The advent of generative artificial intelligence (AI) technology, such as ChatGPT, promises to transform the way businesses operate across sectors. Yet as quickly as novel applications and abilities surface, so does more evidence of the limitations and shortcomings of large language models. Plan sponsors can harness the immense power and utility of generative AI to provide better support to plan participants, but they must do so only after putting prudent processes and safeguards in place.
Identify Appropriate Applications
ChatGPT could be used to help streamline the retirement plan onboarding process by giving employees detailed instructions for how to enroll. Sponsors could use it to help provide information about plan options, as well as answers to employees’ frequently asked questions regarding their retirement plan benefit. The technology could also help craft customized messaging around more routine tasks such as:
However, at present it is advised that such technology be limited in use to non-legally required communications. Because ChatGPT is only as effective as the directives it receives it is possible it may not generate a complete document meeting legal requirements. Thus, at present it is advisable to avoid using this technology for legally required disclosures, notices or legal documents.
Transparency, Security and Oversight
While the opportunities for gains in efficiency are significant, it’s critical to address potential concerns when implementing nascent generative AI systems into your plan.
Training. Employers must establish clear guidelines about which employees can access ChatGPT — and for what purposes. Appropriate training regarding the limitations of the technology should be given to those authorized to use the platform. Workers should be instructed how to handle complex or sensitive issues that require human intervention.
Data integrity and security. At all times, participants' personal and financial information must be protected by strict adherence to privacy standards. Establish clear protocols for sensitive data handling and retention. Implement and document oversight processes to protect information from unauthorized access, disclosure or misuse.
Compliance. Designated individuals should review and approve all generative AI outputs for compliance with regulatory and internal requirements before they’re communicated. Their review should include ensuring that all information provided to participants is accurate and consistent with plan provisions. Audit generative machine learning systems regularly to identify potential issues or vulnerabilities and take any necessary corrective action.
Transparency. Employers should be transparent with workers regarding the use of ChatGPT with respect to their retirement benefit. This includes how their personal information is being used — and protected.
A Prudent Path Forward
Plan sponsors should regularly review and evaluate generative AI integrations to flag any potential issues and identify ways to optimize their use of the technology, while ensuring they remain compliant with all pertinent industry regulations and standards. With guardrails in place, and adherence to prudent processes, appropriate use of AI-powered generative tools in retirement plan management presents an opportunity for sponsors to increase efficiency and enhance outcomes for plan participants.