Don Sesler, The Technology Whisperer: As the owner, I have the honor of leading a team of super-smart individuals who share my passion for helping entrepreneurs and business owners not fall victim to complex technology challenges that can kill productivity. We want to tame your technology so that it works for you and not the other way around. What technology problem is making your life more complicated than it should be? Reach out to me and tell me your story.

AI Cautionary Tale

AI Cautionary Tale

Ever have one of those moments where you press a button on the keypad and immediately realize it was the absolute wrong move? That just happened, and AI was right there with me.

I was deep in the technical weeds of resculpting our internal email and CRM systems to get them ready for AI integration into our business. It’s a tedious process, and I was using AI to help me navigate the technical hoops. As we went along, I did less and less fact checking on the AI recommendations, and one of those recommendations led me to delete a huge chunk of content from my CRM.

Like, gone-gone. Really gone. Not coming back gone.

The truth is it wasn’t that big of a deal. I had the information in other places, which is why I took the risk, but the unexpected vanishing act was a reminder that collaborations between AI and humans are far from perfect.

Why This Happens

AI applications (like some humans) are trained to be fast and confident out of the box, which means they sometimes make half-baked, false statements with a confident tone. It’s easy enough to blame the tool, but the reality is that we humans are usually complicit in the mistake, because we have failed to give the app the full context it needs to correctly answer the question. AI’s goal is to put together a response to whatever prompt it’s given, no matter how vague the question, after just a few milliseconds of thought. If that is good enough for us, that is what it will continue to do.

How To Train Your Bot

Most AI tools can adapt to the questions you ask and the guardrails you set. If you provide professional services like I do, then you’ll want to train yours to be more like a cautious analyst than your local know-it-all barista.

Here are some of the actions I have taken:

  • Ask for sources: Pushing for transparency will help you separate fact from guesswork.
  • Request confidence signaling: Train the bot to tell you when it knows something, and when it is guessing.
  • Avoid pronouns in your prompts: Pronouns can be ambiguous, and AI can make the wrong choice about what you really mean.
  • Set your risk tolerance: AI models value speed above all else, even at the expense of occasional mistakes. Setting some risk tolerance guardrails can help.
  • Watch out for those irreversible actions: When you are deleting, deactivating or publishing something, ask the question “What might go wrong here?”

Some of you have heard me say that most AI tools fall somewhere in the range between a smart college intern and “a useful idiot”, but where they land on that range largely depends on how well we work with them. The better we get at asking good, precise questions and establishing reasonable expectations for the tool, the better AI will be able to help us in this new era.

Are you interested in using AI in your business? Feel free to schedule a free consultation with me, I’d love to hear more about your plans in 2026.