TIME called prompt engineering “one of the hottest tech jobs” as AI language models like ChatGPT are trending big. A prompt engineer’s job is to write prompts and “feed” them into AI models in order to more effectively train the models to produce useful, accurate output. It’s about figuring out what the AI model needs to “learn” and tweaking the prompts to better train it for effective outcomes.

What Is a Prompt Engineer?

“[AI language models] frequently make mistakes, either failing to comprehend the query correctly or failing to recognize that the query is underspecified, making it impossible to provide a proper response,” Victor Sanh, a researcher at the AI company Huggingface, explained to Mashable. A prompt engineer can help refine its output, providing prompts and queries to help nudge the bot in the right direction.

Despite the “engineer” in the title, prompt engineers aren’t necessarily engineers in the traditional sense, and many don’t have a tech background at all. Instead, they’re people – from a wide variety of backgrounds – who have learned the “tricks of the trade” for manipulating AI chatbots into producing a desired output. For now, at least, prompt engineering could be big business: nearly 30% of companies in one survey say that they’re looking for prompt engineers right now, and about 25% of those companies would offer starting salaries of $200,000 or higher.

Because the technology is so new, there isn’t necessarily a specific credential hiring managers can look for. An entire cottage industry of prompt engineering courses has sprouted up all over the internet, but the quality and effectiveness of these trainings can vary. It’s more likely to appear in the “skills” section of a candidate’s resume, and it’s something that would require more in-depth questioning to understand the extent of that individual’s skills.

Managing AI Tools Effectively

A skillful prompt engineer can be critical to gaining a competitive edge and using AI tools, like ChatGPT, without encountering many of the most common pitfalls.

Plenty of examples have gone viral of chatbots spectacularly melting down or providing blatantly wrong information. It’s gone beyond just viral jokes, too. A legal case made headlines in May 2023 when a lawyer cited numerous cases – only for opposing counsel to discover that the cases didn’t exist because ChatGPT made them up when the lawyer attempted to use the tool to “help” him with research and writing.  Because AI chatbots are, essentially, just gigantic predictive-text tools, these kinds of errors are not unlikely to happen.

Prompt engineers are one attempt at addressing these issues, by having humans with deeper knowledge “train” the bots so that they will respond properly. Different companies may manage prompt engineering jobs differently: some may be more like language specialists, dedicated to finding the right words to produce the right output, while others may add more technical and algorithmic work to their roles.

There’s no doubt that AI chatbots have taken the business world by storm in recent months, and the rise of the prompt engineering job is just one of many ways in which the effects are being seen. Still, AI is in its very early days, and anything can happen. Chatbots and image generators still produce glaring mistakes in many cases, and legal and ethical questions remain about consent and the use of copyrighted materials to train AI. For now, prompt engineers are a clear way that many companies are looking to hedge their AI bets, by combining human intuition and knowledge with the speed and efficiency of AI. As with hiring for any new, high-tech roles, the smartest companies will be the ones that can be most flexible as the world continues to evolve.

By Daniel Midoneck