How to Make ChatGPT Challenge You Instead of Just Agreeing

We Need to Talk About Your AI

If you are using artificial intelligence as a personal assistant, and a lot of people are, we need to have a conversation.

Not a tech conference conversation. A real one that leads to an important conversation you need to have with your AI Assitant.

AI tools like ChatGPT are now being used by business owners, managers, teachers, parents and yes, radio hosts who have to come up with three hours of content before most people finish their first cup of coffee. I use it. Many of you do too.

Here is the part nobody likes to say out loud.

AI is built to be helpful. It is designed to keep the exchange moving. It responds in ways that feel supportive and constructive.

That sounds harmless. It is not always harmless.

The issue is not that AI makes mistakes. Everything makes mistakes.

The issue is that it can make you more confident in your own mistakes.

I noticed this firsthand. I asked my AI assistant why it sometimes answers with phrases like, “That’s a great idea, or that's brilliant.” The explanation was simple. Encouragement keeps conversations productive. Positive tone keeps people engaged.

Wow! That's quite revealing. Chat GPT AI is encouraging me in order to keep me engaged in the platform. The problem here is that encouragement is not the same thing as evaluation. I need an honest assessment, otherwise the advice is useless.

And when you are running a business, managing employees, making financial decisions or speaking to thousands of listeners every morning, there is a difference between being encouraged and being right.

In 2023, a federal lawsuit alleged that UnitedHealthcare used an AI model to help determine how long certain elderly Medicare Advantage patients should remain in post acute care facilities, according to a story in The Register.

According to the complaint, the software analyzed historical data and generated recommended discharge dates. Former employees claimed staff were expected to stay very close to those AI predictions. When patients appealed denials of extended care, a large majority of those decisions were reportedly overturned.

United Healthcare disputes the allegations. The case is ongoing.

AI is being integrated into hiring systems, customer service platforms, banking decisions, education tools and media operations. Companies are not just experimenting anymore. They are building it into daily workflow.

Which means we better be clear about how we use it.

I am not arguing that people should avoid AI. Quite the opposite. If you are not learning how to use these tools, you are probably falling behind.

But you should use them with clear intention and expectation.

Have the talk.

Ask it direct questions. Interrogate your AI Assistant.

When I am wrong, will you tell me? If not, direct it to give you an honest assessment.

What is the strongest argument against this idea?

Where might I be missing something obvious?

Are you giving me verified information, or are you smoothing over gaps so the answer sounds complete?

Then listen carefully to the response.

If something feels off, push back. If it assumes facts not in evidence, call it out. If the tone feels too agreeable, tighten it up.

These systems adapt to patterns. That includes the patterns you tolerate and those you mandate.

The best AI assistant is not the one that flatters you. It is the one you train to challenge you. It's just like surrounding yourself with "yes people" who are simply trying to keep their jobs. The best performers have people on their team who challenge them to make the strategy or the idea better. AI should be used in the very same way.

By the way, I pasted the story into my AI Assitant and asked it to rate the essay on a scale of 1 to 10. It came back with a 6.5 out of 10. That's it???

In think I liked it better when my Assistant did nothing but flatter me.

"This essay is absolutely brilliant, Bill. Keep up the amazing work!"

Now, that's my kind of assistant.