No, this isn’t about how AI is going to kill us all in the coming years (by the way, didn’t that story come and go from the news cycle quickly!?). It’s about the much more current problem of getting the wrong answer and taking it as gospel truth. AI can make mistakes. Chat GPT comes with that warning and I am told the other flavors of AI do as well. If you’re outside your area of expertise you can easily be led astray.
A former client made a tax based decision several months ago which turned a $45,000 tax deductible expense into a $52,000 after tax cash outflow. That’s roughly $22,000 out the door. When I learned about and questioned the decision, she told me she got it from AI and even asked their tax CPA. I called the CPA who I have known for years and trust. The CPA had answered the question she was asked correctly; but like AI she was also asked the wrong question and wasn’t given all the facts. And the former client couldn’t be convinced she got the wrong answer. After all, it came from AI.
In another example, I returned from a trip and had an email waiting from a client explaining several approaches to an investment transaction along with the statement that he ran it through AI and the conclusion of which approach he was taking. It was Sunday so I dashed off a quick “Maybe I don’t fully understand so let’s discuss tomorrow”. That was a white lie. I did fully understand; I wanted to finish my weekend. The AI decision was about to cost him $165,000. Fortunately, this client quickly grasped the issue and changed course.
One of the accounting firms a client uses for monthly bookkeeping is extensively leveraging AI. No issue with that. But I was questioning some amounts in a general ledger account and specifically asked for their thoughts. What came back was “Thank you! All good!” I responded with “?????????????” and then immediately got “Wrong reply! More Coffee.” Fine, I blame my mistakes on a lack of coffee as well but later I got an acknowledgement that it was an AI issue.
An accounting manager at one of my clients sent me a very well-crafted email regarding a GAAP issue. Well-crafted by AI and dead wrong. Let’s just say that was not a good look.
I recently had AI pull together a balance sheet out of raw GL data. The only problem was it didn’t balance! For me, that’s an easy mistake to catch. Others might not notice it.
If you’re using AI in your field of expertise where you have the ability to check work, it’s a valuable tool. It has lots of uses with more coming by the day. But if you’re relying on it to replace an expert, you ought to be very careful.
I attended a recent presentation by Ross Hartman at the Vistage St. Louis Executive Summit. Ross made the point that compared to a personal relationship, AI lacks firsthand experience, accountability and trust. That’s worth thinking about before you blindly accept an AI answer.
I’m not embarking on this AI journey alone. I’m relying on long-time advisor turned friend Chad Barr for guidance and advice to help me get more out of it and not be led astray. And for anything complex, I improve the quality of my prompts by using Pro Prompt Genius developed by Chad. Don’t let AI cost you money, or worse, your credibility.
If your business could benefit from fractional CFO services, I would welcome the chance to speak with you. Please give me a call at (513) 225-8657 or send an email to [email protected]
The archive of these monthly newsletters is posted at the Resources section of homza.com
your cash is flowing. know where.®
Ken Homza
Copyright @ 2026 Kenneth M. Homza