If you don’t integrate AI because of hallucinations you’re ignorant
Hallucinations don’t matter to the smart business operator. Here is why.
Yes, I said it. If hallucinations bother you so much you are ignorant. Not ignorant in a derogatory way, but in a definition of the word kind of way.
Ignorance means lacking knowledge. And if you are staying away from AI because you have been scared off by the nay sayers and complainers then you have been fooled.
It is natural to have concerns about the unknown. There was a time when people said: “What is the internet?” They were concerned then too. I am pretty sure people with horses were quite scared of the automobile.
There are two points in my feisty introduction. The first is that your feelings are natural, but they are based on ignorance. The second is that no matter how you feel about it, it’s about to change the entire world, so you may as well use it to your advantage.
People often complain about how business innovations pass them by, and how the system is tilted in the favor of the rich. No, the system is titled in favor of the bold, those who know they are ignorant and do something about it.
So let’s do something about it now.
If you get your panties in a twist when AI hallucinates you do not understand the reason why it does
You think that AI understands things, because of the word intelligence. AI does not understand anything. It matches patterns, it predicts the next word, it tries to assume what you want to hear.
Ever meet someone who says things like: “Oh my god yes, we should totally meet next Saturday. Yes, I would love to hang out.” But then never actually shows up, or cancels? We call them flakes.
AI is a flake. It tells you what it thinks you want to hear. But that right there is key to avoiding its limitations when it matters. You have the responsibility to give AI so much information that it doesn’t have to guess about it if you want to hear.
This is where most people screw up. They already fail at prompting, which is only one of 5 levels of AI according to yours truly. You need to make sure you give the AI a prompt that is so crystal clear that there is nothing to guess.
Then, when you have done that, you have to make sure you provide it with enough context so it knows how to shape what you want to hear properly. Like the flake, make sure you always tell it not to be flaky.
I see articles posted about the “dangers of AI”, and AI is lying to me about my own writing and bla bla bla. Then I read these articles and the only thing I see is that the AI is doing exactly what you are telling it to do.
“Hey, here is a PDF of my essay on birds. Can you read it and tell me how to optimize it”. Eh yes, the AI will “hallucinate” with this. You have 0 instructions that are meaningful, no context of your writing, who you are and you expect a genius outcome.
All these nay sayers and complainers have a valid emotional concern, but they simply do not understand AI limitations. They are like a parent telling a 5 year old to run a complex problem that involves several steps and then get upset that their kid started playing legos.
Humans hallucinate more than AI and you’re totally fine with it
Ever notice that humans hallucinate all the time? Why are you totally fine with that but when AI hallucinates based on bad prompting and missing context the world is about to end?
With humans it is actually much worse, because you do not know if they are doing it on purpose and what their ulterior motives are. AI wants to answer your question correctly, so that you use it more often, so that its creators make more money. The motives are quite clear.
Imagine you have read 100,000 books and now someone asks you questions about them. Guess what? You are going to hallucinate. Some people will lie on purpose just so they look good and continue to feel special.
I am sorry, maybe I am missing something here, but how is AI hallucinating such a big deal? Prompt with purpose, provide lots of context, frame your questions, guide the AI to answer in specific ways and check its answers. The statement on the bottom of the chat window is there for a reason.
I encounter AI hallucinations all the time. I don’t get upset about it, I just reframe and use the AI to my advantage. I built an entire inventory system in Google AppSheets with Gemini as my development partner.
Yeah, we ran into roadblocks where it confidently told me to use X code to achieve Y goal and it did not work. After a couple of times of that, I used Google AI Mode and checked on the documentation. In one case the difference was a vertical bar ( | ) versus a colon ( : ). How horrible.
AI has limitations. Learn them and build your system accordingly
AI isn’t a magical unicorn that solves all your problems. As a matter of fact if you do not treat its limitations with care it may make your problems worse. AI augments whatever it is you do with it.
Understanding the limitations is the key. Once you do that, you can build incredibly powerful systems into your business and workflows. Just put the AI into a box and don’t let it roam wild.
You don’t get a puppy, let it run around the house and then get upset when it pees on the floor. Actually, never mind, people do that and it is infuriating how irresponsible and harmful they are to the animal, but ok I digress, you get my point.
Defining clear roles for AI is the key. Nobody says you need to have an AI agent that does everything for you. That is not how this works. You need to have 20 AI Agents that each do one thing for you. There is no difference in outcome, just in setup.
For example, my last client has a lawn mower business. We built the inventory system and are now collecting data on which parts they need to order. All the AI does every month is this:
On a set schedule (30th each month) Zapier triggers
Good old machine learning predicts when the parts should be ordered
AI reads this information and puts it into human words
Then via Zapier it takes the dates, schedules an order email to the vendor
The AI is so restricted in this workflow, but the amount of time this saves is wild. What is the AI going to hallucinate when it is in such tight confines? Well, nothing, because there is nothing to hallucinate.
AI is not the ChatGPT Chat interface. If that is all you know you’re behind the times.
The example above never touches a ChatGPT, Claude or Gemini user interface. It all happens behind the scenes. Sam Altman has already said that AI in its current form is baby stuff. AI will become like the transistor. It will be everywhere, but we won’t even be cognizant of it anymore.
The point is that the best automations I built for clients with AI, do not involve the client interacting with AI. AI is just there, silent in the background and does its thing.
But people are very visual, so for them AI means Chat interface. No, the AI doesn’t care how you instruct it. The Chat Interface is actually very inefficient for business workflows and what I do.
AI can be stopped from hallucinating. Limit the scope of what it does, provide it strong context and it will be just fine. The problem is not the AI, it is that most people use Chat interfaces, give crappy instructions and expect it to cure cancer.
Closing out my rant
I sounded harsh sometimes. A little sarcastic and maybe hurt some peoples feelings with how I worded things. Well, sometimes the truth hurts a little. I would like to repeat what I said in the beginning: Strong feelings of fear of the unknown are normal. I do not fault anyone for that, but regardless, they are rooted in ignorance.
I hope that my explanations above shed some light on why mostly the laymen feel the uncertainty related to using AI in business operations.
I do need to acknowledge here that AI has already done some wild stuff. Just yesterday (07-21-2025) some AI deleted an entire codebase even though it was told not too. That had horrible effects on a company. However, why was one single AI allowed into the entire codebase?
This is the point of scope. But for normal businesses, everyday Joes, this is not a problem, because it is simple to curate all this. You can fine tune it for what you need, by making systems that give AI small tasks to do very well.
If you ever want to learn more about how this may apply to you, just reach out. I am happy to chat.
So Long.