PegaWorld 2026: Transforming Back Office Operations at Blue Shield of California with Pega Workflow Automation

All right. Hello everyone. Good afternoon. Welcome to our breakout session. My name is Rija Pagarkar. And today I'll walk you through how at Blue Shield of California with Pega, we transformed our back office operations from fragmented manual processes to intelligent workflow orchestration. I will share real examples from our Medicare enrollment automation and inventory workflow management. And I will also share with you how we plan to use AI to take this transformation even further. So with that, let's get started. A brief introduction about myself. I lead product management for platform strategy and operations with a focus on Pega based workflows and operational transformation. In the last seven years, I have, uh, my work has mostly centered on building case management, scalable case management platforms integrated with our core application, which is facets. It's our system of record and also other data sources, internal and external, both.

And more recently, like many of you, I'm sure I have been ideating and, um, kind of, uh, trying to deliver AI backed workflows. So I think many of us are doing the same right now. A little introduction of our company. Blue Shield of California. It's an independent. um health plan nonprofit with more than 4.8 million members. And our mission is to ensure affordable and high quality health care to all Californians, which sounds easier than and done. It is quite a challenge. I think some of you will understand and appreciate with cost of health care. It's a mission that is very. It's it's a very important. And it's also challenging at the same time. And for this mission to succeed and stay successful, we need really strong and resilient operations. Foundation. So a few years back when, when we when we when we grew at a very fast pace, we really needed to rethink about our back office operations and how they are able to support speed, compliance and quality, which are really three Cornerstones for our success.

So this back office transformation began with Medicare enrollment automation. As some of you know, Medicare is a is a federally federal health care program that's offered to individuals age 65 and over, with some exceptions. Of course, that makes it a very complex process with all the strict regulatory as well as compliance requirements. And we wanted a solution for Medicare enrollment that could scale, that could remove manual work, and that could give consistent outcomes, which was really important across all channels, like consistency across all channels. Uh, a little closer look at the guiding principles that we grounded our solution around for Medicare enrollment as as we can see, digitization and optimization and automation is right at the center of, of, of it all. But what we really wanted was a streamlined and simple omnichannel enrollment platform that our brokers and sales teams and other stakeholders could rely on.

Right. And, and for, for those consistent outcomes and efficient enrollment processing, which is critical to their success and for them to keep their focus on our membership, which is really their job, instead of worrying about what happened to the enrollment, whether it got processed, whether it got delayed, uh, things like that. So that was really important. We also wanted a solution that could scale with that growing membership and evolving business needs. Changing business needs so that scalability was critical. And we also wanted our business teams to have more control so they don't have to rely on it for everything. That is always changing. In the case of Medicare and maybe other processes also. So through that configuration based design and removing operations, removing manual work, we we had to provide more efficiency to our business teams. And most importantly, we wanted a platform that could support the highest level of customer service to our members.

I mean, that's the reason we are here. And that was one of the most important factors for the Platform selection, as well as the capabilities that we built on it. So just wanted to show you the the view or the process of our Medicare enrollment before we implemented Pega workflow. As I won't go into every box, but applications were received from all the channels like we call them electronic channels. If they are coming electronically, mostly from. They were mostly file based ingestion through nightly batch jobs. Those enrollments would come from our broker systems, would get loaded to the Legacy system. And that was actually the end of automation because the next day, morning supervisors and team leads really had to pull reports to distribute and assign work in among the team members based on line of business, because even within Medicare there are few variations like individual group and Medicare supplemental.

So. So the work would get assigned. And there were very frequent need for handoffs between the teams also. And there was really no way at that time, except again, to use some kind of reporting and statuses, like identifiers to know which team has to now perform their task or their action and processes. And once they had their work, they had to swivel to multiple systems to fix the business rule failure errors, and in some cases, enrich the data, etc., validate the data. and that was all, you know, it was it would delay the process and potentially pose operational risks. So this was the process that we had as we were discovering and trying to improve it with the help of Pega. So with Pega, we re-architected the process end to end. We still got enrollments coming from our brokers and sales channels. We did try to leverage APIs as much as possible and do away with those file based intake.

We. We. But there were business rules that were run real time. We have integrations now. Real time integrations with CMS for eligibility check, which is very important for Medicaid enrollment and to know any exceptions right at the beginning of the process rather than wait till later, which was the case earlier. And we are able to enrich the data. We still have paper, we still have some members who use paper form, and we don't have an OCR at this point in time. But with all these integrations, we have managed to minimize that data entry by a huge extent. So with the help of some unique identifiers, system is able to fetch data from different systems, including CMS system for Medicare and our own code application facets. So the data is available, it is populated and processors only have to review it and make updates if there is a need for that.

So data entry is quite minimal now. And then all handoffs of course with Pega workflow, that is the core competency. We have those work queues and work baskets that allow internal teams to do the work that they have to do to keep the process moving. So really, we have moved from manual workflow to this intelligent orchestration of work. And that's our current state at this point in time for Medicare enrollment. But there are plans to do more, and I will try to touch upon that in the next slide, which is really taking a deeper dive into the future state or now current state that I just spoke about. This is how the landscape looks like at present, where Pega sits right at the center running those business rules, doing the case management orchestration being that unified platform for the enrollment process. And as you can see, there are integrations with different systems of record which enable the teams to get all their data and all the validations done automatically.

There is no need to go to any system and spend time doing that. Uh, we have integration with USPS database for address cleansing and standardization. And, uh, we have, uh, application enrollment status that is, uh, that is real time synced to our, uh, broker portal, our sales platform also as well as our call center. So anytime our member sales brokers need to know where the enrollment is, uh, in the, in the life cycle, they, they know that they know if they have to do anything or take any actions, because that information is also available and exposed to these systems. So this is our and again, like one more thing, like, and if there are any errors or fallouts happening later on after the enrollment is submitted to our core application, we call them fallouts or, you know, some kind of reconciliation, uh, work that needs to happen that also is ingested into Pega inventory, workflow management application.

We have a separate application for that. And I think that this is a good time now to, uh, go to our next example, because that is related to inventory workflow management. And I would like to go to the next slide. And that next example, which is email inventory management and how we apply the same principle that we just looked at for Medicare. Similar principles to automate and streamline that inventory. So, um, I mean, I know maybe it's like only our company and that we are so behind, but our teams in installation and billing are managing or we're managing large volumes of requests from other internal teams and external stakeholders when it comes to enrollment and group installation and billing. So there are emails. It's all done through outlook where they get requests either to clarify, uh, you know, something or even to make a transaction to do some change, like change the address or change the plan or, um.

Provide clarification on billing and payment, etc.. So they, they are that centralized team to provide all that information. And in the earlier Legacy system, someone needed to manually triage all all this email inventory that came in because there was no automatic way to do that. And then after triaging, it was assigned to the right team to work it like pretty, pretty much manually. So we recently transitioned all that inventory, email inventory, and as well as Other to Pega. And it's kind of a lift and shift right now where we have just created a foundation and we are just using Pega very core like basic capabilities. So today, email requests are ingested in Pega cases are created and they are routed using intelligent rules and skills. And there is of course integration with our core application. So data about that member or group is is taken from facets and it's displayed on the screen.

On the case UI. So users don't have to again, go to any other system in order to work that request. And there is a end to end audit, which is very important for all the requests. Or anytime we touch a member record or make updates to it, we need to have all those audit related things taken care of. There is outreach management because as you can imagine, when it's email, there is bound to be back and forth for, you know, additional information and clarification between the sender and our team. So earlier, there was no way to do that in the Legacy system, but we do have some intelligent and good features now available that notifies them each time there is a The response and that correspondent gets attached to the to the case. So these are some of the things that we have accomplished as of now.

But there is more to do. And we want to now enhance this workflow with GenAI capabilities, where incoming emails on various topics should get classified based on the type and how urgent they are. And we want a we want to automate it as much as possible, like where nobody has to look at it, it just should just get done. But again, that won't be the case always. So in that case, we would use the workflow that we have for our processing teams where they will get the cases. But we want to have AI summarize the request as much as possible. We want AI to give recommendation on complex um requests, like what is the next best action. I think by doing that, we can really start to seeing those operational efficiencies because there is a lot of like burden at this point in time because of this email inventory.

And after we do this or while we are doing that, we also want to look at removing email entirely. I mean, why should we have emails, at least from our internal teams? We want to give them a self-service portal, which is which has those predefined requests that can trigger and a workflow, an automated workflow in Pega. And it should just make the updates and any exceptions should get routed. So that will eliminate that back and forth between the internal team, which is really very, very unnecessary. So we want to get to this point. So what we want to do is not just optimize work moving forward, but get rid of or remove work wherever possible. So I think that's going to be really important. So. Looking ahead, our journey continues towards AI enabled predictive operations. As we look as we saw today, at this point in time, we have standardized our workflows.

We have incorporated some intelligence, we have given control to our business team, real time visibility, etc. but we want to have more AI features to triage to do that. Predictive routing, not just skill based, but based on other criteria like workload, etc. and then we want to auto generate response and auto resolve those requests with the help of AI. And then in the future, we we want to have closed loop operations where systems are optimized and they're continuously improving themselves and not just executing processes, but are learning and predicting what what can be done better and differently. So that's the plan. And. Just to summarize, we have come a long way. We have centralized workflows. We have reduced manual touch points. Um, and this itself, this transformation has resulted into significant impact and value for our stakeholders. But we are continuing to evolve and continuing to improve from here on.

So I think so that's the end of the presentation and we would love to take questions. And for that, I want to invite my team member, Achmea. Please welcome him. Uh. Thank you. Welcome. Achmea. So Achmea is a principal Pega architect with stellaris, that's our sister company. And he played a very key role to all the great things that we looked at, all the integration that we all have today. Uh, he has provided that technical leadership and, uh, we are continuing to improve on that. So we can either take, if there are any questions from the audience or, uh, we can. We can open up to questions. And again, if you would grab a microphone, John, I know you're always anxious, I think. And also if you want, you can share more. This was tremendous about the business piece, if you want, because you've got your technical expert.

Um, one of the things to be asking about if there's, there's things that you'd like to know about technically how they accomplish this with the back office as well. Yeah. Thank you Bill. Great presentation guys. Good job. Thank you. Uh, you had mentioned kind of looking forward and where you're going with all this. And can you share when you did like what, what year you did this? And when you think about how long it took. Can you kind of. I was trying to avoid that piece of information. Um, and then with that, maybe, can you also talk about how you might think about doing it today? I mean, we, you know, I'm fortunate enough to know you've played with Blueprint a little bit. How would you imagine doing something like this today with some of the different tools that we have? Sure. Great question John.

Thank you. So the timeline is roughly 2021 to to now, honestly. But again, that doesn't accurately reflect the, you know, the work that we have done because there were breaks in between. We also have internal processes to have funding and to, you know, uh, do the work. And that's, uh, that's been, uh, there has been some, you know, challenges with that. And sometimes we were blocked. But yeah, to answer your question, Medicare enrollment began in 2021. It was a phased approach. We wanted to go like really, we wanted to ease the business into using this new tool. And I must say, it was a very, very successful, uh, change management and strategy that we implemented. So we just had, uh, you know, we just moved them to Pega with a lot of their old ways and old things still intact, like even the assignment part.

Even though we had the ability for Pega to enable that get. Next, we there was some pushback because they were doing this work for so many years in a certain way. And we didn't want to, you know, just do something forcibly. So for a few months, we actually allowed them to just do it, like still manually, uh, distribute. So just giving an example, but, but, but we finished Medicare in 2024. It's fully operationalized. And now we have to look at taking it to the next level, doing, doing more work on our broker outreach and, you know, automatic update if there is any data is missing. So we want to improve our SLAs, further inventory, workflow management, email lift and shift just happened this year. So yeah, so the timeline is from 2021 to now. And you asked about if there is anything we could, we would do differently.

So my answer to that is I think we would, should have and we will moving forward, spend more time understanding our processes really well before trying to, you know, move them or change them and maybe improve them and get rid of some baggage or some unnecessary work that is happening today instead of automating, automating them. I think some things should not just exist. I mean, they are redundant. So I would, if given an opportunity to do things differently. I think that's one thing. We would put our money behind that work to do that real discovery and, uh, change, uh, some, some processes. No, I think John is trying to ask whether we will be using AI or not. You want to expand on that? Sure we will. Yeah, yeah. And I think. What we will do is we will augment AI with humans. So we will not they will be assisted and they will not be the decision maker like in enrollment.

What we can do is the compliance and the regulatory things. We cannot give it to AI. This should be human intervention needed there. Right. So AI can propose rules, the Pega rules, they will validate and then the human will take the decision. That is how I am seeing it. Yeah. That's that's a way forward. And the paper application that you said, that is also another use case. That paper application that we are getting. That we need to, we need to use gen AI. And we've already done a POC to do that. It was quite successful. And it's but we still have to get through our AI governance and get everything approved. Was there a question? Hi. You answered several of my questions in your last answer, but the order of operation, like figuring out the use case that you started with. How did you come about that you said you didn't do a lot of looking at the process beforehand and kind of lifted and shifted.

But yeah. You talk about the order of operation on how you figured that out. Of selecting Medicare and these other use cases. Sure, sure. I'll try to answer that. So actually, when we first started in 2021, late 2020, our plan was to just have one platform for all our enrollment and back office inventory. That was the original vision. But as we started doing discovery and learning more about each, you know, each like line of business, what we call and their complexities, we realized that Medicare is, was like a different kind of category in itself. And we could not club it with other enrollment processes. And also it happened to be our focus like for growth. We, our company had a very, uh, kind of like, uh, ambitious goals to grow, not just membership like numbers, but also our Medicare star ratings. So because of that, we prioritized Medicaid enrollment to re-architect and reengineer, uh, the process.

And it has been a really successful, uh, you know, experiment, I should say. So that was for, for Medicare. Thank you for the question. Got a, um, probably a tactical operational question for you related to the next, um, button that you were talking about and your transition from kind of going in, assigning work and getting people to actually, you know, kind of just get it pushed to them. Um, were you ultimately able to get the teams over their objection? You said that you kind of let them do it the same way for a couple of months. I'm interested if you were able to ultimately get them over their objections. Um, and if so, kind of how were you were able to manage that? And then a follow up is whether or not, what was your experience with kind of adjusting the urgency, the priority in order to make sure that, um, the work rooted in the sequence that the team thought made sense.

So the second part, you mean the urgency and SLA of the work that gets ingested in the application for teams to work on is the question around that? Yeah. So, so what we're experiencing with is when a CSR hits gets next. Yeah, their confidence level is not as high as I'd like it to be that they're actually getting that piece of work that they should be working on. They're thinking something else might be more urgent or a higher priority. So we're working through that trust. I understood and that's a great question and something that I can really I can share with you our experience regarding that. All right. So just to recap, sorry, what was the first part of the question again. Whether or not they ultimately got over their objections to using it. So we did it. I have to tell you, it was not easy.

So work like, um, work distribution was just one part where I think leads and supervisors. The day we went live, they their thought process was, oh my, what are we going to do now in the morning when we go to when we log in and we, we start working, we don't have any work to assign because teams can get their own work. And we could sense that that was really what was going on in their head. So, uh, and, uh, so as I said that, we let them do that because they didn't know how it will work. And eventually there were other things for them to do, uh, that, that were, you know, that came and they were okay with that. But one other thing that we, we saw there was a resistance. And pushback is when I was showing you the integration with the CMS system, where the eligibility is automatically validated and all the information that processor need to process the enrollment is there.

A Pega is getting it, but they were still just so, I guess, habituated to go to that CMS system themselves and see it for themselves. So for for some time we learned we didn't know it immediately, but we learned that that was what was happening and they were still doing that. So then we still had to do some work with the team to understand if there is, is there any like specific information that they're not getting today? And because of which or some like exceptions or some scenarios that is making them do that. And then we found out that there was, and because there was like maybe like 2% of cases, they were not getting all the information. That's why they were still using it. And then by just, you know, like taking care of that little thing, we then were able to move them entirely out of the the system.

So for SLA and urgency, we do have a fully delegated tables, right? We have business rules around this, that urgency based on the lob, based on the type of work, the urgency is being assigned. And also the SLA works accordingly. Yeah. And get next case. So those are also delegated tables to the business admin users where they control where the case is getting routed to if there is an exception. So which work basket and which skill a user need to have in order to pull that case. So that is configurable by the business admin. But we, we hit a kind of a roadblock with that approach with our inventory workflow management because they have emails, they have fallout, they have billing, they have like at least like 20 different types of work that they have to, the team has to work on. And for them, we had to do a little bit of customization.

So Pega out of the box skill based Getnext case did not work. What we had to do is create another page where the teams could first select what type of work that they want to work on, right? Because maybe on a particular day, they don't want to work on emails because they have some, they want to focus on something else. So they could, they can select that. And then when they click on get next, even if there is some email sitting right that they need to work on, they will not get that. They will get the other work which they have selected. So we did have that kind of customization just to make our users, you know, satisfied. And, uh, work effectively. Thank you. Hey. Quick question. Having worked on a couple of Legacy transformations myself, both in healthcare and finance, we have often seen that at times when we look at back end processes, they look quite straightforward.

But when we actually start implementing them in Pega, they turn out to be complex. So if you can share one such example that you might have faced, and following up on my second question, based on the trust when we go live with such products, considering our end users have been sitting on Legacy for far so long, they develop their own workarounds, certain problems. So how do we build a trust and confidence with them, and how do we end up launching such products or future to them so that they can start actually working on it without really feeling resistance towards it? So if they have workarounds that that supporting their work, right? Yes. How do we take in legacy systems? There are shortcomings and people find a workaround to get through it. But when we are designing entire systems that can help them get through it in one go, certain times there are hesitance from end users to do so.

So how do we gain their trust and build their confidence in these systems? OK right. Thank you for your question. So the first question is, um of legacy in legacy system. The process was simple, but when we transfer it to when we try to lift and shift, and I know that's not the best terminology, but when we move it to Pega it, it, it sometimes is not always that simple, right? Yeah. So, um, I'm trying to think of, uh, of an example. So one was like the example that I was just, uh, sharing where the legacy system did not have any intelligent, um, you know, like logic to, to, uh, assign work, right. They were just basically, uh, processor just had one single queue and they were able to just select the type of work that they wanted to work on, and they would just keep getting that work.

Where in, in Pega, we try to be more agnostic. At first we were like, if a particular processor has a skill set to work on a line of business, they should be able to work on the email and they should be able to work on some other, uh, work related to medical. Just an example. But when we try to do that, because this is out of the box. Pega. Right. Like to have that work basket and to have that skill. Uh, we, it was not, it didn't work. And we had to make that custom logic in the sense to give them that same capability that they had in their old application. So that's one thing that I can think of where we we thought it's going to be simple and it's going to be better. But then that was not the case. And we had to do something to match what they had in Legacy, which was a little bit extra work for us.

Um, so then, um, the work around like we do actually have a lot of, we have a huge roadmap at this point in time. So this is what I just showed you was like bare minimum like for Medicare. Yes. We, um, we, we created those integration points and we have that automatic enrichment and work basket, but there is still a lot of, uh, you know, there are other systems in Medicare enrollment space after Pega because there are, there is like some communication that has to happen with CMS that's taken care of by another system. And they. That's the system that today takes care of letters and communication. And there are fallout that happen. So and we have the same processors who are on Pega doing that work. And there is a there is really a room for improvement here and an opportunity to bring those processes or to get those business rules up front.

So there are no fallouts moving forward. So, so there are some, you know, work that happens and we want to eliminate that. So that's my answer. So kind of going back to what he was asking, I just want to clarify something. So it isn't that it was complicated in Pega per se, right? It's more that the old Legacy process was actually too complicated. And had we not done a lift and shift and focused a little more on improvement, it might have been more seamless, right? Absolutely. Yeah. Thank you. Thank you for clarifying that. So are there more questions I wanted to ask? Achmea if I think we have little time. Do we have time till five minutes? Probably. So Achmea just want to get your thoughts. Now, when we are all ready for bringing in AI into our back office in Blue Shield, because we have the foundation, from your perspective, sitting in the on the technical side of things, uh, what are your thoughts?

What do you think we need before we can go full? Uh, we have taken our first step, like, as you mentioned, that the AI governance approval, that is the first step for whatever I had in the so many hubs. I went there and they said the every company has their own AI governance model. So that is our first step. What we have taken. Right. And the second step would be our cloud migration. So once we will be in PegaCloud with version 25. I think we'll be getting so many AI capabilities that we can use. More features and like more like. Like it's easier and faster way to deploy things. Okay. Anything else? PegaCloud will make us scalable. We are like couple weeks away from going on PegaCloud for our back office. So other thing is talking to the operations team and finding out the use case, their pain points, and then see which AI use case fits to them.

Yeah. That is another exercise that way. That business has to use, but it has to it has to be done in partnership. Right. That partnership is needed. Okay. Thank you. Achmea are there any more questions? Thank you. Thank you for your time.