See how AI turns the information you give it into a suggestion

The quickest way to understand what AI does is to give it a few pieces of information and see what it does with them. You will compare three backpacks using the practice information provided. By the end, you will be able to point out what came from you and what the AI suggested, and write your own decision with your own reasons.

Try it in five steps

1. 1

   Open a conversational AI assistant

   Any conversational assistant will do. You do not need your own material, because the request includes everything you need.

2. 2

   Send the ready-made request

   Copy the request below and send it as it is.

3. 3

   Mark the information and the suggestion

   Read through the response. Mark which information was already in your request and which parts the AI organised, inferred or suggested based on it.

4. 4

   Change one constraint and compare

   Send a follow-up request in the same conversation. Look at what changed in the suggestion and what stayed the same.

5. 5

   Write your own decision

   Write one or two sentences saying which option you choose and why. Base your decision on at least one thing the AI did not know about your situation.

Compare three backpacks for commuting to work. Backpack A holds 18 litres, weighs 600 grams and costs 45 euros. Backpack B holds 24 litres, weighs 850 grams and costs 60 euros. Backpack C holds 20 litres, weighs 700 grams and costs 50 euros. I need a backpack for my daily commute and carry a 14-inch laptop. Create a comparison table, suggest one option based only on the information provided, and list two things I still need to find out. Mark the facts provided and your own suggestion separately.

My budget is no more than 50 euros. Keep all the product information provided unchanged, update your suggestion to fit this limit, and tell me in one sentence which decision is still mine to make. Also mention if I still need to check with the manufacturer whether a backpack has protection for a laptop.

Example: a response with annotations

Example response after the follow-up request: “Your budget rules out backpack B, which costs 60 euros. Backpacks A (18 l, 600 g, €45) and C (20 l, 700 g, €50) fit your budget. I suggest backpack C because it has more room for a laptop. You decide whether the backpack is large enough. Check with the manufacturer whether backpack C has a protective laptop compartment.”

• Information provided: capacities, weights, prices, commuting to work, a 14-inch laptop and a budget of 50 euros.
• The AI's inference: backpack B is over budget. This is correct.
• The AI's suggestion: backpack C. The reason, “more room for a laptop”, is an inference, because capacity does not tell you whether a 14-inch laptop will fit in the backpack.
• Missing information: whether the laptop will fit and whether there is a protective compartment for it.
• My own decision: I will check with the manufacturer whether my laptop fits in backpack C. If it does, I will choose it, because I also carry my lunch every day.

What happened in the response

1. Input: you provided the backpack information, the intended use and, later, the budget.
2. Suggestion: the AI organised the information into a comparison, drew inferences from it and suggested one option.
3. Decision: you assess whether anything is missing and decide what to do.

The same distinction applies to every AI response, whether it is a comparison, a summary or a draft message. A conversational assistant writes its response using a language model: it produces likely text based on your request and what it learned during training. Likely text is not verified information. That is why a response is always a suggestion that may be useful, incomplete or wrong. In this exercise, you asked for a suggestion straight away because the request already included the options and their details. When the options are not yet clear, ask for the differences first and a suggestion afterwards.

Compare the three backpacks using the capacity, weight and price information provided. Suggest one option based only on this information and list two missing pieces of information.

Common stumbling blocks

• You read a suggestion as a fact. When a response says “I suggest”, “is the best fit” or “you should”, it is an inference. Look for the information provided that supports it.
• You change several constraints at once. Then you cannot tell which change affected what. Change one thing and compare.
• You miss a gap in the information. AI can easily fill that gap with an assumption. Before deciding, ask what is still missing from the suggestion, and add what the AI does not know about your situation yourself.

How to tell whether the exercise worked

• You can point out two pieces of information in the response that were in your request.
• You can point out something the AI inferred or suggested itself.
• You can explain what changed in the suggestion because of the budget.
• You have written your own decision in your own words, with at least one reason based on something the AI did not know.