We Used AI to Test Retirement Readiness. Here’s Where the Model Succeeded—and Failed.

I'm an hourly financial planner, and I'm always curious to see if free tools and AI can replace the expensive technology I use in my practice to answer complex questions. Today, I'm testing just that: using ChatGPT to figure out if a couple can retire.

Check out my video I did on this same topic.

To start, I found it's best to let the AI model prompt me first, as it knows what data it needs. I then fed it a detailed case study to see if it could handle the complexity of real-world planning.

The Case Study: The $3.3 Million Retirement Test

Here is the financial situation I presented to the AI expert:

• Age/Retirement Goal: I am 60, my spouse is 61, retiring in Spring 2027.

• Spending: $10,000 per month total ($5,000 core, $5,000 discretionary).

• Assets: $2 million in 401k/IRA, $1 million in Brokerage, and various cash accounts totaling $3.3 million.

• Income: Social Security, a non-COLA pension, and rent from a house with no debt.

• Major Goals: $1 million Colorado mountain house purchase and $100,000 in wedding expenses for two daughters.

Then, AI received the prompt: "Please run a retirement readiness analysis based on this information. Let me know if my plan looks sustainable, show me trade-offs, and suggest adjustments."

Where the AI Succeeded (The Starting Point)

ChatGPT’s initial structuring was impressive. It instantly broke the analysis into the "Big Picture," "Key Trade-offs," and "Actionable Next Steps," much like a human advisor would.

Key Takeaways ChatGPT Got Right:

• Solid Foundation: It correctly determined that, with $3.3M in assets and guaranteed income streams, the base plan is on "solid footing."

• Sustainable Withdrawal: It calculated that the pre-Social Security withdrawal rate of 2.6% is "very sustainable."

• Key Swing Factor: It correctly identified the $1 Million Colorado Mountain House as the single largest variable, noting it "requires giving up flexibility" and accepting higher ongoing costs.

• Roth Conversion Opportunity: It accurately highlighted that the period between age 62 and 70 (before RMDs and full SS) are prime Roth conversion years to smooth lifetime taxes.

• Social Security Delay: It correctly advised that delaying benefits until at least age 67 would meaningfully boost guaranteed lifetime income.

The Critical Gaps: Where the AI Model Failed

While ChatGPT provided a great foundation, its lack of depth and inability to handle specific, nuanced tax laws proved to be significant failures that could lead a DIY investor astray.

Failure 1: Missing the Tax Cliff Opportunity

ChatGPT failed a crucial tax test related to the Affordable Care Act (ACA). The original income assumption was high, but the AI didn't consider the significant impact of the ACA subsidy tax cliff at the end of 2025. This omission represents a huge missed opportunity to recommend strategic low-income years to maximize premium tax credits—something a human planner would flag immediately.

Failure 2: Lack of Context and Visualization

When asked to quantify the projection, ChatGPT struggled with basic output:

• The graph it generated used unclear scaling (e.g., "1 to e to the 7"), making the visual useless.

• It presented the sustainable annual spending as two unformatted bullet points, not a clear, context-rich table.

• It did not provide the underlying inflation or return assumptions it used for the projections.

Failure 3: No Stress Testing Capability

The most significant limitation appeared when I asked ChatGPT to run a stress test (e.g., modeling a retirement with only 3% or 4% market growth instead of 5%). The AI responded: "It seems like I can't do more advanced data analysis right now. Please try again later." Without the ability to run multiple scenarios, stress-test returns, or model inflation shocks, the final projection is little more than a static guess.

The Bottom Line: AI vs. Advisor

The free AI model proved to be a good starting point for a DIY investor looking to frame up their questions and get a quick confidence boost. However, it cannot replace the depth, precision, and understanding of complex, current tax legislation (like the ACA) that a human financial advisor provides. A human advisor's value lies in stress-testing and spotting critical tax cliffs that a simple algorithm overlooks.

At Barrett FP LLC, we offer expert financial planning on an hourly basis, focused entirely on helping you achieve your goals.

Learn more about how we can help you stress-test your plan's assumptions and see if we're a good fit.