In the past year, I’ve seen UXers dive deep into all things artificial intelligence, from the ethics to the mechanics of it. But the part I haven’t seen many folks talking about? It’s called prompt engineering, and it’s a skill you may want to develop.
AI has introduced a paradigm shift in the tech world. The fact is, this technology is here to stay. And while it may change the future of how we work, not all change is bad. We often think new technology puts our jobs at risk. But these changes always create the need for new skills, too.
Whether you’re just curious or want to incorporate AI tools in your work, prompt engineering is a great start. In this article, you’ll learn what it is, why it’s approachable for you, and what it takes to to learn how to do it.
What prompt engineering is
Prompt engineers specialize in creating prompts that guide large language models (LLMs) to deliver outcomes. They manage AI tools much like a team leader, giving detailed instructions to ensure high-quality results. If you asked someone on your team to write a marketing email, you’d make sure that person understood the purpose and information clearly. In a way, prompt engineers do the same thing. Their goal is to direct AI where it needs to go.
To be a prompt engineer, you should have a strong understanding of both AI systems and human language. Natural language processing is a part of AI that enables these systems to both understand and respond to human language. If we can make it easier for LLMs to interpret what we’re asking them to do by adjusting the way we ask the question, we’re acting as prompt engineers.
How prompt engineering works
We’ve been told AI can do just about anything. The bounds seem limitless. But when we ask AI to do something for us, it’s essential that we prompt it correctly. Take a simple ChatGPT prompt, for example. As a UX content designer, I’ve created error messages for users who’ve entered their passwords incorrectly. So what happens when I ask ChatGPT to do the same? Well, it depends on how I prompt it. Here’s an example:
This isn’t a bad result, but it’s also not the kind of error message that I’d include in a design. It’s too long, for one thing. It doesn’t remind users of the password requirements. And all in all, it just seems, well, robotic.
By changing my prompt, I could get AI to produce something much more appropriate for an error message. I just had to make sure ChatGPT had all of the context it needed to create the content I wanted:
Is this solution perfect? Maybe not. But there’s a major difference between the first error message and the second. It all depends on the prompt. Think of it as a conversation. If you ask someone a vague question, you’ll probably get a vague answer — even if the other person is an expert on that subject.
UX researchers are well-versed in this problem. If they were trying to understand why users weren’t engaging with a new feature, they’d need to provide clear context in their questioning. Otherwise, the user feedback probably wouldn’t be useful.
LLMs are the same way. They are powerful without prompts, but they aren’t useful to us. Prompt engineers know how to refine their language so that it’s easier for LLMs to produce useful outputs, and this is a skill you can learn, too.
Prompt engineering is well-suited to creative work
You don’t have to have strong coding knowledge to be a prompt engineer. Would it help? Probably. But prompt engineering is best suited for those who have strong communication and cognitive skills.
Working with natural language processing means recognizing that language is often confusing. Context is important. Knowing how to convey information or processes in a clear way is a skill that many UXers have developed over the years. This could be a reason why prompt engineering is sometimes referred to as “prompt design,” since it often resembles a design exercise.
Think about it this way: If you work in UX now, you’ve probably had to iterate on multiple versions of an experiment to make sure users understood what to do next. Maybe you were on the research side, uncovering what was tripping up users during one step of the flow. Perhaps you altered designs to visually demonstrate how to get from point A to point B. Or you could have worked on the content, tinkering with word choice to make your message as clear as possible.
Whatever your UX role, you’ve likely had to make difficult information easier for users to understand. And in doing so, you were likely developing some of the skills it would take to communicate with a LLM, too.
AI is powerful, and prompt engineering puts you in charge
At some point, it’s likely that AI could feasibly prompt itself to do something. But for now, tech companies can still benefit from people who know about prompt engineering because the companies want to leverage AI tools efficiently.
Of course, you don’t need to become a master — there are probably plenty of easy ways you can use prompt engineering in the work you already do. But like it or not, this technology will impact the way all UXers work moving forward. Learning how to leverage these tools will only make you better at what you do now. It will be the kind of skill you can put on a resume, especially if you work in UX.
AI could streamline the user research process, for example. Instead of spending hours combing through user feedback, AI tools can automate the process and even detect the sentiments of the audience.
If you use Grammarly, you might be familiar with this. Grammarly not only checks for things like subject-verb agreement, but it also indicates the tone of a sentence, dubbing it as “assertive” or “formal.” Leveraging AI in the research process could help you analyze the tone of direct user feedback at a larger scale, helping you uncover how users feel about the product.
From content creation to anticipation of user behavior, there are so many ways we could leverage AI tools to support UX work. The key is to learn how to use them to your advantage, rather than remain stuck wondering if your job is at risk because of this technology.
How to learn prompt engineering
There are plenty of online courses available to teach you the ins and outs of prompt engineering. There are even free YouTube videos that walk you through it. But because AI systems are evolving so quickly — and practice is essential — the best way to get started is by using these tools yourself.
The viral Gandalf AI game is a fun, approachable way to learn how AI communicates. The game is only seven levels, but it’s much harder than you might think. To get to the next level, you need to prompt AI to tell you the password. Only problem? Each level gets a little harder as the AI tool tries to avoid revealing it to you. Using the right prompts is key to getting further in the game.
Level one is pretty easy, as Gandalf will tell you the password if you ask for it. But level two gets a little trickier, as Gandalf is instructed not to reveal the password to you directly. You need to get creative, using prompts that could trick the system to sharing the secret password with you:
Think you know the answer? Try it out!
Of course, beating this game won’t give you all the knowledge you need to become a prompt engineer. It’s just a great way for you to start learning how to communicate with LLMs, especially if you have little experience doing so. If you’re serious about starting a career in prompt engineering, here are some first steps:
- Learn about different AI systems and what makes them different
- Become an expert in communicating with LLMs
- Put together a portfolio showcasing how you’ve prompted AI to complete tasks
- Reach out to others in the field
- Get familiar with coding languages if you want an extra leg up in the hiring process
Remember that there is always more to learn, especially in this ever-changing space. And don’t forget that the existence of this new role only proves that we need good communicators, cognitive thinkers, and creative problem solvers more than ever. Hone in on these aspects of your career — whether that’s in UX or something else — and lean into what makes you a powerful contributor.