I’ve come across numerous articles discussing the potential emergence of “AI prompting” as a distinct profession in the near future. I’ve also noticed a surge in courses aligning with this trend.
I respectfully disagree.
In my opinion, prompt engineering will become a skill, akin to writing. The growth of AI is undeniable, and its influence will span across many industries. Hence, I feel that a proficient professional should be adept at creating effective prompts in order to get the best output. This ability will be assessed as a valuable skill.
While there may be dedicated roles centered on prompt creation in the immediate future, given the ongoing evolution of this field, I foresee it ultimately becoming a standard skill set for professionals.
Disclaimer: I am talking for the prompts that users enter within the UI of the LLM, not the prompts that are used by developers in the training of an LLM or in the development of software applications.
Update 2025:
Prompt writing MUST become a skill, akin to writing.
Update 2026:
In 2023 I said prompting wouldn’t become a profession. In 2025 I said it must become a skill.
Today I think it’s not even that.
The models got too good. You can write a bad prompt, a lazy prompt, half a sentence with a typo, and still get a useful answer. They infer what you meant. They ask when they’re not sure. They fill the gaps you left.
All the tricks we collected are dead. “Act as a senior marketer.” “Think step by step.” The magic templates. The prompt libraries people sold as courses. The model doesn’t need them. It barely notices them.
The real skill was never prompting. It was reasoning.
It was then. It is now. And I believe it will stay that way.
Breaking a problem down. Knowing what you actually want. Knowing which question to ask. Spotting when an answer sounds right but isn’t. Deciding what to do with it.
A good prompt was only ever reasoning written down. The model no longer needs the writing to be good. It still needs the thinking behind it to be good, and it can’t do that part for you.
So my position today: prompting is a commodity. Reasoning isn’t.
The disclaimer still stands. Building AI systems (agents, pipelines, products) is a different job. That’s engineering, not prompting.
