Bonds are an important part of diversified investment portfolios. Their stable income stream and relatively low risk compared to equities make them an especially important component of pension and retirement planning.
Bonds are issued by various entities such as governments, municipalities and corporations. Each type of issuer offers different levels of risk and tax treatment. Issuer creditworthiness also affects bond risk and yield, with higher-rated bonds considered safer but with lower yields, while lower-rated bonds have higher yields and a correspondingly higher risk of default. pose a risk.
To protect employee benefits and protect against mismanagement of funds, the US government established the Employee Retirement Income Security Act (ERISA) in 1974. This is a federal law that protects individuals in private sector retirement and health plans. Employers and plan providers must comply with ERISA’s strict requirements, especially when it comes to investments selected for these plans, such as bonds. This can be a complex and resource-intensive process, often involving extensive manual audits and detailed risk assessment processes.
“Not every bond is required to comply with ERISA, but benefits administrators use ERISA to make sure bonds meet internal rules,” says Soundarapandian. A, Head of Data Science Practice at Hexaware. Bonds are required to be ERISA compliant in order to be purchased by institutional investors.
Verifying that bonds are ERISA compliant has historically been a manual process involving a thorough review of documents and legal history. Bond prospectuses can range from 100 to 400 pages in length and must be reviewed against an extensive checklist of ERISA requirements. Manually auditing a 401(k)-plan document can cost up to $7,500 and take about 45 hours of a skilled professional’s time. Yet despite the critical nature of compliance verification, there are a few automated solutions that streamline the process.
Generative AI is changing the equation, though. It has emerged as a transformative force in bond investment risk management by automating and optimizing complex analytical tasks. Gen AI models can sift through large amounts of unstructured data about bonds, identify key information, and compare it to risk parameters.
“It knows exactly where the compliance information is in a 300-page document,” says Soundrapandin. “It will retrieve relevant parts, check against a compliance checklist, and provide a recommendation in less than 20 seconds.”
Hexaware’s BondReco is an innovative application of Gen AI in this area. The software automatically classifies bonds into ERISA-eligible and non-eligible categories to ensure their investment safety. This can reduce audit fees by thousands of dollars. Using technologies like Azure Form Recognizer and Azure OpenAI, BondReco not only increases accuracy in digitizing data but also generates reports that validate investment classification.
BondReco is built on top of a highly secure enterprise-wide language model accessed through Microsoft’s Azure cloud. Hexaware did extensive fine-tuning to adapt the model to different industries and company types. “We have covered all sectors and vertical markets,” says Soundrapandian. “The application explains every decision and creates an audit trail on the back end.”
With BondReco, the cost of detecting an ERISA clause in a document can be reduced to approximately $500, demonstrating the financial and operational benefits of adopting AI-powered automation in the financial risk assessment landscape.
As powerful as AI is in reducing the time and cost of ensuring ERISA compliance, Hexaware takes an extra step to ensure quality. “We always have a human who checks the output,” says Soundrapandian. This should give bondholders an extra measure of confidence in automation.
To learn more about evaluating the safety of your ERISA bonds with BondReco, please visit the Microsoft Marketplace.