If you are using ChatGPT or Claude or other major AI Chatbot on a regular basis, you know that it’s a quality-in-quality-out proposition. You also know the key to successful results is specificity. The more specific you are in your prompt – or instructions – you give ChatGPT, the more accurate and compelling the output will be.

When I work with executives to understand how to use a chatbot correctly, I emphasize there are no shortcuts unless you want your results to short-change you. The more time you spend perfecting your prompt, the more perfect your replies from ChatGPT are going to be.

What happens if you leave out a crucial instruction? Sometimes, if you miss adding a single word, the result can be disastrous. 

Here's a real-world example. One of the superpowers of the leading chatbots is they excel at summaries. We record conference sessions in Otter, a mighty AI voice-to-text transcription software. We then download the transcription and upload it into ChatGPT or Claude.

For our "AI Quote of the Week," we often ask for top "AI quotes." We instantly get back a selection of quotes, saving us minutes to hours if we had to scour the transcripts to find the best quotes.

Here's the danger: When you ask a chatbot to extract the quotes, you must be very specific and ask for the quotes verbatim. 

Why? Your chatbot is likely to "clean up the quote" with grammatical changes or even change the quote to "sound better." After all, the chatbot is trying to please you.

So, if you ask for a quote from a transcript but forget to tell the chatbot to make sure the quote provided is word-for-word what the person said, you can get back a very different quote.

This mistake is easy to make and the kind you find out about only after you publish the quote. So, remember to add the word “verbatim” anytime you ask for a quote from a transcript. (-Kevin)