AI Prompt Engineering. What is it? Prompt engineering refers to the process of asking questions and interacting with a Large Language Model (such as ChatGPT or PMI Infinity) to draw from vast amounts of data and get answers that are customized to your needs. This is a new area of expertise in every professional field, including project management. For many, it might feel like an overwhelming concept, but with a little instruction and a lot of experimentation, anyone can become a Prompt Engineering master. There are oodles of resources for individuals who want to improve the crafting of perfect LLM prompts, including a free course for PMI members. Here are a few tips to get you started on your Prompt Engineering journey.
Make AI your friend
What is the point in having a real-life Star Trek-esque “Computer” if you can’t have a little fun with it? One of my favorite suggestions from PMI is to give your AI a name and talk to it like a person. I’m partial to Leonardo DiCapri-code, myself.
Say please and thank you. No, really. Recent studies show that the language you use with AI can influence where the LLM pulls their sources from. Polite language is more likely to elicit responses based on more formal, and therefore typically more credible, sections of the internet. Demeaning language is more likely to turn your LLM into a bridge troll who makes anonymous comments on strangers’ Instagram accounts. Jokes aside, the LLM is learning from you, too, so the language you use really does matter. How you phrase your prompt, like including please, helps the LLM understand better when you are asking a question, and gives subtle context to your requests. Plus, treating it more like a conversation with a colleague is more likely to get you the results you are looking for. AI benefits by not only providing information but can help your professional development communication skills, and I know a lot of people who can use those skills; not to mention it is free practice.
K.I.S.S.: Keep It Simple…Sort of
Prompt Engineering is fundamentally computer coding but with words and written language. It is only going to do what you ask it to. Think of it like an extremely smart individual who, though brilliant in their field, God help them if they are sent to the grocery store by themself. Something as simple as “canned corn” can cause confusion, requiring them to call their partner every 5 minutes with endless questions: What brand? Low sodium or regular? 15 oz or 24 oz?? As the person at home who is familiar with the recipe, the answer to these questions seems obvious, but if left unanswered, the result could be not getting what you need.
So keep it simple, but as specific as possible. Give context like, whose voice this is coming from, what their profession is, who the audience is, the occasion, etc. Tell the LLM what tone you are looking for, whether it is casual, personal, formal, or something in between. Tell it how you want the information presented: paragraph, speech, social media post, report, middle school love letter…the possibilities are endless.
Stack your AI prompts like building blocks
In most cases, a single prompt will not be enough to get you to the result you want. Prompts are building blocks – stack them on top of each other for best results. Break your problem into its most basic pieces, and start one piece at a time. Instead of starting over with new prompts, return to the conversation you already have going with AI. The history portion of your LLM is a very valuable resource that you can return to as often as necessary.
You cannot think of AI as a shortcut; in most cases it will absolutely save you time and energy, but it won’t be instant. There will still be an investment of time and creativity to elicit the type of responses that make AI a great partner.
Be patient, trust the process, and give lots of feedback. Tell your AI helper what parts of the prompt it fulfilled according to your expectations and what parts you want revised. Think of yourself as a director, giving stage directions and corrections. “This time, with more emotion!” (Which actually is really close to something I have said to ChatGPT before.) Try something really cool by asking AI how it would have asked the prompt to give better results; It’s like saying to a friend, how can I communicate better to be more helpful? Whoa – no life coach needed.
Here’s me leading the AI session at the PMI Mile Hi Symposium 2024
Embrace integration with the robots
Most of all, just dive right in. Working with your AI buddy really can be like brainstorming with a colleague. Through the process of giving feedback, making revisions, and condensing ideas, you will develop a clearer vision of your end goal. You will come up with new ideas and inspiration.
If you are new to Prompt Engineering, I hope this gives you some tools to get started. If you use AI right, it isn’t cheating. It’s evolving and embracing technology. Use AI to its full capacity, but also verify the work and statistics and other information, and finesse the AI outputs to fit your style and situation. Lastly, here are the courses that I’ve taken in case you are interested: