Should You Use AI for Retirement Planning?

AI can be useful in financial planning. Here’s what to know to leverage it and what to beware of in using AI for retirement planning.

Consumer use of artificial intelligence is expanding, and financial planning is no exception. More people are turning to large language models (LLM) like ChatGPT or Claude AI to help them grapple with questions about their financial futures. But is the strategy a reliable substitute for the guidance of a human fiduciary advisor? Should you use AI for retirement planning?

“Yes and no,” said MIT Laboratory for Financial Engineering director, Andrew Lo. While the latest versions of AI chatbots are great at some elements of planning, they’re terrible at others. There are also some serious risks. “If ChatGPT gives you bad advice,” Lo offered as an example, “there is no actual fiduciary duty in the sense of bearing consequences. So you’ve got to take their advice with a pound of salt.”

AI for retirement planning

Here we take a look at the pros and cons of using AI for retirement planning to help you understand how to best leverage the technology to meet your needs.

Where AI fails

LLMs are great for a lot of things. They can put together a wealth of research around pretty much any subject within a few blinks of the eye. But they’re far from perfect, and they often make mistakes.

“Sources such as ChatGPT and Copilot have proven to be valuable tools, [but] they have also shown to provide misleading, fragmented, or even false information when it comes to personal finance recommendations,” said Alan E. Becker, senior investment advisor and RSG Investments CEO. “Creating an accurate retirement strategy based on your unique goals is one place where [AI’s] possibilities seem lacking.”

According to personal finance site Credit Karma, 66% of Americans who have used generative AI also asked it for financial advice. Becker and Lo say there are some major risks to be aware of.

1. Legal responsibility.

Laws require human advisors to follow regulations around conflicts of interest and fairness and to act in their clients’ best interest. Neither LLMs nor the companies that created them face such restrictions.

“If ChatGPT gives you bad advice, if you end up getting traded ahead or front-run by one of these large language models, they will not go to prison,” Lo told MIT. “There is no actual fiduciary duty in the sense of bearing consequences.”

2. Inaccuracy

LLMs pull from a vast library of materials, so answers sometimes rely on outdated or biased sources. AI also needs informative, well-written user prompts to guide its analysis. That need can lead to interpretive gaps around subjects like tax optimization or regulatory nuance.

“A recent UK study by consumer group Which? asked numerous AI tools for investing advice and found some offered tips that would specifically breach tax law,” Becker said. “Researchers also found that if they added a flaw to their question or prompt, AI rarely caught it and would offer advice regardless.

Lo’s short-list of topics to avoid includes: Medicare tax optimization, when to begin Social Security, state-specific legal advice, actuarial precise calculations, and trying to understand real-time regulatory changes.

3. Unrealistic goals

Talking to a chatbot about finances can often lead to outrageous – if not impossible – suggestions. Ask about potential early retirement, for instance, and LLMs may advise you to boost your income by $50,000 a year or sell your home and move to a state with a cheaper cost of living. The former is a pipe dream for most people, especially those above age 55. The latter fails to evaluate social factors like long-time friendships, relationship to place, or proximity to family members.

“Where these programs truly fall short is in the objective versus the subjective,” said Becker. “They can offer objective facts (when they get them right), whereas a financial adviser can offer subjective insights.”

4. Privacy concerns

“Never share sensitive personal information with a chatbot,” wrote financial advisor Jill Schlesinger, including Social Security number, driver’s license information, passwords, account numbers, and medical records.

What AI does well

It’s tempting to pull up an LLM and type something like, “I have x amount of money saved so far. Will I be able to retire at 65?” But that approach will yield murky, incomplete results at best. Experts like Lo say AI’s true power is unlocked when you ask well-defined questions with clear frameworks.

“For DIY investors, chatbots like Claude or ChatGPT are useful tools for researching financial products, tax rules, and [research] harder questions like when to claim Social Security,” Tableaux Wealth director of financial planning, Luke Delorme, told Boston College’s Center for Retirement Research. They also show great potential for “combining and analyzing separate data sources to gain new insights, such as identifying opportunities and vulnerabilities.”

Lo points to a few areas as sweet spots for implementation.

Explaining the logic of diversification. AI can help you understand why it’s important to do things like balance your assets within a variety of categories like stocks, bonds, and cash holdings to smooth out volatility and protect savings from catastrophic losses.
Behavioral coaching. LLMs are good at assessing saving and spending habits, and helping fine-tune your budget. They can also provide useful insight around the value of stock investments versus, say, putting all your money in CDs.
Scenario exploration / stress test assumptions. Use AI to model best, worst, and average outcomes related to decisions or situations like early retirement, market downturns, long-term inflation, or unexpected healthcare costs. The information can help you get a better grasp of how to weather potential financial crises and plan ahead.
Summarizing regulations and tax rules. Should you retire at 65 but postpone applying for social security until 70? Which retirement accounts should you withdraw from first and why? LLMs can provide valuable insight into research-based questions like these.

Technology can also help you better understand the possible costs of long-term care (LTC) later in life. New AI-powered platform, Waterlily, for instance, compares an individual’s information with similar data from 50,000 families and draws from a dataset with over 500 million data points to provide personalized LTC predictions. Answer a brief questionnaire and the LLM does the rest. Reports include year by year analysis of changing healthcare needs, projected costs, and suggestions like savings or investment strategies to help pay for LTC insurance.

Overall, Delorme and Lo say it’s important to always double- and triple-check AI responses. They also recommend that you ask LLMs to reveal any assumptions about the information they provided and analyze it for potential blind spots. Use the tool to hone your financial understanding, goals, and have better, more informed conversations with your financial advisor.

“Early signs suggest [AI] will expand what’s possible,” said Delorme. “That’s a reason for optimism on behalf of retirement savers. Those who work with advisors may find those relationships becoming richer and more proactive.”

“AI can help educate you,” Schlesinger wrote. “It shouldn’t be a substitute for judgment, a financial plan, or a human advisor who actually knows your life.”