For more than a hundred years, the Teachers Insurance and Annuity Association of America-College Retirement Equities Fund (TIAA) has been helping teachers manage their retirement funds. In recent years, onboarding new institutional clients has become a significant challenge because of the complexity of possible client configurations.\n\nTo dramatically reduce the complexity of those configurations, the non-profit financial services provider has developed an intelligent system that leverages digital twin and graph database technology to automate the process.\n\n"At TIAA, we have a very complicated retirement product offering, based on all the regulations the IRS has," says Alex Pecoraro, managing director and head of retirement services technology at TIAA. "In order to do the setup, it requires quite a bit of business knowledge, and we have whole teams organized around doing that. We've been trying to move away from doing configuration to presenting the offer to the client, letting them make choices about what the product offering is that they want to do in terms that they recognize, and then translating that back into configuration."\n\nA big piece of TIAA's Defined Contribution Retirement Solutions is Outsourced Services, which calculates benefits on payroll remittance from TIAA's institutional clients \u2014 including universities, medical facilities, and government institutions. TIAA's Outsourced Services consist of more than 600 features, which can yield more than one trillion possible client configurations.\n\n"Retirement plans are very complex, from regulatory rules to plan document rules," says Paul Magro, senior director of business management at TIAA. "These rules can be done at different levels \u2014 the client level, the employer level, the location level, the employee level \u2014 so they're very much what I would call non-linear."\n\nBecause TIAA was relying on human associates to help clients navigate this process manually, the non-linear, yet highly interdependent nature of the data meant those associates had to become experts in particular types of offers. Specialized teams manually crated and tested the technical configurations against a client's desired operational model.\n\n"The way the business was structured, because these offers are so complex and you need to understand all the inner works of all the downstream systems, we were very 'functionalized,'" Magro says. "You had 'X' amount of associates that could handle certain types of offers."\n\nIf a big client or multiple clients were attempting to process a large number of the same type of offer at the same time, TIAA couldn't just assign other associates to help because they didn't have the expertise in that particular offer.\n\nAutomating onboarding\n\nIn 2017, TIAA responded by initiating a project called Institutional Client Onboarding Next-generation (ICON), which has since won it a CIO 100 Award in IT Excellence. ICON is an automated solution intended to assist associates with the onboarding process. Eventually, Magro says, TIAA would like to achieve a self-service process for plan sponsors.\n\nPecoraro says the key to ICON was changing perspectives. Instead of viewing it as a technical configuration problem, the team adopted a product adoption approach. ICON guides plan operators through plan service options based on their selections, and the results are validated against a digital twin that takes the product metadata, operational flows, overlapping interdependencies, nuances, and business rules and incorporates that data into the product offering and resulting system configurations.\n\nAs the foundation, Pecoraro's team created a graph database that represents the 600-plus features, with control nodes used to represent the complex grouping logic. Data nodes represent the data fields required for implementing a feature, and relationship links denote dependencies, validations and exclusions. Pecoraro says the resulting graph contains more than 1,000 nodes and 1,200 relationships of product configuration rules.\n\nThis graph is the digital twin. A custom-built "Physics Engine" translates the graph information into consumable and generalized services that exposes the basic relationships of decisions users need to make. The team also created a business service that exposes the nuances of the decisions in business context via an API for each of the roughly 30 services in the offering. When the operational user makes selections from the presented Plan Services Offer, the selections are saved via the API into a JSON document, which can then be validated against the underlying graph model.\n\nThe ICON project has dramatically reduced the amount of time and expertise required for client onboarding, Magro says. He says a new hire recently used ICON for an extremely complex offer and completed the job 70 percent faster than it would have taken a very experienced associate to do. Magro says it normally would have taken a new hire a year or more to reach that level of proficiency.\n\n"It was a very exciting result to say the least," Magro says. "It really took a lot of the complexity out of it."