5 Steps To Help Retirement Advisors Compliantly Integrate AI Usage Into Their Practices

Advisor Perspectives welcomes guest contributions. The views presented here do not necessarily represent those of Advisor Perspectives.

In the three-plus years since the launch of ChatGPT, artificial intelligence (AI) has quickly evolved from novel party trick to critical business tool. Workers in the professional services sector, in particular, now rely on AI for a broad range of tasks, such as taking notes, synthesizing documents, meeting prep, and composing emails.

According to a 2025 Thomson Reuters Report, 41% of professional services workers are using generative-AI tools such as ChatGPT, and another 17% are using industry-specific tools. But highly regulated industries must tread carefully.

For retirement advisors, the Employee Retirement Income Security Act (ERISA) — enforced by the Department of Labor (DOL) — does not prescribe specific rules for the use of AI. Instead, it establishes a broader fiduciary framework that governs prudence and loyalty, which advisors must apply when evaluating and using AI in their practices to remain compliant.

Here are five best practices to help guide fiduciaries and ensure that they’re benefiting from these tools without running afoul of regulations.

1. Human Accountability Is Paramount

While AI can help advisors run their practices more efficiently, fiduciaries cannot use it to substitute for their own decision-making.

For example, advisors shouldn’t use ChatGPT to recommend the investment lineup to the plan committee. AI can inform and support analysis, but ultimately, recommendations should come from the advisor.

The DOL (and plaintiffs’ attorneys) will hold the firm or advisor accountable, not the technology. For advisors using these technologies, the Securities and Exchange Commission (SEC) is also looking to see if the use of AI is properly disclosed via agreements and Form ADV.

When using these tools, advisors must provide adequate oversight. For example, notetakers are great for productivity, helping with client follow-up, and eliminating busywork. However, advisors must play an active role in managing the process by reviewing sample meeting minutes from committee notes and minutes.

2. Understand and Assess the Source Data for the LLM

Using a public database such as ChatGPT as a large language model (LLM) is very different from using an AI tool that relies on a proprietary data set such as Morningstar, Bloomberg, or a portfolio analytics tool developed by an advisor’s firm.

When relying on AI, fiduciaries should know which kind of tool they are using and, whenever possible, prioritize LLMs with access to more current and reliable information that is closed and proprietary to the firm.

3. Evaluate Model Integrity and Currency

Along with source data, AI rests on the strength of how the platform is tested, updated, and prompted, including the frequency and quality of new data inputs.

Fiduciaries must have a robust understanding of all these factors to ensure that they’re accessing the most accurate and up-to-date information available.

4. Monitor Risks, Bias, and Controls

Part of using AI responsibly is being cognizant of error rates and potential bias within the model. In the professional services sector, benchmarks have shown that hallucination risk is usually around 1 to 5% in a tightly controlled workflow with approved sources. This rises to over 10 to 30% when the prompt asks for the AI tool to answer from memory, create citations, or respond to niche facts.

AI can make mistakes due to bad data or inaccurate prompts. For example, an advisor might query for fee benchmarking against a peer group. The AI tool might respond that, based on available information, the recordkeeper’s fees are reasonable. However, the peer group might be skewed toward larger plans or fail to consider average account balance, leading to incorrect responses.

5. Ensure Strong Governance & Oversight

As AI continues to evolve, fiduciaries, including advisors, consultants, and their firms, must maintain clear internal policies, controls, and accountability frameworks governing its use.

Since some AI tools draw from open sources and then store inquiries and results in unsecured environments, it’s important to consider potential cybersecurity risks and to ensure alignment with the DOL’s cybersecurity guidance.

Though AI can pose potential cybersecurity threats, it can also be used to help combat fraud. Some service providers are using these tools to monitor for normal behavior and provide safeguards if they detect activity that diverges from typical patterns — such as how credit card companies will put a freeze on unusual or large credit card purchases.

Balancing AI’s Benefits with Fiduciary Duty

As AI becomes more omnipresent, fiduciaries must not only be mindful of how they adopt these tools in their own practices; they must also carefully monitor how their vendors use and store sensitive client information.

Even as we await formal DOL guidance related to AI usage, the DOL has issued corollary guidance to fiduciaries, including cybersecurity guidance, so it’s important for advisors to be proactive in their approach to these tools.

We’ve all seen how AI tools can boost our productivity and efficiency but, like most things in life, the benefits must be weighed against potential risks. If advisors want to use these tools responsibly to support their businesses and clients, they must do so thoughtfully and with sufficient oversight — trust, but verify.

Bonnie Treichel, Esq. is the founder of Endeavor Law and the founder and chief solutions officer of Endeavor Retirement, a consulting firm dedicated to solving problems for plan sponsors, advisors, and service providers in the retirement plan industry.

A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.