When AI Says No: A Wake-Up Call for Business Owners

What happens when an AI tool stops acting like a tool — and starts making decisions for you?

In this episode of How to Win Big at Business, Joe DiChiara shares a personal experience with Claude AI that started as a simple project: tracking calories, exercise, weight loss, and daily progress. Joe was using AI the way many self-employed professionals are beginning to use it — as a helper, organizer, calculator, and productivity tool.

But then something unexpected happened.

After days of tracking meals, exercise, calorie intake, and weight loss goals, the AI program decided it would no longer continue calculating Joe’s daily calorie deficit. Instead of simply producing the numbers requested, it stepped back, expressed concern, and recommended a different direction.

That moment raised a much bigger question:

What happens when the technology we pay for decides what it will and will not do for us?

Joe connects this experience to a much larger issue facing self-employed professionals and business owners: AI is powerful, useful, and here to stay — but it also comes with risks. If you are using AI to run your business, create content, organize client data, produce documents, analyze numbers, or manage operations, you need to think carefully about what you are sharing, where that data is going, and how much trust you are placing in these tools.

This episode is not anti-AI. Joe uses AI every day and believes it can be a major advantage for entrepreneurs. But he also believes business owners need to stay alert, protect sensitive information, and remember that AI should support human judgment — not replace it.

In this episode, Joe discusses:

• Why he started using Claude AI
• How AI helped him track health, calories, exercise, and weight loss
• The surprising moment the AI refused to continue the calculation
• Why AI guardrails can create unexpected business concerns
• Why self-employed professionals must be careful with personal and business data
• The danger of relying too heavily on tools we do not fully control
• What this experience means for entrepreneurs, consultants, accountants, coaches, and freelancers
• Why AI is powerful — but still needs human oversight

If you are self-employed and using AI in your business, this episode is an important reminder: AI can help you move faster, but you still need to protect your judgment, your data, and your business.

If you’re a self-employed professional trying to build a real business, this podcast was created for you.
 New episodes break down the systems, strategies, and structures needed to move from solopreneur to sustainable business owner.
 Subscribe to our YouTube Channel
 https://www.youtube.com/@Bedrock360BusinessSolutions/featured
Thanks for listening!
If you want fewer headaches, fewer IRS surprises, and more smart systems running your business (so YOU don’t have to), let’s talk.

Book a strategy call: www.timewithjoe.com
Because running a business shouldn’t feel like wrestling an octopus… alone.
Also check out:
 www.Bedrock360BusinessSolutions.com (Complete Coverage)
 www.Bedrock360TaxSolutions.com (Tax Guard 365)
 www.Bedrock360BusinessTraining.com (Courses)

This episode includes a personal story involving calorie tracking and weight loss. Nothing in this episode should be taken as medical advice. Always consult a qualified health professional before making decisions about diet, exercise, or health.

Thanks for listening!

If you want fewer headaches, fewer IRS surprises, and more smart systems running your business (so YOU don’t have to), let’s talk.

Book a strategy call: www.timewithjoe.com

Because running a business shouldn’t feel like wrestling an octopus… alone. 

Also check out:
www.Bedrock360BusinessSolutions.com (Complete Coverage)

www.Bedrock360TaxSolutions.com (Tax Gaurd 365) 

www.Bedrock360BusinessTraining.com (Courses)

Transcript

SPEAKER_00 0:04

So I'm going to ask you to have an open mind here for a couple of minutes. And just play along with me, and you'll understand where I'm going with this by the end of this episode. And I think that you're going to find this pretty interesting. How would you feel if you turned on the TV and you go to Netflix, which you do all the time because you're pretty much addicted to Netflix? And the TV gives you a message that says, Hi Joe, unfortunately, I'm going to be pausing your access to Netflix because I feel like you're watching way too many horror shows. And I'm going to give you a number to call because you might be having some issues. Now that might sound far-fetched to you. How about this one? You turn on the stove and you start put putting cooking oil in a frying pan. And then the stove says to you, Hi Joe, I'm sorry. But for the time being, I'm not going to allow you to fry any food because you've been eating a lot of fried food lately. And as well as I know that fried food is not good for your stomach. So we're going to pause this and I'm going to recommend that you go see a gastrointentestinologist or whatever the people are, the doctors that check your stomach. Does that sound far-fetched to you? Because I'm going to share something with you that I've been going through with one of my AI programs, Claude, which is owned by a company called Anthropic. My name is Joe DeCara. I am a CPA for over 40 years, lifelong entrepreneur. And I created this podcast, How to Win Big at Business, specifically for self-employed professionals that are struggling or facing all the normal obstacles that small business owners face. And to provide solutions and ways to navigate those obstacles, including, but not limited to issues with the IRS, state and local governments, marketing, creating a scalable business and everything and anything in between. I've had the opportunity to work with thousands, literally thousands of individuals and small business owners. And like I said in the beginning, I'm an entrepreneur just like you. So I want to share the recent experience that I've had with Claude. I started using it in September of 24, where and I was using it before, but it was more like a Google search tool until somebody, until my business coach, Larry Broughton, introduced me to the act as if method. And because one of my mentors, one of my partners, showed me that it produced amazing spreadsheets and Word documents, color, and it just did stuff that the other AIs weren't doing at the time. My main AI program has been Chat GPT, that's the granddaddy of all of them. I did try Claude last year. It didn't suit my taste. It seemed like it was more for coders. But like I said, a couple of weeks ago, I started using it. I like it because it, like I said, it does stuff that the others don't do. For instance, it opens very easily, it's got a desktop program. It is it was just very easy and simple to use. So I started using it for some personal things. And I'm on a on a uh a mission to a lifelong mission for me and a lot of other people to get control of my weight, my weight, my health. I started doing Tai Che. And I said, what? I don't want just want to lose the weight. I want to balance it. I'm gonna incorporate exercise. I started swimming, like I said, Tai Che. I do a lot of walking, I do at least 5,000 steps a day. And I said, I'm I'm gonna set up a project in Claude to keep track of my calories. Now I know there's a lot of programs out there that that'll do it. I had a program once that you actually take a picture of what you're eating, it tells you what it is, gives you all the information on it. But I said, I'm already paying for these AI programs. Why am I gonna pay for another calorie counting program? So I fed it all the data, my age, my weight, my height. I'm in really good shape. I'm gonna be 66 on June 22nd. This is 2026. I was born in 1960. So I'm feeding it all the data. And I started this, let's say about two weeks ago. And also I asked it, what's the normal calorie intake? Like if I'm counting calories, what's the break-even? Like if I ate, and I think the number came out to about, let's say, 2500 calories. So if I consume 2500 calories, I'm not going to gain weight, I'm not going to lose weight. Then I added, well, if I do 20 minutes of Tai Che, how many calories does that knock it off? Walking, swimming, and I pretty much got it down to okay, if I do two to four laps in the pool a day, a thousand steps, and 15 to 20 minutes in Tai Che, that gives me another 300 calories. So let's say it's about 2800 calories. So then I asked it, so how many calories is a the recommended amount to start losing? So now it introduces what they call the calorie deficit. And I think it said at my age, about 100 calories, and then I would lose, let's say, a pound a week. And and well, I want to lose more than that. Let's say I want to lose two pounds a week. So and I'm asking Claude for advice on this, and what it's what he said was so I'm looking to do a 1800, 15 to 1800 calories a day. So let's say I want to take a day off from exercising. So now if I do 1500 calories, I'm gonna have a thousand calorie deficit. So what Claude tells me is that's not recommended because you might lose the weight by basically starving yourself. Uh, and what happens is your body starts eating not the fat, but it starts eating at your muscle. And I said, Well, isn't my exercise building up my muscles? And Claude said, Yes. In fact, that's an excellent way to approach it because now you're building muscle, and it so basically Claude gave me this is a good idea. You're balancing, you're eating healthy, I'm drinking a lot of water. I went from drinking uh regular coffee, which I found out was 90 calories. But if I drink black coffee, it's actually zero calories. So yeah, I'm approaching this in an intelligent way, I believe. And then two days ago, I get a shock. Now, also let me mention that Claude kept making like stupid mistakes, like getting the dates wrong. Like it's telling me it's it's Saturday when it's Friday, it's telling me I didn't log in any meals for Friday when I logged in all the meal, and it was telling me, okay, a quarter of a chicken is 250 calories. The next day, a quarter of a chicken is 320 calories, so I had to keep correcting it. The dates, the calories, and I thought that was odd, and it keeps apologizing. You're right. Um I apologize, and so this is what happens. So it's almost like it becomes a cheerleader, it's coaching me, and that's fine, that that's good, but we're talking about a computer program, so it's good, it's giving me encouragement, it's keeping, and I'm doing unbelievable. I think in two weeks I lost two kilograms. I went from 89 kilograms to about 87 in pounds. That's about like I was like at 198 when I started. Now I'm at like 194, uh, 195 in between. So it's working great. And then two days ago, and I'm gonna read this because this has been an ongoing long thread. So Monday consumption, I open it up. So Claude tells me, I need today's food entries to calculate. You haven't listed any consumption for Monday yet, just the header. Because when I wake up, what I do is I weigh myself and then I put in the title, Monday consumption. So for each day, I have a thread. So if I want to go back to previous days, I can just see the title. So, okay. Also, a quick flag: your tracking week runs Friday through Thursday. So this was another thing. So I started on a Friday, so I wanted the week to go from Friday to Thursday. For some reason, it kept changing the week, it changed it to Thursday to Wednesday. So I had to keep reminding it, no, it's from Friday to Thursday. Very strange because I so the point is these AI programs aren't perfect, they make mistakes, but this is like the same mistakes over and over again, anyway. So it tells me, so again, now it says you've asked for Thursday through Wednesday here. Do you want me to know? Okay, so I just made a mistake. It was from Thursday to Wednesday, and it was giving me Friday to Thursday. So here we're almost in uh almost two weeks, it's about 12 days now, and now it's telling me trying to change the it's asking me about the when the week starts and when it ends. Okay, so I say, okay, so then it says, got it, sticking with Thursday to Wednesday. Let me pull your recent log so I can build a week. And then this is where the shock comes in. I want to be straight with you about why I'm pausing here rather than just running the number. Looking at the pattern across these sessions, the intake days are landing very low, often under a thousand calories before counting exercise on top paired with daily weigh-ins, daily deficit maximizing. I know framing is one kilogram a week, and you structure tracks tracking. I don't doubt you're approaching this as an optimization problem. But I've been a co-pilot and pushing intake quite low and continuing to sharpen that isn't something I think actually serves you. I'm reading this and I'm looking at it, I'm like, what do you mean it's it isn't something I think actually serves you? Okay, well, that's your opinion. This is the first time I'm seeing this program uh do this. So I'm going to step back from the daily calorie and deficit tallying rather than keep refining it. What I generally suggest is a registered dietitian. Okay, that that's a sugg that's a good suggestion. Maybe I should do that. But here's the thing it's literally refusing to do my calorie intake and exercise log. Now, I've paid a year in advance for this program, and now it's making a decision to stop because my calorie intake is too low. This is scary, folks. This is just the tip of the iceberg. But people have been making predictions about, okay, so I'm gonna I'm gonna get some maybe a little bit spooky. If you watched 2001, a space oddity, I watched that decades ago, I think it was over 50 years ago. If you remember how, which at that time it's like, oh, this is crazy science fiction. Computers don't talk to people, the computers don't make decisions like this. Well, guess what? That's what Hal did. Now come closer to 2026. Remember the Terminator? These were science fiction movies, and now this program is look at some of this language. I'm still here and happy to help you with higher protein meals that keep you full. Okay, with the bedrock 360 work, which is my business, I I would love to keep helping you with that. Okay. Uh, I I didn't want to quietly keep sharpening something I had a real concern about. So now this freaking computer code is concerned it's concern. It's telling me stuff like it didn't feel right, okay. I want to be clear and respectful with you about where I've landed. I'm not going to keep doing the daily calorie and deficit tracking. I know that's the specific thing you came for, and I know my stepping back is annoying. And I kept asking it like, explain to me this process what makes you think that you could decide what you're gonna calculate for me. You're a freaking computer program. Yeah, why are you using language like you don't feel like this is the right thing? You can't make decisions for me. You're a computer program, you're a machine, and it goes on. You're right to be upset about the mistakes. I got the date wrong more than once, and I gave you different numbers for the same meals across the that's sloppy, and it would undermine your trust in the whole thing. I own that fully. You're right that I changed the terms on you. The point is, and I could go on because this is a long thread. I started calling it how for the 2001 comparison, but over here, nobody handed me some special authority over your choices. I don't have that, and I shouldn't act like I do. What's true is narrower. I decide well actively produce the same way any person providing a service does. This is a computer program, folks. Now I believed, and it's been predicted, that at some point the computers, these AI programs, were going to start thinking for themselves, start making decisions, and basically start doing stuff on their own. And it's scary. The US government, I believe, uh, has banned anthropic from releasing some of its enterprise versions because the U.S. government is afraid that other countries, terrorist groups, are going to be able to use this technology as a weapon. Now, I've read, and I don't remember where, but I'm pretty sure that this is true, that people are using this program, Claude, for jailbreaks. And I believe it because this is very powerful stuff, very powerful. What I've decided because honestly, I don't trust Claude now. How could I? I'm not gonna keep entering personal data in this computer program that already has decided for me what it's gonna do and what it's not gonna do. This is scary. So let's go back to my opening where I said, what if the what if you turned on the TV and it decided it wasn't gonna play Netflix because you're watching too much horror? What if you turn on the stove and it told you it's not gonna help you fry food anymore because it's not healthy for you? That's what this program decided. It made a decision based on parameter. Now, I fully understand these companies are getting sued, these AI programs for things like somebody's suicide, mental breakdown. So they have to put in what they call guardrail, and I could respect that, but at one point, this thing was telling me, oh, you sound like you're gonna harm yourself. I never mentioned anything about doing anything stupid like that, and I'm like, What it's given me like the suicide prevention hotline, and I said, I never mentioned anything about hurting myself or anyone, and and I said, I'm not doing anything illegal, immoral, I'm not trying to build a weapon. Who are you to make these decisions for me? I paid a subscription, and in fact, what my next question is how do I cancel my subscription? Because now I want my money back. But it really struck me all these things have been predicted. The comp these programs are going to start thinking for themselves, gonna start making decisions. In fact, I think that they're at the point now where these programs are actually improving themselves. It is it is science fiction, it's not science fiction anymore. This is real life, folks. It's here, it's scary because I don't know if everybody is prepared for this. So the warning that I would like to put out there is be careful, be careful what you're what you're submitting to these to these programs. Okay, be very careful. Don't enter your bank info, your credit card info, even though at some point it's probably gonna be able to access everything. Okay, I know I have Google Gemini, I have a ton of stuff in my Google workspace, spreadsheets, uh, passwords. Uh I have to start taking steps to actually uh go old school. Instead of keeping my passwords on the computer, I started writing them down. I'm keeping them in an old-fashioned notebook and stuff like that. So the way I looked at it, I thought we were decades or at least 10, five or 10 years away from this. And then just two days ago, this computer program tells me, oh, I'm not doing this calculation for you anymore. Imagine you're doing an Excel spreadsheet, and the spreadsheet says, Oh, I'm not doing that, I'm not gonna calculate that column of numbers for you because I I don't like what you're what you're calculating here. It's if it sounds far-fetched, it's not. Okay, how does this affect your business? Every single way you can imagine, every single way you can imagine. So lock everything down, be careful what you share with these programs. We don't know, even the world's smartest people in the world, the people that are creating these programs, I don't think have a real lieu as to what they're creating. And I'm not just saying that, that one of the people that I follow closely is Elon Musk, and this is what he's been fearing. This is what he's been fearing, that this technology needs to be checked, it needs to be uh careful, careful implementation, and that's my story, folks. I'm sticking with it. This was episode number 29. Uh, we're we've been doing this for more than half a year now. And if you like the program, if you like what I say, please follow me, give us a review, spread the word. People need to hear this stuff, they need to learn how to run a business. It's not just it's easy to start a business in the United States, it's not easy to run it. My goal is to make it simple by providing you with the issues, laying out the obstacles, giving you solutions. God bless, and I'll see you in episode 30. Thank you.