People are turning to artificial intelligence (AI) for lots of tasks—think meal planning, budgeting, and even interior design. But can tools like ChatGPT really handle something as critical as financial planning? I decided to test this by asking ChatGPT to create a retirement plan, focused on a retiree aiming for an annual income of $100,000. The scenario involved a 40-year-old who plans to retire at 65 in a fully paid-off home in Tennessee, earning $125,000 a year. After setting these parameters and asking if the plan was realistic, I sought an expert’s opinion from Eric Franklin, a Certified Financial Planner at Prospero Wealth.
Insights from a Financial Planner
When I reached out to Franklin, he was intrigued—he’d never before critiqued an AI plan. With over 23 years in technology, he seemed like the right person for the job.
His first reaction was more about curiosity than expectation. “I had no idea what I was going to get,” he admitted.
The results were, well, surprising. It wasn’t that ChatGPT’s output was blatantly awful or that I had significant financial resources; rather, it fell into a vague area where things feel a bit off, much like the concept of the uncanny valley in art.
Initially, Franklin thought the output seemed plausible. “It passed the sniff test pretty well,” he noted, reflecting on how it looked realistic at first glance. But he quickly recognized that the plan I presented covered only a small fraction of what a comprehensive financial advisor would usually consider—like stress tests and the potential impact of life changes.
As he examined the details, it became clear that using ChatGPT as a retirement planning tool could be risky for someone without a solid financial background. It made several incorrect assumptions, overlooked inflation, and created scenarios that were far from realistic. This guided a unique consideration: too many people might be tempted to rely on AI for retirement planning, thinking it could provide straightforward answers with minimal input.
“That calculation is literally wrong,” Franklin remarked, as he shared the plan with his team, and they unanimously agreed.
What Did ChatGPT Get Right and Wrong?
- Age: 40
- Job: reasonable cost of living
- Income: $125,000
- Retirement Age: 65
- Timeframe: 25 years
- Annual Retirement Spending: $100,000 (in today’s dollars)
- Emergency Fund: Done
- Home Ownership: No plans for downsizing modeled
- Current Retirement Savings: $100,000 in 401(k), projected 6% growth
The 4% Rule
ChatGPT suggested that a future retiree would need $2.5 million by age 65, based on the 4% rule, which would enable annual withdrawals of $100,000 over 30 years.
Expert Critique
Franklin pointed out that most experts don’t favor the 4% rule for calculating retirement savings needs, raising a red flag. He explained that its foundation rests on data that’s not particularly reliable in today’s context. “In many real-world cases, it turns out retirees often don’t even use what they saved, leaving behind a substantial amount,” he said, stressing the need for a more nuanced approach to retirement spending.
The $2.5 million figure ChatGPT generated also ignored inflation, leading Franklin to explain that what seems like a hefty sum today would lose its value over time. “In 25 years, the purchasing power of what’s now $100,000 will be significantly less,” he added, opening up further complications in the potential plan.
Missing Components
ChatGPT proposed an investment strategy with an 80/20 split of stocks and bonds at age 40, shifting towards a more conservative 60/40 by retirement. Franklin estimated that the initial $100,000 savings could realistically grow to about $430,000 by retirement, aligning closely with AI’s predictions for new contributions.
However, the AI’s suggested saving rate—approximately 21% of income—was not only ambitious but inaccurate, resulting in a significant shortfall. “If you do the math, you end up about $500,000 short,” he stated, noting that the calculations were unexpectedly off.
Simplistic errors like these can lead to major confusion for users who think they can rely solely on AI for this kind of planning. “It was a bit startling how simple tasks weren’t accurately calculated,” Franklin said.
Final Thoughts
“Following this plan could really put someone at a disadvantage, both mathematically and psychologically,” Franklin warned. Most individuals face changing circumstances that don’t fit into a rigid plan. Additionally, ChatGPT’s output failed to consider Social Security income, offering only vague references to health care costs being offset by future withdrawal strategies.
In sum, while AI can produce seemingly sensible plans, it lacks the subtlety, experience, and empathy a human financial planner would bring to the table. “There’s little room for flexibility,” Franklin concluded, doubting that anyone at an income level of $125,000 could comfortably put away such a substantial portion of their earnings for decades.
