Everyone has that coworker, friend, or family member who shares information that can be best described as confidently incorrect. You know the type. Never at a loss for a fact. Treats opinions as facts. Speaks first, checks never. Completely well-meaning. Impossible to slow down.
If you recall the TV show Cheers, I bet a name just popped into your head. Cliff Clavin. Postal worker. Bar philosopher. Human search engine with zero filter. Everybody loved him. Well mostly. And nobody trusted him with anything important. Fast forward to today and imagine Cliff with an AI subscription!
The Prompting Trap
We have spent three years learning how to write better prompts. Yet almost nobody is talking about what to do with what comes back. Learning how to prompt is clearly useful. Frameworks, mnemonics, cheat sheets offered on LinkedIn. All serve to ensure your input is structured.
Heck, I even built one. I called mine PIE: Prologue, Instruction, Epilogue. (Yes, I have a terrible sweet tooth!) I like to think of the prompt box like an oven. You stick your tray of ingredients in, assembled exactly the right way (you even fluted the pie crust!) and then ding the timer goes off and you open the door.
You lift the tray out and look at your pie. But what if you're baking experience consisted of using an easy-bake oven when you were six? Do you know what a good pie looks like? Can you tell the difference between a pie that looks done and a pie that actually is done? Sure the recipe got you to the oven. But the moment that door opens, you need something the recipe never gave you. Judgment.
When Close Enough is not Good Enough
Oh sure, for an apple pie, close enough is probably good enough. Who even knows what nutmeg is anyway? However for that Lemon Meringue pie to be just right you need to know how your oven runs, and when to pull the pie to let the carry over heat take it the rest of the way.
At some point baking becomes more then just following the recipe. Almost counter-intuitively it is the very structure or the recipe or the prompt that can promote false confidence.
We confuse precision with accuracy. A clean, well-structured prompt can return a beautifully formatted wrong answer. Structure is about form. Accuracy is about truth. Not the same thing.
We confuse speed with confidence. AI does not pause. No "let me think." No hesitation. We have learned to read the pause in a person as a sign they are actually thinking. AI skips the pause. We read that as certainty.
We lead the witness. Lawyers get called out for this. How you frame a question shapes the answer before it starts. A leading prompt gives you your own assumptions back dressed up as analysis. That is not research. That is a mirror.
Enter Carla Tortelli
And then there was Carla Tortelli. The character was a sharp-tongued waitress, who called it like she saw it, and questioned everything. And she never missed a chance to challenge whatever came out of Cliff's mouth.
Critical thinking has always had two parts: thoughtful input and examined output. AI did not change that. It just made the input side feel finished while the output walks out the door unchecked. I built PIE to get you to the oven. That is what newcomers do when they start prompting. They recognize the input needs structure, just like baking a pie.
But over time, working with AI actually requires cooking. Baking follows a recipe. Cooking is a conversation. You taste, you adjust, you taste again.
The Discipline of Output
Output when using an LLM requires the same discipline. Resisting the temptation to go with the first answer is hard. Treat the output like Carla treated Cliff.
Ask the output to show its work. Ask it what it is not telling you. Ask it to argue the other side. Does the answer hold up when you push back? Would you stake your reputation on it? Or your next career move? Because being confidently incorrect has a cost. In a coaching conversation, an AI-generated insight delivered without examination is not helpful. It is a liability with a friendly interface.
So stop. Take the time. Read it twice. You can even respond with a snarky comment!
Be Carla.