For about a year, "prompt engineer" was the hottest job in tech. Listings floated pay north of $200,000. Breathless articles promised that whispering the right incantations to a language model was a new technical specialty, and everyone rushed to trade lists of magic phrases—"act as a world-class expert," "take a deep breath and work step by step."
Then the job quietly faded. By 2025 the standalone "prompt engineer" title was thinning out on job boards even as the underlying work spread into nearly every role.1 That's the tell. A genuine, durable technical skill doesn't evaporate as a job while becoming more common as a task. Something had been mislabeled.
Watch what actually produces a good result from an AI. You have to understand the task well enough to describe it. You have to supply the context the model doesn't have—who it's for, what it's really trying to accomplish, what "good" looks like here. You give it an example or two. You look at the first attempt, see what's off, and say specifically how to fix it. And when you catch yourself making the same correction twice, you write it down as a standing instruction so you never have to make it again.
None of that is engineering. That is management.
You're not prompting. You're managing.
Working with an AI is remarkably like managing a bright, fast, eager new hire who has no context, boundless energy, and no memory of yesterday. What you get back is only as good as the brief you gave. Hand that person a vague one-liner and you'll get something plausible and wrong. Hand them a clear objective, the relevant background, an example of what you're after, and a quick round of feedback, and the work comes back excellent.
Every capable manager already knows this pattern: brief, review, correct, and turn the correction into a rule. The people who are good at directing people turn out to be good at directing AI, for exactly the same reasons—and the ones who were never good at handing off work struggle with the model the same way they struggled with their teams.
The best AI users are the best communicators
So here's a prediction I'd bet on: the people who get the most out of AI won't be the most technical. They'll be the best communicators. They can take a fuzzy intention and lay it out in a structured way—here's the goal, here's the context, here's what matters, here's what I need back—and be precise about the ask. That skill has a name, and it isn't "prompt engineering." It's clear thinking, made explicit.
The flip side is familiar, because we've all done it. You're busy, you fire off a two-line prompt without really thinking about what you're asking, you get mush back, and your first reaction is that the AI just isn't very good. Usually the AI was fine. We simply hadn't done the work of saying clearly what we wanted—the same reason a rushed, half-explained request to a colleague comes back wrong. It isn't a character flaw; it's the cost of not slowing down to think. But sloppy input gets you sloppy output, and skipping the thinking on the way in is a reliable way to manufacture workslop on the way out.
What to do about it
Stop hunting for magic phrases, and stop trying to hire a prompt engineer. Build the skill that actually matters: the ability to specify a task clearly, supply what's needed to do it well, and give feedback that improves the next attempt. It's learnable. Better still, it's the same skill that makes you better with people—so the practice pays off in both directions at once. If you want to get good at AI, get good at explaining what you want.
It never belonged in the engineering school
The name did real damage. Calling it "prompt engineering" sent everyone looking for a technical answer to what was never a technical problem. The abilities that separate the people who thrive with these tools—defining the task, framing the context, communicating precisely, giving useful feedback, turning lessons into process—are the ones we've always taught under a different heading.
I spend my days teaching those skills to MBA students, and I've come to think that's exactly where this supposedly new one lives. Prompt engineering never belonged in the school of engineering. It belonged, all along, in the school of management.
Footnotes
-
The standalone "prompt engineer" title—hyped in 2024 with $200k+ listings—declined sharply on job boards by 2025 even as prompt-related skills spread into many roles. See, e.g., Salesforce Ben, "Prompt Engineering Jobs Are Obsolete in 2025—Here's Why": https://www.salesforceben.com/prompt-engineering-jobs-are-obsolete-in-2025-heres-why/ ↩