techbullion.com

A Practical Approach to Modernizing Retirement Services

By Ellen F. Warren·Published November 18, 2026·Finance News, Fintech Careers

Karthikeyan Marappan is Director of Technical Delivery at Corebridge Financial

Technology modernization continues to reshape retirement services as organizations address legacy infrastructure, growing data demands and the emergence of AI. Karthikeyan Marappan discusses how these changes are influencing technology strategy, operational priorities and the experience of plan sponsors and participants.

Karthikeyan Marappan is Director of Technical Delivery at Corebridge Financial, where he leads technology strategy and transformation initiatives across retirement services. During a 25-year career spanning software engineering, application modernization, data and technical delivery, he has developed deep expertise in retirement plans, wealth management and the complex financial systems that support them.

Marappan’s career began in application development and progressed into senior technology leadership as the industry itself moved from heavily mainframe-dependent environments toward cloud platforms, API-driven architectures, advanced data capabilities and artificial intelligence. His responsibilities have included more than 60 mission-critical applications and an annual IT portfolio exceeding $10 million. Recent initiatives include a multiyear technology roadmap covering more than 100 applications, API-first integration, payroll and plan-servicing modernization, cloud migration and the application of AI to participant and plan sponsor servicing.

An IEEE Senior Member, Marappan has authored and presented technical papers at IEEE conferences and written two books examining AI, big data and modern cloud service management. In this interview, he discusses what retirement-services organizations have learned from years of modernization, where AI can provide practical value, and why technology decisions in a highly regulated financial environment still begin with understanding how the business actually operates.

ELLEN WARREN: You began your career as a software engineer in India more than 25 years ago, working with global technology companies before moving into technology leadership in the U.S. What drew you to technology and financial services, and how did those early experiences shape the work you do today?

KARTHIKEYAN MARAPPAN: I have always liked using technology to solve a problem that affects someone directly. I started in 2000 as a software engineer in India, working on a billing system for a U.S. telecom client. The fundamentals were simple but important, starting with understanding the requirement, knowing what else a change will affect and delivering it without introducing new problems.

I joined Polaris Software Labs in 2002 and moved onsite to support AIG’s insurance and retirement platforms. Working on production support, system design and disaster recovery for recordkeeping systems gave me my first real understanding of retirement services. These systems hold the savings people are depending on for their future, so reliability has a very practical meaning.

As I moved into project and delivery leadership, my technical background helped me understand what developers were dealing with at the application level and how those issues connected to the larger business and technology decisions shaping a transformation program.

EW: After working with retirement services technology for much of your career, where have you seen organizations underestimate the complexity of modernization?

KM: The level of integration is easy to underestimate. An application may look relatively self-contained on an architecture diagram, but in practice its data feeds dozens of other systems. A change to core recordkeeping can become a much larger effort once you identify the downstream applications and processes that depend on it. The technical change itself may be relatively straightforward compared with the work required across dependent systems, teams and operations.

The complexity also extends beyond the technology because many processes that appear automated still depend on considerable manual work. Employees may correct files or resolve exceptions behind the scenes every day, allowing a process to function even though the technology does not fully support it. Payroll processing is a good example. Files arrive in different formats, and staff may need to address those differences manually before the data can move through the process. Before we change the technology, we need a clear view of where the business is going and how the work actually gets done, including the system dependencies, spreadsheets, manual interventions and workarounds that could otherwise carry the same problems into a new platform.

EW: Your work spans the retirement plan lifecycle, from prospecting and onboarding through administration, participant engagement and investment management. How does that end-to-end perspective affect the way you approach a technology problem?

KM: A single request in retirement services can move through several teams and applications. A hardship withdrawal, for example, can involve participant service, eligibility and compliance rules, payment processing and data held in different systems. Improving one component without understanding the rest of that flow can create a new problem somewhere else.

Newer plan structures make those dependencies even more visible, with multiple employer plans and pooled arrangements involving different rules within the same plan and participants who move between employers. The technology has to support that complexity throughout the plan lifecycle, from the RFP and onboarding process through administration and the participant experience. I usually begin with the business capability we need to support and work back to the systems involved because applications will change over time, but the underlying business capability gives us a more stable basis for making technology decisions.

EW: Technology decisions in retirement services affect plan sponsors, participants, advisors, operations teams and compliance functions. How do you establish priorities when those groups have different needs?

KM: Regulatory exposure, customer impact and financial impact provide a useful starting point. Distribution processing, for example, touches all three. An error can create a compliance issue, affect a participant directly and require significant effort to correct, so it deserves a different priority from an internal report that is simply running more slowly than we would like. We also look for improvements that solve problems for several groups. Automating fee processing can reduce manual work for operations, give finance and compliance clearer information, and improve accuracy for plan sponsors.

The people affected by a change need to be involved early. Operations, compliance, advisors and technology teams can make better decisions about the tradeoffs and priorities if they are working from the same information.

EW: Your recent work has included a three-to-five-year technology roadmap covering more than 100 applications and an API-first integration strategy. What did that scale of transformation teach you about deciding what genuinely needs to change?

KM: Age alone is not a reason to replace an application. We gave each application in the roadmap a target state based on what the business needed from it. Some could be retired, some needed replacement or extension, and others were still doing their jobs well. Core recordkeeping illustrates the point. Some of these platforms have been operating for decades and remain extremely reliable. For these systems, strengthening an existing core to support growth can make more sense than accepting the cost and operational risk of replacing it, particularly when participants and plan sponsors depend on uninterrupted service.

APIs give us more flexibility in making those decisions because a well-designed integration layer allows established core platforms and newer applications to work together without forcing every component to move at the same time. We can modernize where there is a clear business reason and continue to get value from systems that remain dependable.

EW: Data strategy, governance and integration have also been major parts of your work. Where have you seen stronger data architecture make the greatest operational difference?

KM: Participant information is often distributed across systems built at different times for different purposes. Bringing that information together gives service teams a more complete view of the participant, which can improve response times, reporting accuracy and the quality of service. Modern analytics have also changed the relationship between business teams and information. People who once depended on IT to produce static reports can increasingly answer questions themselves using current data, allowing problems to be identified sooner and decisions to be made without waiting for another reporting cycle.

Strong data governance supports both capabilities through clear ownership, consistent definitions, appropriate security and reliable integration. AI raises the stakes further because its effectiveness depends on the quality and trustworthiness of the underlying data.

EW: Your recent work includes developing AI solutions for participant and plan sponsor servicing. Where do you see the most practical applications for AI in retirement services, and what needs to be in place before it can be used responsibly at scale?

KM: The strongest applications tend to involve repetitive work where people need to locate, interpret or route information quickly. In a call center, AI can help a representative find relevant information during a conversation with a participant or plan sponsor. Plan onboarding also involves documents, setup activities and repeated exchanges of information, so there are opportunities to reduce manual effort there. Other practical uses include document review, request routing and account-history summarization.

The more AI becomes involved in these processes, the more important it is to match the controls to the risk of the activity. Data must be governed, access should be limited to what the AI needs for a specific task, and actions need to be traceable. Human review remains important if a decision can affect a participant’s money or create a compliance issue, and integration needs to happen through defined, governed interfaces. I apply similar criteria when judging AI hackathons by looking at whether an idea solves a real problem, whether the architecture is secure and capable of scaling, whether the technology works in practice and whether there is a viable business case. I use the same criteria for AI work in a regulated business because they keep the focus on practical value.

EW: Retirement services organizations operate under extensive regulatory, security and audit requirements. How have those obligations shaped your approach to innovation as architectures become more distributed?

KM: Compliance has been part of my work since early in my career, including SOX and SOC1 reviews, so I tend to think about it as part of system design. Retirement services technology also has to account for Department of Labor requirements, FINRA, state regulators and legislation such as the SECURE Act. Clear ownership becomes especially important as systems extend across cloud platforms, APIs and outside vendors. Business areas need to understand the rules and data they own, and technology teams need to know where controls belong and how they will be monitored. Vendor evaluation has to consider security, data protection, integration, scalability, regulatory requirements and cost before a solution becomes part of the environment.

The same discipline applies to AI, where access should be limited to what is appropriate for the task and every activity should be traceable. We have to establish clear boundaries early to give teams a practical framework for developing and deploying new capabilities responsibly.

EW: Your career has taken you from mainframe environments through cloud migration, APIs, modern data platforms and AI. Which changes have had the greatest effect on how retirement-services organizations operate?

KM: Mainframes have provided tremendous reliability in this industry, but connecting older environments to new digital capabilities was historically difficult and changes could take considerable time. APIs made a major difference because they allowed us to expose capabilities from core systems to newer applications without replacing everything at once. Cloud platforms added flexibility and scale, and modern data platforms made it possible to bring participant information together across systems. Straight-through processing has also improved contributions, distributions and fee processing by reducing exceptions and manual rework. AI is now beginning to affect both customer service and the way technology teams work.

We still need to enroll participants, administer plans, process transactions accurately, protect information and meet regulatory obligations, regardless of the technology supporting that work. Earlier in my career I worked on disaster recovery; today we talk more about cloud resilience. The technology and terminology may be different, but the need to provide dependable service has not.

EW: Alongside your industry work, you have published and presented technical research through IEEE and written books on AI, big data and cloud service management. How does that work influence you as a practitioner?

KM: Writing makes me examine why something worked and whether the same approach would make sense in another environment. Day-to-day delivery tends to keep you focused on the immediate problem, but writing a paper or book requires a closer look at the reasoning behind a decision and can expose assumptions that are easy to overlook when you are moving quickly.

I completed a postgraduate program in AI and machine learning at the University of Texas at Austin to deepen my understanding of the technology, and my involvement with IEEE exposes me to people working on problems in other industries and technical disciplines. I also enjoy mentoring engineers, participating in senior-level interview panels and judging hackathons because these activities keep me connected to new ideas and different approaches and give me an opportunity to contribute to the development of other technology professionals.

EW: As AI, changing participant expectations and continued modernization create new possibilities for retirement services, where should technology leaders concentrate their efforts over the next several years?

KM: Technology strategy has to begin with the capabilities the business needs, the areas creating the most friction and the risks that need attention. For participants and sponsors, some of the most valuable practical improvements include cleaner payroll integration, more straight-through processing and fewer transactions that require manual correction or a phone call. Trusted data, clear ownership and well-designed APIs provide the foundation for more personalized service and give organizations a stronger base for introducing AI where it can help employees work more effectively or improve service, with controls and human review appropriate to the risk.

Retirement products, regulations and participant expectations will continue to change, so the technology environment needs the flexibility to respond. Ultimately, the measure of a modernization program is how reliably and effectively the organization serves the participants and plan sponsors who depend on these services.