Have you heard of prompt engineering?
This latest job is all about refining generative AI prompts so they provide accurate, reliable, and credible information. Data-led startups are asking for prompt engineers to join their teams and over the next few years, businesses across a wide range of industries will start to request them too!
What is prompt engineering, and how do you become an AI prompt engineer?
I’ll answer these questions and relay insights from some leading lights in the prompt engineering industry on how this new role has impacted the marketing sector.
Prompt Engineering Essential Insights
- Prompt engineering involves creating prompts to guide large language models (LLMs) like ChatGPT, similar to providing context for effective search engine results.
- Prompt engineering helps digital marketers make more targeted and engaging content, as well as saving time.
- SEO tools like AnswerThePublic can enhance prompt quality and resonance with the target audience.
- Experiment with different prompts through testing, allowing for variations in length, target audience focus, features, benefits, and calls to action.
- AI prompt engineers are becoming a viable career, requiring a combination of skills such as passion for AI, programming, clear communication, and marketing/psychology knowledge.
- Staying updated on AI regulations and incorporating empathy into AI content will be essential for prompt engineers.
Prompt engineering is the science of creating prompts to guide the responses of large language models (LLMs) like ChatGPT.
Think of providing a prompt to ChatGPT like searching for something on Google. The more context you can give Google, the more likely the results will match what you’re looking for. Prompt engineering works in the same way.
Ankit Prakash, founder of Sprout24, a platform that provides insights about SaaS products, including ones that generate and use AI, puts it succinctly: “Think of prompt engineering as a translator. It allows us to convert human [i]nquiries into a language AI can understand and respond to effectively.”
How Does Prompt Engineering Impact Digital Marketing?
How is prompt engineering helping people who work in digital marketing?
It helps them create more targeted and engaging content. Personalization in marketing is the key to high conversion rates. Prompt engineering makes it easier than ever for marketers to create content that resonates with the needs and pain points of their target audience.
Prompt engineering can also help digital marketers save time as they can create clear and efficient prompts faster, spending less time refining the output. This means less time spent on repetitive tasks and more time spent focusing on the bigger picture.
3 Tips for Better Prompt Engineering
If you’ve ever used a generative AI tool like ChatGPT, Bard, Midjourney, or DALL-E to create content (and three out of four organizations already do), you have given it a prompt to get started. This means you’ve already been doing some prompt engineering, even if you didn’t know it.
Here are my three expert tips to carry out prompt engineering like a pro.
Use the Right Tools
Whether you’re creating SEO-optimized content or want to ensure your output is clear, accurate, and credible, there are a wide range of tools available.
For example, take ChatGPT and SEO—you might want to ask ChatGPT to include specific keywords in your prompt to maximize the odds of your content ranking in the search engines. AnswerThePublic is excellent for this as it uses autocomplete data to tell you what people are looking for in Google, YouTube, and Bing.
This also means you can ensure the prompts you use resonate with the needs and wants of your target audience.
Why not check out my AI tools to not just enhance your prompts, but to enhance how you use AI? Here are some of my favorites:
- AIPRM. Discover a library of prompt templates and share your own with other AI users.
- ChatGPT Prompts for Developers. This tool offers pre-written prompts to inspire you to write high-quality code.
- ClickUp AI Writing Assistant. This tool gives you pre-made prompts to help you create you best content and enhance productivity.
- PromptPerfect. This plugin helps you write more concise prompts with less risk of misinterpretation.
- Prompt Perfect: This tool helps you optimize your prompts to make them more accurate and relevant.
Break Tasks Down
Imagine you’re creating a Google Ads campaign to promote a landing page on your site. The project sounds large and daunting, so you break it down into smaller chunks to make it more manageable. For example, researching the keywords, writing the ad copy, deciding which assets to use, and so on.
The same logic applies to the prompt engineering process. Let’s say you want to create a long-form blog post. The best way to approach the process is to split the blog into sections (for example, an intro, an outro, and multiple header sections) and prompt engineer these sections individually.
Another strategy is to create a detailed outline where you ask the AI tool to consider individual steps in the process. I’ll talk more about specific prompt engineering techniques later in this article.
This doesn’t just give you more control over the process but means the AI tool can create better content. AI tools have attention spans like we do – if they lose focus, it can lead to disjointed content and inaccuracies!
Try Different Things
Prompt engineering is all about experimentation. Even making small tweaks to the prompts you provide can make a massive difference to the results you get.
A large part of the prompt engineering process is testing. AI prompt engineers will experiment with different prompts and see which output people respond to best. This helps them identify which prompts are most effective.
For example, let’s say you’re a digital marketer that wants to use AI to create copy for a Facebook Ad. You can ask the AI tool to create adverts that:
- Are different lengths.
- Focus on different target audiences.
- Target different features and benefits.
- Have different call to actions.
You can then run each advert and see which one leads to the most conversions.
Prompt Engineering Techniques
With prompt engineering you get out what you put in. An ambiguous, biased, unrealistic, or unethical prompt will lead to low-quality output.
AI prompt engineers use various techniques to get the right content for their needs. Here are a few different techniques that you can apply to your own digital marketing.
(Always make sure to double check your output for accuracy and to tweak it to suit your own unique tone of voice. ChatGPT still requires human input!)
Zero-Shot Prompting
Some AI platforms have already been trained on enough data that they can carry out specific tasks without any additional clarifications or training. This is known as “zero-shot” prompting.
Zero-shot prompting is great if you need a quick answer to a question or want to see how an AI platform responds without providing any detailed context or prompting.
For example, here I ask ChatGPT when AdWords rebranded to Google Ads.
Few-Shot Prompting
Few-shot prompting is when you provide an AI platform with a couple of examples of what you want it to achieve. This provides it with context and makes it more likely to give an accurate answer.
You might also hear this technique referred to by the number of prompts provided, for example, “One-shot prompting” or “three-shot prompting.”
For example, let’s say I want ChatGPT to come up with a tagline for a marketing campaign for biscuits and provide some examples of previous taglines to help it with the overall theme and tone of voice.
Chain-of-Thought Prompting
Chain-of-thought prompting is when you guide an AI platform through a process by giving it a series of sequential prompts that it can elaborate on further. This means you can ensure the platform provides the right information in the right order.
In this example, I ask ChatGPT to put together a guide to writing a blog, by giving it a series of steps.
Least-to-Most Prompting
With least-to-most prompting, you provide the AI platform with a basic prompt to start with. If you don’t get the results you want, you give a more detailed and specific prompt until you get the answer you need. This is a helpful way of training an AI platform and understanding its limitations.
For example, I ask ChatGPT to write a product description for a doggy raincoat, and then gradually tweak the detail of my prompts until I get the result I’m looking for.
Adversarial Prompting
AI platforms often come under fire for providing incorrect or biased information. For example, an image generator was found to use harmful racial and gender stereotypes in the images it created.
You can use adversarial prompting not just to test any vulnerabilities an LLM might have but also to encourage platforms to think outside the box.
With adversarial prompting, you provide contradictory instructions or intentionally try to confuse the platform. It might sound strange, but it does deliver results.
Here’s an example. Let’s say I ask ChatGPT to write a product description for a sugary cake. I then ask it to write a product description for a healthy, sugar-free cake. This means ChatGPT gets more creative when writing the second piece of content.
Contextual Prompting
Contextual prompting is when you provide an AI platform with as much information as possible. This helps it create accurate and relevant content. Essentially, you’re providing the AI platform with a brief before it starts!
This is especially important if you’re working with a platform that does not have access to real-time data. For example, ChatGPT-3.5 (the free version) only goes as far back as January 2022.
Let’s say I want ChatGPT to create a short piece about home decor. I tell it about the latest trends so it can create something timely and relevant.
Prompt Engineering as a Career
So, if all this talk of creating and optimizing AI prompts has piqued your interest, how do you become a prompt engineer?
More and more companies are hiring AI prompt engineers. According to the 2023 McKinsey Global Survey, 7 percent of companies that have adopted AI have hired prompt engineers.
How to Become a Prompt Engineer
Prompt engineering is still a relatively new role, so there is no definitive route into it as a career just yet.
However, hard and soft skills that will benefit you if you want to become an AI prompt engineer include:
- A passion for AI, machine learning (ML), and natural language processing (NLP). You need to have a solid understanding of how these systems work and how you can use them to your advantage.
- Experience with programming languages like Python. You need to know how to organize data sets, build custom prompts, and automate tasks.
- Clear communication skills. You will need to explain your findings to people who may not understand AI and how it works.
- Marketing and psychology know-how. A marketing background isn’t essential, but it can help ensure your content is well-optimized for conversions. Likewise, a knowledge of psychology can help you understand how people use language and help identify potential biases.
Bijan Vaez, is the CEO of ProductShots.Ai, a company that uses AI to help people create innovative product photos for use on websites and social media. He advises that the ability to reverse engineer is among the most vital skills for a prompt engineer.
“You need to be able to connect with customer needs and try to work backward to how you would interpret and communicate that to the LLM. You need to follow a diamond shape of exploration, diverging out until you find a prompt structure that works well.”
Prompt Engineer Salaries and Prospects
According to ZipRecruiter, the average salary in the U.S. for a prompt engineer is $62,977. So, if you are earning this, you are in the upper half of all US earners.
However, many top companies and technical startups want the best of the best to train their AI systems, which is reflected in the salary and perks offered.
In early 2023, Anthropic, an AI safety and research company put out a job looking for an AI prompt engineer to join the team. The company offered a salary of up to $375,000, plus equity and benefits.
Prompt Engineering Day-to-day
Is there a typical day for AI prompt engineers? There isn’t.
AI prompt engineers carry out a lot of different tasks during the course of the working day, including:
- Creating, refining, and analyzing prompts.
- Communicating and documenting their findings to other teams, stakeholders, and customers in a clear, jargon-free way.
- Building Tutorials and Tools to help people understand how prompt engineering works.
- Reading up on the latest trends and AI news.
Senior AI prompt engineers may also be involved in training and mentoring junior engineers.
As prompt engineering is still a new career field, roles and responsibilities may adapt and evolve more in the future.
The Future of Prompt Engineering
There’s no doubt about it—AI is here to stay. This means that it’s essential for digital marketers to know how it works and how to get the most out of it.
It’ll certainly be interesting to see where prompt engineering goes in the future, given that it came out of nowhere at the start of 2023.
But what does the future bring for prompt engineering?
Ankit Prakash says, “As AI becomes more sophisticated, the role of the prompt engineer will shift towards ensuring AI applications remain aligned with human values and needs, all while keeping abreast of AI advancements.”
What does this mean?
It means that prompt engineers will need to ensure that AI output is ethical, accurate, and unbiased. They will need to train AI systems on diverse data, check and correct output for potential issues, and collaborate with relevant partners (for example, social scientists) to make sure content stays aligned with the right values.
Ethical AI initiatives like The Partnership on AI, The Algorithmic Justice League, and The Future of Life Institute have been set up to help prompt engineers fulfill their obligations.
As well as understanding how to add empathy and the human touch to AI content, it’s important for prompt engineers to stay up to date with how we can use it in line with government regulations.
For example, the European Union (EU) AI Act is now provisionally agreed, although it won’t officially become law until 2025. When it is implemented, sectors that use AI may need to carry out risk assessments and ensure that users have the right to access and remove data used in AI models.
FAQs
Conclusion
AI has definitely made its presence known in the world of digital marketing. For example, ChatGPT is impacting influencer marketing by helping marketers develop innovative campaign ideas and write persuasive pitches to influencers!
I’m absolutely fascinated by the work AI prompt engineers do. They research and test multiple iterations of prompts to identify the ones that generate the best possible content. Even the smallest changes to a prompt can make a massive difference!
Even if you’re not interested in prompt engineering as a career, you can use prompting basics to enhance the AI content you generate and save time on basic tasks.
What are your thoughts on prompt engineering? Would you consider a career as an AI prompt engineer? Let me know in the comments!