PegaWorld iNspire 2023: Citi modernizes the Forbearance process to work smarter with increased controls

Citi executed a multi-year, multi-release program of initiatives to revamp and modernize the Forbearance systems into a lean and adaptive model. The program accomplished this by leveraging the Pega Platform to reduce risk, increase controls around enrollment, set-up, and life-cycle management, as well as improve risk reporting and analytics. Citi launched the first release of FDM (Forbearance and Deferral Manager) in 6 months (September 2021). Continuous releases will continue through 2023.

Transcript:

- I am, like Dale mentioned, I am a application development group manager in PBWMT, handling multiple cross-functional teams in core collections. So, going from the challenges that Dale mentioned to the guiding principles that helped design an all-encompassing solution, operational excellence, omnichannel experience for agent and customer channels, enhanced risk reporting, shortened time to market, configurable scheduler, reconciliation of actions taken across the ecosystem, controls and management. We really wanted to enhance our controls and analytics, data governance and management and integration framework. In business process management, we were looking for workflow automation, real-time activity dashboard, real-time surround updates and case view and audit trail. To gain competitive advantage necessary, we really were looking for A BPM product that not only has this capabilities, but also able to help us effectively manage risk controls and compliance. As you heard before, Citi Pega relationship has been long term. We have our own Pega COE, who actually helped us in the initial stages answering a lot of questions regarding the framework that they're using. And then Citi architectures also knew about Pega capabilities. Pega vendor showcased us the Pega capabilities and helped us design an approach that would meet most of our challenges. With Pega, we were able to readily get workflow automation, real-time activity dashboard case view and audit trail, and by leveraging further capabilities in-house and Pega's project, product, we were able to meet all the needs that we really wanted to successfully achieve, plus enhance our system significantly. Now onto to the solution architecture. So here we are trying to show you the architecture that we have for our new system. Our old forbearance system was fragmented and it really lacked the business process design that real needed to effectively manage the interdependencies between the subprocesses. While working on the designing the solution, we were looking to optimize and streamline processes. So we revamped the system and broke it into three core areas. Eligibility, enrollment and lifecycle management. As you can see here, the respective channels would invoke the forbearance, would invoke the microservices in order to do eligibility check. And then based on the results, they will invoke the enrollment request. As soon as the enrollment request hits forbearance and deferral manager shown here in orange, a case would be created and the account will be updated real time at Citi's, all applications, plus the product processors, in real time. By leveraging the power of Pega's case management, we were able to achieve better visibility controls and coordination across the systems. The microservice provider effectively directed transaction volume to the system of records, and enabling real-time data retrievals and updates. We were able to go live, tech-live MBP in six months because of the microservice able to direct the traffic between the legacy flow and the new flow. Based on that, we were able to do like some sort of a production parallel, like using fire and forget rollout request, which were controlled by our business and risk. So, we actually gave them control in their hands to determine which arrangement for a specific portfolio they really wanted to go live with. Our solution enables unified interactions and experiences across multiple touchpoint. Program execution. So this is how we executed our scrum model. Our collaborative approach entailed engaging and combination of onshore, offshore Pega resources, Citi's business, risk, technology teams, all coming together along with our vendor TCS team, all working actively in a scrum. This execution model played a pivotal role in accomplishing our objectives. By leveraging Pega and working closely with their subject matter experts in a managed service model, we were able to achieve the desired outcomes. This partnership with Pega helped us gain valuable insights and accelerated our product learnings. Establishing an agile team from scratch while managing 22 plus sprints, interacting, bringing together about 15 to 20 development teams, handling close to 1,600 user stories was an impressive endeavor. All that was we were able to achieve because of strong leadership alignment across the stakeholder community, unified vision, shared objectives facilitated towards consistent progress across the various program stages.

- Thank you, Kinnary. So as Kinnary mentioned, the two things that stood out, we had 15 to 20 development teams and about, and on average, 1,500 user stories. When it comes to the actual journey itself, all this looks clean, I can assure you it didn't start that way. So when we talked about the vision for the program and why we needed to change and, really, where do you start sometimes? Getting a program off the ground? So this was an enormous undertaking. We began our current state assessment in which we deconstructed the architecture, looked at all of the customer journeys and created our vision in 2019. And we were prepping for a late Q1, 2020 build kickoff. So obviously we had a little bit of a challenge leading into Q1 of 2020 to start anything. A lot of our resources were then decked against the pandemic, and so we had a little bit of a delay going into 2020. But despite that delay, we began probably, I'd say, Q4 of 2020 is when we actually started and were able to get everything off the ground and get these teams, and as Kinnary mentioned, our first of many production releases occurred in July of 2021. So that timeframe from when we started building to our first delivery was pretty quick. So as many of you know, delivering any project on time and on budget, that's usually the the two things anyone's gonna ask you, you just need to be on time and on budget, and I dunno about you, but most of any project, small or large, that's very difficult to do. But we essentially did just that. Despite our delay, you know, we did revise our timelines, but we stayed very true to to budget and very true to what we expected to deliver and when. So, while our delivery model was very successful, the key thing I would say here on this slide is when we were in the run phase, we had to, and Promitti mentioned this in the customer decision hub, I talked about the numerous controls we had, and all the monitoring, and you have millions of customers that this impacts and we had to migrate all of the old cases also. So not only did we have new cases that we were then gonna migrate to this environment, we had all the old cases that we had to, to move over as, as well. And there was zero tolerance for error, which is, you know, causes a little bit of stress and sleepless nights, especially in July of 2021. But as Kinnary stated, we had a very rigorous testing and testing environment, and when we went live, we did not have any major issue and still do not to to this this day. And that is attributed to the many teams that we'll talk about in just a little bit. But I would also attribute that to Kinnary, herself, because while she's just one person, and this was an army of people working on this program and testing and you name it, going through 1,500 user stories is not easy task. And Kinnary knew what that vision was and that vision was to break our release into program type, and then by line of business so that we could then monitor how that went into production. If we did see an issue, we could very quickly get underneath it, learn from that experience and go to our next release. So our foundational delivery was complete in late of last year, 2022, with enhancements added functionality and we have had many internal teams and internal servicing units look at this platform, and want to also add to it. Unfortunately, you know, based upon our initial scope, we could not include a lot of those, but there is an enormous future for this platform. And, we do expect program completion in Q3 of this year. And then the platform itself will become its own entity, which will then integrate into some future enhancements that Kinnary will touch on. So, we're gonna talk a little bit about results. So, results, so this was a highly visible project, but one that brought many of our groups together and collaborated on the execution model. So these partnerships fostered much stronger awareness and commitment to the successful delivery of the product. As I mentioned earlier, sticking to why we're doing it, everyone that's in project management knows that you, a lot of times, get so immersed that you're just focusing how are we going to complete this and what needs to get done, and you often forget about why you're doing it in the first place. And that is one thing that we had in common. And you can always say you have aligned goals to get a program completed, but when this, you have a multi-year program, multi-year funding, this was more of a family-like bond that has continued. And, right now, as led by Andrew Peterson, it is slated to complete this year. The program was also recognized at our Citi expo in 2022 with our off offshore partners. The transformation, oops, sorry. Yeah, the transformation of our forbearance process using Pega is crucial to our future of our bank for several reasons. So, the first, first was our platform improvement. So creating that foundation for case management using a center out model of centralized rural logic was huge. So, while we talked about center out in the introduction of Pega world this morning, getting to that layer cake model, and, or centralizing even more, is definitely in our future, but we definitely had to start somewhere. The case management platform did result in a 50% reduction in our enrollment backend manual work which was significant, because that's an impact to our operation. So while our efficiency was greatly increased through the real-time enrollment processing, we went from several hours for an enrollment to occur to literally under three seconds. So within under three seconds, a customer could have either talked to us on the phone or maybe they were in a digital channel, they would actually see that enrollment in their account if they logged in. So, Pega's automation capabilities enable us to scale our forbearance operations effectively, 'cause that was another challenge. And we went, which went from several, excuse me, but as the volume of forbearance request fluctuates, we can easily adjust our resources and processes to handle increased demand without compromising the efficiency. So, from a data and analytics, so to summarize all these areas, data, as I mentioned in the analogy, was sitting in several systems. So when we had to extract data or parse data for the purpose of analytics or even strategy, we were very reactive to the environment. This allowing us to then also centralize that data and have the Pega reporting was instrumental in our ability to now be strategic. So it enhances our partnerships with risk operations and really brought everyone together. So, now we're gonna talk a little bit about the future ideation and application. So, Kinnary?

- Sure. So like Dale mentioned, you know, we are no longer in the reactive state as we were before in the forbearance space. And that allows us to look at automation. Automation is going to be our primary focus, including automating controls and monitoring task, and also looking out for intelligent routing for remedial items, as in when they're found out in the exception process. Also, we probably will look into something that was brought up in the keynote speech today about process mining. The other focus item is going to be ongoing evaluation of integration opportunities to streamline further systems as part of product strategy, like what Dale mentioned. By strategically leveraging automation, integrating systems and streamlining contingent servicing functions, we plan to drive greater efficiencies, reduce operational cost, and provide a superior customer experience. All this was not possible without a whole crew of team behind both of us. We are immensely grateful to the extensive roster of individuals who played an important role in making this transformation a real success. It is with heartfelt appreciation we acknowledge the invaluable contribution of collections and recovery, technology, business, risk, operations, architects, project management, a lot of teams out there, plus our esteemed partners at Pega and TCS. Plus our esteemed partners at Pega and TCS who really held our hands and made this possible. This community's total dedication and support is what really made this program a triumph. Thank you. And we are ready for questions.

- [Audience Member 1] I have a question solution architecture.

- So we did not use the microservices provided by Pega at this time. You know, that is on future, but right now we used our own microservices, which interact with our front end.

- [Audience Member 1] Okay.

- So the microservice was sitting in between Pega, layer and the channels.

- And part part of that, too, was due to, when I mentioned we had to start some somewhere, and given the nature of this work, actually one of the main decision engines that converted, or there were two parts to that, Accordiant was actually one of the existing for our branded cards from an eligibility standpoint, but the microservice, I would say our future is to then integrate with Pega for that. But we knew to, based upon the timing, the program complexity, and getting to a foundational end state, we had to utilize our internal microservice for that reason.

- [Audience Member 1] Got it. And I just have one more question.

- Sure.

- [Audience Member 1] You talked about the cards, you know, white labeling and all that at Costco, so, do you guys handle the card disputes, as well, as part of your division, or it's a different NOB?

- Disputes would be separate.

- [Audience Member 1] It's a separate one?

- Yep, this is all just collections, but you bring up a good point. One of the, you know, I wouldn't say a challenge, but part of our program scope was to ensure that there was easy integration with other Citi servicing platforms. So as we built, we wanted to make sure that it was basically an easy tie-in, and wasn't gonna be complicated for an area like that, and, or, we can then kind of converge at some point.

- [Audience Member 1] Perfect, thank you.

- Uh-huh.

- [Audience Member 2] I was listening to what you were saying with regards to the data that we're using for reporting and being able to be proactive with that information. I was wondering how you implemented the actual storage of that data, whether like you're using it within Pega itself, or you're using some sort of an ETL to put it somewhere else, so all the reporting doesn't necessarily maybe impact your production systems, or what exactly you did for that particular piece? Where does your data reside with regards to that kind of endeavor itself?

- So the data related to a case is in Pega, so the data related to the case is in Pega. But what we have done, if you looked at our architecture, all the systems that touch anything related to forbearance, we put all the data into EAP so that it is readily available and not siloed.

- [Audience Member 2] Okay.

- Okay.

- [Dale] Okay.

- So we are NDMing every night, full data, with all our batch processing and everything that has happened, and we send several tables to EAP. So it's every day updates.

- And one other piece to that, too, that while we were in motion with this program, there was also a lot of, this was probably one of the pioneering efforts with EAP, in terms of large volume, not the absolute first, but that was something that we knew we also had to integrate into our model because we had other data sources we also had to feed, ideally cloud would be, you know, a good environment for that, right? But that was another component that we knew, yes, you know, we would like to get there at some some point, but we knew from a program, from a program stability and scope standpoint, we needed to stick with the current model, including EAP.

- [Audience Member 3] Good afternoon. May I ask what you found the most challenging technical and project management issues were and how you dealt with them? And as a follow-up, is there anything that having gone through the process you would do differently, or advice that you would pass on to other people going through a similar process?

- Do you wanna answer first then I'll take? We'll both take that one, but-

- So I think if I heard you correct, you said challenges with project management, is that what you asked for?

- [Moderator] That's correct.

- [Audience Member 3] Technical and project.

- Okay, so the technical stuff, let me go first with that. Since we were actually migrating all our rules and system knowledge from one application to the other, I think at times we realized that the decomposition of the old systems was not very accurate, you know, because we were building requirements off the old system. So, I think if we had to do this all over again, we probably will factor in a lot more time for decomposition and also for everybody to understand how the existing systems are playing. Like Dale mentioned, if you touch one, if you move one, something happened, right? So some things like that we did see during our testing and then we were able to quickly correct it with the additional user story. But that was in technical, in program management, I would say-

- [Dale] I'll take over that part.

- Okay, Dale can cover.

- So, from a program management standpoint, I would say it's twofold, maybe even three, but the first I would say are resources. So as I mentioned, we kind of crossed over, so when we started in 2019 and crossed into 2020, one of, I think, my biggest fears was resources, 'cause with a program this large, you can churn resources, people move into different areas, new people come in, and ensuring that we had stable and consistent resources was probably number one. And we were able to do that, as I mentioned, kind of that family-like atmosphere, but the more commitment we gained and the more eyes we we got on, the easier it was to keep those resources. The second was testing, I would say. So, testing 1,500 user stories, obviously not all at once, sometimes it felt that that way and the amount of scripts that we had to go through knowing that we essentially had to execute flawlessly. The testing, you know, I would say was something we learned from each iteration and each release, where we could have been maybe a little less, or where we could have been a little more thorough. So, you know, that's why we went with that program type and line of business because they each behave differently, and sometimes have different requirements, both from an eligibility standpoint, all the way through their life cycle. The third, I'll just throw this out there, was when you're in a life cycle, there are many things that can occur with an an account. So, if somebody is on a forbearance program and they have another life-changing event, or they go through bankruptcy, or many things, we had to make sure that we, and Kinnary touched on this, the old requirements, we had to ensure that as we built the new, that we accounted for every possible instance, but also kind of updated, and had a way to actually get that into reporting so that it wasn't sitting in other systems, but all kind of together, which, in my mind, you know, leads us to, well, we're not tied into the customer decision hub, that's probably a big component futuristically, I would say where that could fit in, so.

- [Kinnary] Nice.

- Sorry for the long-winded answer, but.

- [Audience Member 4] Thank you so much for the presentation and it looks like the project is definitely initiated in a very interesting time from 2019, and all the way going to now. My curiosity, as you already mentioned, that between new requirement and old requirement, that's definitely need manage, I was wondering, do you might to use the example of like a covid-related deferral and customer assistant program? How does that actually throw a wrinkle or disruption to your project and how did you manage it?

- So when covid occurred, and we had covid deferrals we had to offer, that had to be built yet in parallel 'cause we knew when we were, this goes back to probably program challenges, where do we start? Do we start with deferrals? Do we start with forbearance? And we knew to build the foundation, forbearance was the best place to start that because it was the most, I'll say complex, from a covid deferral standpoint. When we built the deferral model, we based it on our, I would say more natural disaster, but the ability, so the goal with our deferment platform, is to be able to kind of flex it. So, the needs of different deferment options, sometimes they're one month, sometimes they're two, being able to essentially just change rules and instantaneously still, so, if somebody, let's say, was on a forbearance prep program and, or, just came off of a deferral and then needed another, but the requirements were different, that it would be easily changeable.

- [Moderator] Any other questions?

- [Audience Member 5] Thanks for that. Did you have a historical system and what did you do with like historical records? Did you have to sort of archive or bridge them into a new forbearance system?

- [Dale] Yeah, we can't really hear.

- [Crew] The legacy data migration.

- [Dale] Oh, the legacy data migration?

- [Audience Member 5] Sorry, is that better? Yeah, the historical cases. How did you deal with those? Did you archive them? Do you still need to have access to them?

- [Dale] Do you wanna take that one?

- [Kinnary] He's asking for the, are you asking?

- [Crew] Your old data.

- [Audience Member 5] Your old data, your old cases, your old historical data.

- So when we had existing forbearance arrangements sitting on the old system and how we migrated those to the new?

- Is that what you're asking?

- So, in that, I'll take it from a business standpoint and then I'll let you speak to it from a an IT standpoint, from a business stand standpoint, that was probably, you know, our next biggest fear, 'cause while the new enrollments we knew we had no issue and it was functioning very well, and then migrating the old, we did that in the same sequence. So after we released the new enrollment capability, we, I think right around probably two weeks after that, started a migration of old, but the biggest part I think that was successful there was looking at, and this was something we kind of learned through the process, so if we, let's say you're migrating 200,000 accounts from one to to another, looking at and testing a variety of different program types and scenarios so that we knew how they were going to behave, so, we essentially would migrate of that variety first, test those, see how they reacted in their new setting, I'll call it, and then that allowed us to do kind of a mass migration over, I think, it just took a few days, Kinnary?

- [Kinnary] Uh-huh.

- That's correct. Is there anything else you want to add to that? - So, basically, you know, we knew what other data fields that our new FDM would need. So we built a one-time file from our old system and what we did was we went by the same arrangement type and portfolio at a very large level, and, also, looked for, you know, being able to hit all the business logic in the first week. Like, so we really had a list of accounts which we knew are in certain stages in their forbearance and we kept an eye on them. So that's how even post install, post migration, all eyes were on those things and then we didn't find any issues.

- [Audience Member 6] Yeah, absolutely. I really appreciate all the info. My question is specific to other Pega products or services that are potentially incorporated, or if you have plans to incorporate them, for example, within the eligibility model, are there things like AI decisioning on the roadmap to help inform that process?

- Yes, that probably will be in the future. But, yeah, we have that, the same thing that promo Promitti talked about, the customer decision hub, something like that we are going to probably look into for the eligibility, as well.

- [Audience Member 6] And are there products in the Pega portfolio that would help assist with those some other things? And this is where I'm not as familiar with the whole landscape of Pega products.

- I didn't hear your second question properly.

- Hi, this is Sushawn Tera, I'm from Pega. To answer your question, yes, Pega has decisioning capability and products that help facilitate the yeah. And we can maybe talk after the session, and, yeah.

- [Moderator] Yeah, you could use that for like monitoring things like that. Any other questions? Okay. Just getting my steps in.

- [Audience Member 7] Yeah that's right. Just looking at your execution model. so with your COE and your competency center and your dev teams, was that stood up as part of this program? Did you already have a bunch of that stuff in place? Is that what you'd already used on prior programs or projects? Just curious if how that went?

- So, most of the Citi's internal teams, yes, we had everything already out there, but with Pega that is how we use, we got some of the LSA's onshore, offshore and used them in a managed service model.

- [Audience Member 7] Okay, okay, got it, thanks.

- [Moderator] Any other questions? All right.

- [Audience Member 8] Oh, hey, thanks, a great presentation. I have a question. You mentioned that you had 20 plus teams working on this project, how many were actually Pega SMEs, Pega experts out of those 20 teams?

- Pega experts, it's actually just one team.

- [Audience Member 8] Just one team?

- 18 to 20 teams are the ones we integrate for the whole architecture, right? Because we have to update some system of record here, some system of record there. That's how, yeah, it's just one team. The main thing that was in orange was the Pega team.

- [Audience Member 8] So 20 teams, it was the whole ecosystem for collections, right? That basically what involved, right? To build the collections project?

- Yes, to a large extent, those are usually that much integration we have for major collections project.

- [Audience Member 8] Thank you.

- [Moderator] All right, I think we're at time. Thank you very much, Dale and Kinnary, great job.