AI Chatbots Aid Basic Retirement Planning but Fall Short on Complex Advice, Experts Say

People use artificial intelligence for tasks from work presentations and shopping to scientific research. But can AI help workers with the toughest calculation they face: whether they can afford to retire?

Americans already turn to AI for financial advice. About 20% say they use chatbots for this purpose, according to a September study by AI company Pearl. About half of those who use AI at work also apply it to retirement planning, double the rate among workers who don't use AI, MissionSquare Research Institute found in a separate study.

The need for retirement advice is acute. Americans now expect to work four years longer than they would like because of rising living costs and inadequate savings. The median balance for workers with retirement plans stands at $40,000, well below the $1.5 million they say they need to retire comfortably.

Social Security, the financial backstop millions count on in retirement, could cut monthly benefits by as much as 20% in six years unless lawmakers act to shore it up.

Given those facts, people might ask ChatGPT or Claude, "Here's what I have saved so far. Will I be able to retire at 65?"

Some experts say AI provides a solid starting point for basic retirement questions. "I'll say, 'Come up with some financial planning ideas or even run a Monte Carlo simulation to see how much I can spend every year,' and it might not be perfect yet, but it's starting to be able to get to a place where it's producing some pretty valuable output that I think will be beneficial to people," said Luke Delorme, director of financial planning and a Certified Financial Planner at Tableau Wealth in Great Barrington, Massachusetts.

A Monte Carlo simulation is a mathematical model that runs thousands of potential outcomes for a retirement portfolio. It factors in best- and worst-case scenarios, such as a bear market's impact, and projects the odds that savings will last a person's life.

"Such simulations are the perfect thing for a computer program to do. Eventually, I think that those tools will also become pretty powerful," Delorme told CBS News.

Yet generative AI has limits for complex retirement issues, from tax impacts to longevity risk. Boston University economist and retirement expert Laurence Kotlikoff told CBS News that AI may do more harm than good in giving retirement advice. The apps struggle with Social Security nuances and other issues, and they rely on what he calls flawed traditional financial planning advice.

"It's being trained on Wall Street's guidance, and Wall Street's guidance is all about maintaining and collecting and expanding its assets under management, so that has nothing to do with proper economic-based advice," said Kotlikoff, who developed a retirement planning tool called MaxiFi.

AI estimates retirement savings based on expected longevity, a common planner framework. But planning should use maximum life expectancy to avoid outliving money, Kotlikoff said. He has found AI often gives wrong information on Social Security scenarios, which involve 22,000 pages of rules.

"Then you are off to the races of having the wrong analysis done for you," Kotlikoff said. AI "is like the hottest new thing — you can't criticize it because otherwise you don't sound cool or you are defending your job or company."

But "I don't give a s*** about feeling cool — I'm here to make people feel safe," he added.

Andrew Lo, a finance professor at the MIT Sloan School of Management, told an MIT publication in April that AI struggles with tax optimization and regulatory nuance. Unlike human advisers, it faces no legal duty to act in a client's best interest.

Lo said users should ask AI to list where it might be wrong, along with its assumptions and uncertainties.

CBS News tested this with a hypothetical 50-year-old single woman earning $70,000 a year. She has median retirement savings of about $185,000 for her age, mostly in S&P 500 index funds. She contributes 12% of her income to retirement and expects $2,400 a month in Social Security at full retirement age 67.

Anthropic's Claude, OpenAI's ChatGPT and Perplexity assessed if she can retire comfortably at 65. Claude and ChatGPT said yes, but it would be tight, with risks of running out of money. Perplexity was more negative, saying she likely cannot without cutting spending or boosting income.

The chatbots based models on her living to 90, not a maximum of 100, and skipped exact tax impacts. They also ignored long-term care costs. Claude adjusted its view after noting the short planning horizon, shifting from "a tight but doable retirement" to "meaningfully underfunded without course correction."

A larger issue in retirement planning is fear of investing, which leads to errors like holding cash or CDs with returns below inflation, Luke Delorme noted. Savings erode over time, raising the risk of running short in retirement.

Delorme thinks AI could help the two-thirds of Americans without financial planners grasp these ideas. But he doubts AI alone can beat behavioral hurdles.

"It's much more behavioral than it is a technical lack of knowledge," Delorme said. "I don't know if today that's going to help people overcome their fears of things, like the fear of investing, which is such a huge obstacle."