Artificial intelligence is here to stay. How far it will go is still up for debate. 

But at its core, AI is just a tool. And like any tool, it can be incredibly effective when used correctly—and surprisingly harmful when it’s not.

Let me start with a cautionary tale…

~~~

Story 1: When AI Refuses to Be Wrong

Last week, I found myself arguing with AI about Ohio 529 plans.

I knew that Ohio allows a state tax deduction for contributions to out-of-state 529 plans. Not in a “I think so” way. Not “I read it somewhere.” I knew it because it’s literally my job to know these things!

But while researching another state’s rules, I kept seeing Ohio show up on lists of states that don’t allow that deduction. Google kept pushing AI summaries and linking to questionable sources that were just plain wrong.

That stopped me. 

So, I tried different searches. Same result.

I went directly to ChatGPT and asked it point blank. 

It told me I was wrong. 

I asked for more sources. The sources supported the same incorrect conclusion. 

So I pushed back—and this is where it got interesting.

ChatGPT didn’t budge. It didn’t say, “This might depend,” or “You may want to verify.” It doubled down.

Confident. Clear. Wrong.

Eventually, after I kept pressing and provided better sources, it corrected itself—but not before arguing the wrong position with surprising conviction.

That’s the risk.

AI doesn’t just get things wrong occasionally or with uncertainty. It can get things wrong confidently. And if you don’t already have a baseline understanding of the topic, you’d have no reason to question it.

What really caught me off guard was that I told ChatGPT multiple times it was wrong—and it refuted me. Normally, it feels like a “yes-man.” It supports what I write, tells me I’m on the right track, even compliments my style… blah blah blah.

But here? Nope. ChatGPT was just like…. “Nah… you’re wrong.”  Meanwhile, it was actually the entity that was unequivocally incorrect. 

So what would have happened if someone (me) had taken ChatGPT at face value? Well, I probably would tell a client with an American Funds 529 plan that this year’s contributions weren’t deductible. They wouldn’t have reported it on their state tax filing and they would probably miss out on $125 of tax savings.

A big deal? No.

But what would have happened when the client eventually realized I was wrong?

The loss of trust would amount to a lot more than $125 to me. 

Story 2: The Day AI Told My Client Not to Retire

Okay… one more cautionary story.

A client of mine was preparing to retire and we had everything mapped out. The final step was figuring out health insurance. Since it was nearing the end of the year, we decided to use COBRA coverage to avoid resetting a deductible and to reduce risk for the remainder of the year. The plan was solid, and everyone felt good about it.

Everything was moving forward—or so I thought.

A few weeks later, I got a call filled with concern. The client was questioning whether retiring now would be a mistake… “now that COBRA isn’t an option.”

That caught me off guard.

“What do you mean COBRA isn’t an option?”

“Well… I read that it wouldn’t be available to me.”

Now I’m thinking I missed something. Did I make an error? “Did your company send you something? How did you discover this?”

After some prodding, the client finally revealed that ChatGPT had told them.

Based on my research, knowledge of their employer’s plan, and the documentation we had already reviewed, I was able to confidently reassure them that COBRA was indeed an option. Their plan was still intact. There was no reason for concern and health insurance wouldn’t be the bane of their existence for the next few months.

But here’s the problem… it caused them days of unnecessary distress over their decision before they discussed it with me.

~~~

Some areas where I would use AI in your finances?

1. Breaking down complex topics into plain English
This is where AI really shines. It can take complicated financial jargon and make it more understandable, helping you get your bearings on a topic.

2. Asking questions and “what if” scenarios
Sometimes the hardest part is knowing what to ask. So explain your financial situation (without sharing personal details) and ask what you should be researching or considering in your position. It will likely bring up topics you hadn’t thought of yet, or questions you should run by your advisor.

3. Formula-based calculations
There are a lot of calculations that AI software can run for you. Loan amortizations, basic tax questions, time value considerations, debt to income ratios when you are refinancing – It’s really endless. AI is usually very good at math, especially when it’s working on calculations based on proven formulas.

Here are a few areas where I would not use AI for:

1. Making final investment decisions
AI can help you learn, but it shouldn’t be the one implementing your portfolio. Portfolio construction is nuanced to each individual. At the very least, get a professional second opinion.

2. Implementing tax strategies
Again, nuanced. Many AI tools also use outdated sources and tax tables, so if you did try to use AI to implement a tax strategy, I would recommend using one of the upgraded, paid tools.

3. Making final decisions on insurance or estate planning
All of these concepts are very unique to individuals and within a landscape of changing laws. Professionals are necessary. Bring your questions directly to a human.

4. Anything requiring accountability
AI has no accountability. If you use AI to write your last will and testament, and it gets it horribly wrong… there is no accountability. You are stuck.  

~~~

Overall, ask your questions. Explore new topics. But, just like with most people you meet, approach AI tools with a level of suspicion. Be aware that confidence does not equal accuracy. If you see something contradictory, do your own research or ask professionals!