory obligations I’m joined by Leith Dennis, Droit’s Head of Knowledge Engineering for Adept. Leith welcome to Droit Talks About.
Leith: Yeah, thanks for having me. I’m excited to be here.
Hopeton: So before we begin, can we start with a quick intro and if you could let the listeners know what your team actually does at work that would be great.
Leith: Of course. Okay, so I lead Droit’s knowledge engineering team and I also oversee the product development of our core Adept platform. And in order to explain what our team does it probably makes sense that I first start with a bit of an explanation of the Adept platform itself. Adept is what allows our users to make decisions against formally defined rules. These are rules typically capturing either regulation or internal policy and it makes these decisions in a manner that is both repeatable, explainable, and auditable. So the platform is comprised of three key components: we have the Decision service. This is the actual callable service that allows users to provide input information and receive a regulatory decision as a response. We have the Logic Viewer, which is an application that allows users to see a human-readable diagram of the rules as well as traverse through that diagram the execution of a particular decision. And then we have the digital library. This captures a digitized version of regulatory source text and this allows us to link between the visualization of the rules within the logic viewer and the actual text that implies those rules.
Hopeton: Okay, so those three applications make up the Adept platform. So how does that tie in with the work that your team’s focused on?
Leith: Right. What our team actually does is break down regulations treating these natural language documents as data and we use that to create formal logic models that we actually load into our Adept platform and we go through this process in order to build these rules
of, first, starting with the text itself. Reading and analyzing that source regulation,
forming a clear and complete understanding of the rules inherent within it, defining an appropriate data model to capture those rules, constructing formal logic against that data model, and then annotating that logic back to the actual source text. Through this process we create these knowledge models that can be loaded in Adept. Once loaded, a user can provide a specific set of input facts and get returned an actual output conclusion. For example, the way a given rule might operate if we look at something like business conduct requirements. We may have a rule that takes as input certain static party data, product economic details, certain execution context details, and it may return as a conclusion, a “yes” or “no” answer as to whether that transaction is permissible under those rules.
Hopeton: Okay, so in other words it’s less about interpreting the rule and it’s more about computing it. So as you explained you start with this complex unstructured human readable legal text you your team then applies a structured model that translates it into machine readable logic and then from there you compute the outcome, and I guess by doing so you eliminate the subjectivity around manual interpretation, right?
Leith: That’s exactly the idea.
Hopeton: Okay, so all of that sounds like quite a job for your team. How do they even begin to untangle and structure something this intricate?
Leith: Yes, so it certainly is complex work and can be, and I’ll also acknowledge a lot of that complexity too comes from the sort of interconnected web of sources that you need to consider
for any particular rule. For a given rule we’re not just looking at the primary law or regulation that might define it, but we also need to take into account things like any related regulatory guidance, any exemptive relief issued by regulators, certain best practice interpretations from industry bodies, and even the interpretation of individual market participants legal and compliance teams. So all of this context needs to be factored in when we look at building out a rule and our team is a specialized set of individuals from various backgrounds, but we’re united in the fact that we can take a structured data lens of thinking and how we approach building out these rules. We leverage the models that we have and how we construct logic, and we apply that to the entirety of this context so through that exercise we can define complete consensus-driven versions of these rules.
Hopeton: Okay, alright so I wanna shift the conversation a little bit, not entirely away from the Adept platform, but to this ‘Explore Mode” that we’re talking about today. I actually have the patent filing here with me, and it’s entitled, and I’m gonna need your help with this Leith, “methods and systems for regulatory exploration preserving bandwidth and improving computing performance.” What does that mean?
Leith: Yeah, so certainly a bit of a mouthful, but we can try to break this down so as you’ve said we would refer to this a bit more concisely as Explore Mode. What this software functionality actually allows is for us to take all of the rules that we’ve built within our platform and expose to users entirely new ways of interacting with them.
Hopeton: Okay so what I’m hearing is that our rule set has become more dynamic. Would I be correct in that?
Leith: I think that’s fair, exactly, so Explore Mode as a functionality provides users a more efficient, and in a way more complete manner, of analyzing potential outcomes with respect to evaluating these complex rules. So previously if you wanted to know if a specific transaction was permissible, you feed into the system every single detail related to that specific transaction and you get back from the system a definitive answer, either “yes, you can execute this trade” or “no, do not trade.”
Hopeton: So with this Explore Mode, you can provide partial inputs and see, for example, what types of products or services could be offered or what variations are possible under your existing rules. Am I getting that right?
Leith: That’s exactly it, yeah. It is really a whole new way of using the platform
and even a new way of looking at compliance. Not just looking at “you can or can’t do this specific thing,” but flipping around to “what are you capable of doing?” or put differently, what is the possibility space available to you. Maybe a bit more simply, it allows you to move beyond a simple yes or no decision based on fully defined inputs and provides the capability to explore the possibilities available to you as constrained by a set of rules. So really it provides a more open-ended means of using the rules to interact with your data. I can move from asking have I sufficiently onboarded this particular counterparty to trade this specific interest rate swap I can instead ask which counterparties can I trade this swap with or what types of products could I trade with this counterparty so you can explore any possible combinations of what is available to you subject to the same rules. And importantly, this new mode of evaluation preserves the same level of precision and auditability that we have within Adept.
Hopeton: Okay so just thinking about the… let’s talk about the pre-trade side for a second.
What you’re saying is that teams can test scenarios on the fly and I guess immediately see how regulatory obligations change with different inputs.
Leith: That’s right, so for pre-trade, we’re going from just a binary yes/no response on a specific transaction, so for example no can’t trade that yep and we’re allowing you now to answer questions of the shape where you get an answer here’s all the trades that you can do so it’s definitely a shift and one that can potentially even open up new revenue opportunities for users allowing them to efficiently understand the space in which they’re permitted to operate and then on the post-trade side as well. I’ll mention there are tremendous benefits here too. This functionality as it’s implemented offers efficiency and evaluation particularly when dealing with really large data sets.
Hopeton: Okay so let’s dig into that a little bit. Can you say a little bit more about the kind of efficiencies that we’re talking about here?
Leith: Yeah and to get to that, let me digress just a bit into some of the technical details of how Explore Mode is implemented. So the way it works fundamentally is by taking our knowledge models and converting them into a series of database queries that can then be applied on top of your data. In this model of evaluation, essentially, we’re moving to a world in which we bring our logic to your data and this enables really efficient evaluation of bulk data sets. So within a post-trade reporting use case, this could allow you to process through millions of transactions or positions in order to determine eligibility with incredibly high throughput. So that means faster processing and that of course means lower cost. And importantly, within this new mode of evaluation, we’re still retaining all of Adept’s decision fingerprinting capabilities.
Hopeton: Okay, you might need to explain that… decision fingerprinting… what is that?
Leith: Yeah, so decision fingerprinting. We can think of a fingerprint as a unique ID for a specific execution path through our logic. Or maybe a bit more simply, we can think of it as a label for a decision where two decisions that share the same label will be guaranteed to share the same explanation at the most detailed level. If we are using Droit as a control or a QA for transaction reporting a user would, on a daily basis, process through all of their transactions for the day compute eligibility through Adept and then look to compare that determination against whatever they’ve gotten from their primary reporting system, and through this exercise of comparison against primary that user may identify thousands or tens of thousands of breaks. So obviously sorting through and investigating each one of those decisions individually is an unaccomplishable task.
Hopeton: Huge, yep.
Leith: So this is where decision fingerprinting comes into play. It allows users to take
those tens of thousands of breaks and very precisely group or bucket them into maybe 20 or 30 problems. So within those buckets, those decision fingerprinted groups, any representative single decision can be used to investigate that category of problem. Through our platform, we of course allow this introspection of the logic through the Logic Viewer and through the Digital Library and if a user understands precisely why that decision was made they now have the answer for that entire fingerprint group.
Hopeton: Okay.
Leith: So this allows users to really focus their remediation efforts They can understand how do I has made a decision and then use that to focus their upstream investigation to fix whatever rule or data issue they may be observing within their primary system.
Hopeton: Got it so, I see how teams could really change the way they interact with the rules and their data quite frankly across the the entire trade life cycle. And hopefully Explore Mode will make it easier for them to experiment, to get answers fast, and to trust the outcome. Leith, really appreciate the insights and the work that you and your team are doing.
Leith: Thank you.
Hopeton: And a big thank you to our listeners. If you want to learn more about the patent, check out the press release on our website droit.tech, and as always thanks for joining us and stay tuned for the next one. Bye for now.