Episode transcript The transcript is generated automatically by Podscribe, Sonix, Otter and other electronic transcription services.
Hi everyone, here is Ricardo Vargas and this is the 5 Minutes Podcast. Recently, I did an excellent course provided by PMI to PMI members called Talking to AI Prompt Engineering for Project Managers. And it was a great course. And I think it's extremely valuable for the project managers and for our work and managing projects. And by the way, just to start, this is not a sponsored episode. Okay, so PMI has no idea that I'm recording this, but I want to share with you one aspect and one specific topic on this course that I think is extremely valuable for us. And this is not about one specific technique, but some general guidelines for you when you are using tools like ChatGPT, Claude, Gemini, or Copilot, and you were doing prompts, and this is to help us to get to the answers we want in a much faster way. And the first thing that these guidelines say is to be specific. One of the worst things for you to have, I would say, a bad experience with this kind of generative AI tools is not being specific. If you go now on ChatGPT and just ask ChatGPT, please make a drawing of a house. Do you agree with me that there is a massive risk that the house that Dall-E, for example, will give to you, has no relationship with the house that you have in mind? Why? Because there are billions of different ways to draw a house. So you need to be specific, and it's the same for your prompt. The second thing is to provide context. You need to explain the context in which you are looking for an answer to avoid what I said many times, which is this kind of posterized answer. And just to remember, when I say pasteurization. Pasteurization is a process where you heat up, for example, the milk to produce cheese and then make it cold in a very short time frame. And this kills the bacteria. Okay. The bad and the good bacteria. This is why, for example, a cheese produced with pasteurized milk is always a much more flat risk cheese than a cheese that uses unpasteurized milk. But this is just to explain. So it's the same here. If you do not provide context and you are not specific, you will have a pasteurized answer. It means that the answer is just a generic answer that will serve nobody. The third thing is about tone, and this is super important. Every time I ask ChatGPT, for example, to give me an answer, or I would say to give some insights on something I'm doing, I always put the tone I'm expecting the answer. For example, just to share with you, many times I write look, write this in a positive tone, a friendly and informal tone because this is who I am. Also, I always ask ChatGPT if I can reply and write in plain English. For example, if I want to say something or, for example, I'm preparing a script for something. Why? Because I'm not a native English speaker. So if you hear what I'm saying or writing very complex or elaborated words, the first thing you will know is look, Ricardo did not write this or is just reading a teleprompter that he has no idea of what is behind. And this, for example, just to be honest, if you go on, for example, LinkedIn and you see the posts, and of course, I know many people behind this post, and I say, look, this post has nothing to do with the person that is signing the post. Why? Because the person is using ChatGPT and does not even care to write, ask GPT to give a post in a tone that looks like him or her. So, tone matters. You know, people will notice that it's not you. And I'm not saying that it is unfair for you to use ChatGPT, please. I'm not saying that, but it looks very bad if you post something that people know does not look like you. So, tone matters, and one of the best ways for you to give context is to make sure that the tone is right and specific enough by providing examples. Give some examples. For example, I did this by copying and pasting some examples that will help the generative AI process give you a better answer. Another thing you need to know about your audience is that many times, the answer that will come will be an answer that is not suitable for your audience. And I love to use the concept of PMI when I say it on this podcast PMI I have no doubt that 99.99% of the people listening to this episode know that I'm referring to the Project Management Institute, but I had situations when I said PMI and some people in the audience of course, not related to project management, they thought that I was talking about Philip Morris International, the cigarette manufacturer. So, did you see how this is different? So you need to know your audience, okay? So, you need to craft the message. If you are preparing something, it will be shared and discussed with children. So, you need to write in a way that children are able to answer. So, you need to create a prompt that will provide you with an answer based on your audience. The next one is about clarity. You need to make your request to ChatGPT or to Gemini with clarity, with no doubts about what you were expecting to be done. One of the best ways to do that is by using a structure. You can even structure and say, I want your answer to cover one, this, two, this, and three to help and guide you to have the best answer. And last but definitely not least, you should experiment, test and refine most of the time, if not always, the first answer I get is not exactly what I was looking for. So many times, I take pieces of the first answer to craft the second prompt. I say, for example, I like this part of your answer, but I want you to refine that part. So, with two, three, or four interactions, ChatGPT or Copilot will give me the answer I'm looking for. So it's not for you to believe that you will just make your initial prompt. Write two lines, press enter, and then ChatGPT will give you back a perfect answer because this does not work okay. Remember AI and generative AI tools; they are basically huge calculators and probabilistic calculators. They don't think okay, despite of many people think that they have a soul and feelings, and I'm not discussing that today, they are statistical and probabilistic machines. So, if you are specific, give context. Make sure that you are setting the tone. Provide examples; know your public. Make things clear. Create a structure and experiment, test, and refine. You will get a perfect result, and if you are a member of PMI, do not miss the chance. It's a very fast course. It's very well presented and well delivered, and it will help you a lot in your work as a project manager. I hope you enjoyed this podcast, and I will see you next week with another 5 Minutes Podcast.
