Since the boom in popularity of generative AI, new terms and jobs have emerged, one of them is “prompt engineer”.
So, what is prompt engineering and what does a prompt engineer do?
Darin Stewart, VP analyst at Gartner told Digital Nation a “prompt” is simply the instructions provided to a large language model (LLM) or other AI when instigating an interaction or making a request.
Newsletter Signup
Get the latest insights and analysis delivered to your inbox.
SIGN UP
I have read and accept the privacy policy and terms and conditions and by submitting my email address I agree to receive the Digital Nation newsletter and receive special offers on behalf of Digital Nation, nextmedia and its valued partners. We will not share your details with third parties.
“When provided directly by an end user, prompts are usually natural language requests such as ‘list things to do in Portland, Oregon’ or ‘write a 500-word article on generative AI in the style of Bob Woodward’,” he explained.
“Information to be considered in responding to the request along with instructions or examples of how to carry out the task may also be included.”
The prompt is the instructions to the AI of what you want it to do, Stewart explained.
“When you give it a task with no further information or guidance, such as asking a question like ‘what is the airspeed velocity of an unladen swallow?’ you are using a zero-shot prompt. You can also provide examples of the type of response you want,” he said.
Giving a single example is a one-shot prompt and multiple examples are a few-shot prompt.
“There are other styles such as chain of thought, where you lead the AI through the process of how to solve a problem or execute a task. You can also include information for the AI to use,” he said.
“For example, you could upload a long news article as part of the prompt and ask for it to be summarised according to a specific format you also include in the prompt.
“This is called in-context learning and does not change the underlying AI in any way (no training or tuning) and will be forgotten by the AI as soon as the current interaction session is concluded.”
Who performs prompt engineering?
To be a prompt engineer, it doesn’t need a degree from a university. Stewart said anyone who asks a question or interacts with an LLM or LLM-based service like ChatGPT is effectively doing prompt engineering.
However, he said if the prompt engineer does not understand how to structure their prompt, what information to include, or what examples to provide, their results will usually be “dubious” at best.
Stewart said prompt engineering is a new skill that is rapidly growing into a distinct discipline.
“Professionals with at least a basic understanding of generative AI, a good understanding of the information resources they can (or cannot) use in a prompt, and above all the ability to clearly express their request, will be best positioned to act as prompt engineers,” he said.
Like any other position, Stewart explained that someone claiming to be a prompt engineer should have demonstrable and verifiable prior experience.
“They should be able to demonstrate their ability by creating a prompt to address a scenario provided by the prospective employer,” he said.
“There are a lot of people claiming the title of prompt engineer who have no expertise in the discipline, and this can be extremely dangerous for the companies that hire them.
“Similarly, there are a lot of prompt engineering boot camps and training programs popping up online that appear to be written by the people who failed the interview for a prompt engineering job.”
Business benefits of prompt engineering
Prompt engineering is applicable to any role or job that wants to leverage generative AI to accomplish a task, according to Stewart.
“If a company is going to interact with generative AI and incorporate it into their business processes (and they will whether they know it or not), they need someone who is skilled at prompt engineering,” he said.
“All knowledge workers should have at least a rudimentary understanding of prompt engineering.”
Stewart said generative AI has the ability to free up professionals from many of the mundane and repetitive tasks required of their roles.
“It can enable workers to focus on the aspects of their job that require insight, expertise and creativity,” he said.
“A prompt engineer can create prompts to facilitate these interactions and provide them to the enterprise as a whole, so non-specialists can apply them to their own work without having to become an expert in generative AI or reinventing the wheel by creating prompts for common or recurring tasks.”
He added, “The prompt engineer can also help protect the organisation from many of the potential hazards of working with LLMs and generative AI.”
