TIAA modernizes the customer journey with AI

The financial services organization overhauled its customer service operations, weaving together machine learning models, generative AI, and customer journey mapping to improve customer experience and predict trouble points.

Customer experience is a key pillar of TIAA’s mission to help employees of nonprofit institutions achieve their retirement goals. And just as retirement decisions can be complex, so too can be easing customer journeys, from contributions to investment decisions to beneficiary designations and more.

Achieving end-to-end visibility — in real-time — into these journeys, some of which involve highly customized legacy contracts typical of a 100-year-old company, would enable TIAA to better serve its customers across the various decision and touch points along the way from enrollment to retirement.

And so the New York-based financial services organization launched its Journey and Service Operations Center (JSOC) initiative, which leverages proprietary and off-the-shelf machine learning models to help walk institutional and individual investors through challenging financial decisions, ranging from 401(k) rebalancing and retirement planning to calculating tax implications on cashing in an IRA at a particular age.

Sastry Durvasula, chief information and client services officer at TIAA, says the multilayered platform’s extensive use of machine learning as part of its customer service line partnership with Google AI makes JSOC a formidable tool for financial and retirement planning and guiding customers through complex financial journeys. It is not for simple tasks such as changing one’s address, he notes.

“There is no journey map anywhere for servicing these clients on the most important financial decisions or transactions they’re making,” says Durvasula. “This is our first foundational step in automating that and making it as proactive as possible.”

Putting customer experience on the map

The cloud-native project, which began in January 2023, touches all layers of the financial services company’s digital infrastructure — from the front end to the back end — enabling customers and call service representatives to perform a great many tasks more smoothly, thanks to a data visualization UI that resembles a subway map. The map displays each step of the customer journey and assists customers with questions when they get stuck, Durvasula says.

“We are now for the first time giving our customer service reps a subway map so they can see where a customer is in a journey and may have questions,” he says, noting that retirement planning requires many steps and TIAA could not locate a commercial offering that would give its customers access to all the data they need to develop a plan.

To make that possible, JSOC reaches into the company’s cloud infrastructure to access data using a range of technologies, including AWS Gateway, Aurora, OpenShift, Secrets Manager, Athena, and Kafka, to name a few.

TIAA has also equipped JSOC with AI operations (AIOps) functionality to “proactively understand what is happening with anomaly detection, incident response management, root cause analysis, and predictive analytics of different customer journeys,” Durvasula says.

For example, JSOC includes an incident recommendation and resolution engine, customer anomaly detection engine, and aged customer incidents dashboard, which combine to help call center representatives simultaneously troubleshoot and predict customer challenges.

The third component of the system employs generative AI to empower TIAA call center representatives to better assist customers on complex financial journeys and use sentiment analysis to assess customer satisfaction, Durvasula says.

“Reps can make a call to the customer who may be stuck somewhere on their journey, in the middle of a transaction. It could be a digital aspect of the journey or even a physical element because there are physical steps involved in complex transactions. Some transactions start digital and end with paper in the back end,” Durvasala says, adding that here is where Google’s gen AI is used to determine customer sentiment and satisfaction. “The third component is gathering customer feedback and making sure we understand the sentiment to improve our services and journeys.”

Ray Bellucci, head of recordkeeping and chief administrative officer for retirement solutions at TIAA, says JSOC is already having an impact, making it easier to identify complexities in TIAA participants interactions so the company’s customer agents can proactively address their challenges.

“The simplification of these complex customer journeys helps instill confidence in our participants,” he says.

JSOC, which TIAA opted to build in-house because existing commercial products did not have all the tools needed, has earned TIAA a 2024 CIO 100 Award for IT innovation and leadership.

The power of AI

JSOC’s AIOps functionality has thus far gone live on mobile platforms, leading to several positive outcomes, including a 95% reduction in time identifying the root cause of a problem and 80% improvement in identifying anomalies, according to TIAA. Once the AIOps aspect of JSOC identifies an issue, it directs the failure to a human to intervene or uses machine learning with humans in the loop to “self-heal” more complex patterns, the company says.

The AIOps functionality will be extended to all customer journeys and platforms by year’s end, according to TIAA.

The platform was built iteratively, with outcome-based solutions released along the way. Using this method, TIAA was able to identify real friction in customer experiences and bring together app developers, customer service agents, and customer experience and design professionals to resolve the issues as they arose, Durvasula says.

One analyst says a solution like JSOC that uses both generative AI and machine learning models and is built in collaboration with customer service teams is ahead of the pack in an era still dominated by AI experimentation.

“You can’t just go to an IT organization and ask them to build an AI model for retirement planning. That’s very business-oriented and you need subject matter experts who understand the topic and can provide guidance and knowledge to the AI engineers,” says Kevin Prouty, group VP and general manager of the tech buyer practice at IDC. “That’s pretty advanced.”

For TIAA, finding internal subject matter experts who understood the various processes in order for IT to collect and record all that “tribal knowledge” was challenging; so too was locating, parsing, and organizing data from a wide range of systems, including legacy platforms that required special integration, according to a TIAA representative.

“You have all these different, disparate systems that have different sets of capabilities but there is only one end customers trying to do the transaction. So how do we bring these things together?” Durvasala asks. “It is an integrated view of AI in the back end, whether Splunk AI or Dynatrace for customer journeys or the Google model that we use for the customer side of the house which uses the model garden. We have integrated all of that alongside our own proprietary model and recommendation engine. This is the most complex ecosystem I have seen.”

SUBSCRIBE TO OUR NEWSLETTER