Building an AI Portfolio: Tips for Gaining Hands-On Experience

Now it's important for you to get even more hands-on experience using AI so you can really become comfortable using it and knowing how to use it as effectively as possible. So I want to give you some ideas of how you can sharpen your skills and improve on what you've already learned and ultimately build a portfolio to show other people the value and your knowledge of AI. So how do you go about doing this? First of all, hands-on experience is really how you truly understand the capabilities of AI, and it's how you get good at doing something.

You've got to push and prod and see where it's good at doing certain things, not so good at doing other things, and it's really when you try things out yourself that you're going to truly understand its capabilities and what you can do to bring more out of it and what maybe doesn't work so well. So I've got some project ideas—things that you can do to get some real-world experience. As you do these, these are just ideas for you to start with.

I want you to experiment with them, try different things. You can't break it. You can just try different approaches if things aren't working for you.

Think differently about how you can write your prompts, and I want you to experiment and see what works well, what doesn't work well, and play with it. Push its capabilities and see where you can take it. These are meant to just be a starting point. Don't stop with them, but these are to get you going so that you can have some ideas of what you can do with this.

Also, as you do these things, consider making your own AI portfolio. Document your journey, think about the lessons that you've learned, and figure out a way that you can present that to people so you can showcase your abilities. This could be in various ways.

You could write reports to present to your Management to show the worth and value of AI and how it can improve the business—to show why it's worth you and other people learning and using this. You can write blog articles. You can create social media posts so you can track these things, boil them down into individual little nuggets, and then start posting multiple times on social media so that people can see that you're using and understanding how to use AI.

And you want to highlight how AI helped you. How did it help you to maybe do something faster, better, beyond your capabilities? Because AI can do things that you couldn't do without the help of AI. So these are all different ways to think about how AI is empowering you to do more, better, faster. One thing is definitely keep your chats. Don't just delete them because you want to be able to highlight your prompts.

If you post something on social media, people will often ask, "Hey, what was the prompt that you used to do that?" So make sure you keep those so you can talk about them. Track things like the amount of time you've invested. How much time did it save? Did it do something with better quality? What accomplishments, what real-life tangible benefits did this have that you wouldn't have been able to do if you didn't have AI? And when you're writing these, you can write them yourself, or you can even use AI to help you write those summaries.

You can input the things that you did and have it write a report and put it into nice language. And whatever format you decide to use to share this with others—whatever is best for your particular career—this can be a way to highlight your knowledge and to highlight that you're learning and you're staying on the cutting edge of technology. So first project idea: really practicing prompts and trying to get better at prompts.

I want you to experiment with prompts in the sense of starting with short, generic prompts and seeing how that works. Try not including many details and look at the response. Note things like how long it is, what's the tone, what's the writing style, what kind of audience it's targeting. And look at what it does.

Look at the assumptions that it's making, because it's got to make many assumptions—assumptions based on how long the output is—because if you don't tell it what you're looking for, it's going to make all sorts of assumptions. Then go back and either revise the original prompt or follow up with additional details—things that you want to see.

So take note of what it did and then think, what would you have done differently? How would you have changed this if you had been more clear? Then look at the response, see how it changed, see whether it should be closer to what you want, or maybe were there ways that it didn't listen to what you wanted. Then do you have to clarify that again because maybe it misunderstood what you were wanting or didn't listen to a certain instruction. I want you to keep doing this for new chats with different prompts and try this again and again, and see how it works if something's vague and how it works if something's detailed.

Sometimes being vague can actually be useful. Sometimes you don't have something in mind, and it can be nice to start with something short and vague and see what it comes up with—see what it highlights.

If you're coming up with ideas and you don't really know exactly what you're looking for, that ideation stage—where you're coming up with these concepts—could benefit from leaving things open-ended. You don't always have to go in with something specific. But when you do start to clarify something and really want to be specific, see how that works and see how well it listens—or doesn’t listen—to your specific goals when you put those in. Try it with different types of tasks. Your goal here with trying to experiment with different prompts is to become a Prompt Engineer, which is a fancy way of saying, “I know how to talk to AI to get the most out of it.”

You're pushing and prodding, trying to figure out what all of its capabilities are. As a reminder, on this screen you can pause the video if you want to review things that we've shown earlier. These are just ideas of some of the things—just a few of the capabilities—you can try, which we've discussed before. Try out lots of different things.

Try the vague prompts, try the specific prompts, and find things that are useful in your day-to-day jobs and in your day-to-day life. Now, as you do these prompts, let's say you're not getting the result that you want. What are ways you can improve it? Well, we already talked about some of those ways, but another idea here is you can use this prompt:

Tell the AI: “Act as a Prompt Engineer. Review the following prompt.” So the idea is if you have some sort of prompt that you're going to paste in below, you want to say, “You, AI, act as a Prompt Engineer. Review the following prompt and help me optimize it. Ask me any questions required to get the details you need to optimize it.”

Sometimes ChatGPT or Copilot will just do the thing you tell it to do, and they don't always ask questions. Sometimes they do, but by telling it's okay to ask questions, it's more likely to follow up with a question rather than just do what you said—because its default assumption is to just do what it's told. But if you give it permission, it's more willing to ask. That can increase the quality of your response.

The other approach too is, in this prompt, you can also tell it what's not working well. Say, “I'm having a challenge getting this kind of result. How do I revise the prompt to get you to do what I'm hoping you'll do?” Then once it outputs a new prompt, go to a new chat and try that new prompt out. That's important because you want to open up a new chat to do that.

If one chat stream is thinking about how to improve a chat, the context of that is it's trying to improve the prompt. When you create a new chat, it wipes the slate clean and focuses on executing the prompt with a fresh set of eyes. So always use that new prompt in a new chat.

As an example here, let's say I was going to do something very generic. If I was going to say, “Write a blog post about the Hawaiian Islands, ” I'm not really saying the length, not saying who I'm targeting. It's very open-ended, and it'll give you an idea of what the default writing style is like—how formal, informal, and so on. Read it for tone, read it for length, read it for order.

What kind of things is it talking about? It made a lot of assumptions because I gave it very little information. Now, if I had something more specific in mind, I'm going to stop this. I'm not going to wait for the whole thing to finish.

I could either type in a new prompt here, or I'm going to go back and edit this message. I'm going to put in a much longer one: “Write a blog post about the Hawaiian Islands, listing interesting things to do for tourists who have never been there before and want to learn about the local culture, food, and unique activities.”

Keep recommendations specific to each island, listing the islands in order from most visited to least visited. See how much more specific that prompt is. I'm telling it who I'm trying to talk to, the kind of things I'm talking about, and the order of things—because it might've chosen a different island order.

I wanted to put the most popular islands first. Then it gives something that I think is much more targeted at people who want to go learn about the islands. It's listing cultural experiences, local food, and unique activities.

It literally took what I said to do—list those things—and is listing that for each of those islands. See how much more specific that is. This is what I want you to try—not just for this particular topic, but for your own.

Find something that is interesting in your job or in your life and try this out. Go with the vague things, see what it does. Try the specific things and see how much you can pull out of ChatGPT or Microsoft Copilot and see what it can do for you.