“Prompt Engineer” Sounded Like the Job of the Future: What Happened?

Only a few years ago, prompt engineering was being described as one of AI’s hottest new careers, with eye-catching salaries and suggestions that knowing how to talk to a chatbot could become a profession of its own. The skill survived. The job market evolved differently.

The prompt engineer is not dead, but it also did not become the enormous standalone profession that some early headlines predicted. In 2023, one widely discussed Anthropic vacancy for a Prompt Engineer and Librarian offered compensation reaching $335,000, helping turn the title into a symbol of the generative-AI job boom.

By 2025, however, a research analysis of 20,662 LinkedIn job postings found only 72 prompt-engineer positions—less than 0.5% of the jobs examined. That study is a preprint and represents a snapshot rather than the entire global labour market, so its figure should not be treated as a universal measurement. But it illustrates an important shift. Companies still need people who know how to instruct AI systems effectively. Increasingly, they also expect those people to understand software, workflows, evaluation, data, security or a particular business problem. The prompt became a skill, it did not necessarily become the whole job.

Why everyone suddenly wanted to become a prompt engineer

The excitement made sense in early 2023. Generative AI systems were powerful but unpredictable. Small changes in instructions could produce dramatically different results. Someone who understood how to structure those instructions seemed capable of extracting considerably more value from the same model. Media coverage amplified that opportunity. Anthropic's unusually high-paying vacancy became one of the most prominent examples, and a new career narrative emerged: learn how to write excellent prompts and companies may pay heavily for that expertise.

It sounded unusually accessible too. Unlike machine learning, prompt engineering appeared to require less advanced mathematics or computer science. Strong writing, logical thinking and experimentation could produce noticeably better AI responses. Those abilities remain useful, the mistake was assuming they would necessarily remain scarce enough to support a huge standalone profession.

AI itself became easier to instruct

One reason the role changed is that the models improved. OpenAI's current prompt-engineering guidance still recommends clear instructions, relevant context and carefully defined output formats. But it also makes an important observation: newer, more capable models tend to be easier to prompt. Users increasingly do not need to memorise elaborate collections of secret phrases to get a useful response.

Models can understand ordinary language better, ask follow-up questions, work from examples and reason through more complicated instructions. Prompting therefore remains important, but basic prompting is becoming easier for ordinary workers to perform themselves. That weakens the case for hiring a separate specialist simply because they know how to phrase requests cleverly.

The work became more technical, not less valuable

There is another side to the story. Production AI is considerably more complicated than writing one impressive ChatGPT prompt. A company building an AI customer-support system needs to know whether the system works consistently across thousands of requests. Someone has to evaluate incorrect answers, design instructions, connect external data, test edge cases, control costs, protect information and determine when a human needs to intervene.

The 2025 prompt-engineering study found that employers sought more than prompt-writing ability. The roles it examined also emphasised AI knowledge, communication and creative problem-solving. Current job listings carrying prompt-engineering titles similarly often combine the skill with NLP, programming, model evaluation or workflow development.

So the more durable version of prompt engineering looks less like:  “Find the magic words.”

and more like:

“Design and test an AI system that reliably accomplishes a useful task.”

That is a much broader profession.

Employers are hiring around AI in different ways

The wider AI labour market did not disappear while the prompt-engineer title remained niche. Microsoft's 2025 Work Trend Index found employers considering roles such as AI trainers, data specialists, security specialists and AI-agent specialists as companies redesigned work around AI. Its 2026 Work Trend Index goes further, arguing that AI is increasing the importance of judgment, clarity of intent and work design. That tells workers something important. Knowing what to ask an AI system is useful. Knowing what problem should be solved, how to check the answer and what should happen next is more valuable.

What this means for African workers

For African professionals entering AI, the lesson is not to abandon prompt engineering. It is to avoid building an entire career plan around the title. Prompting can strengthen many existing careers. A software developer can use it while building AI applications. A cybersecurity analyst can combine it with AI-security testing. A lawyer can use AI effectively while developing expertise in privacy, regulation and governance. A marketer can build repeatable AI-assisted research and content workflows. A data analyst can use models to explore information while retaining the statistical knowledge required to recognise a bad answer.

The valuable combination is increasingly: domain expertise + AI capability. 

That is particularly relevant in markets where dedicated AI vacancies remain more limited than general professional roles that are gradually incorporating AI.

Be cautious about courses selling the title rather than the skill

The prompt engineering boom also created a large training market. There is nothing wrong with learning prompting. Clear instructions, context management, examples, structured outputs and evaluation are genuinely useful techniques. The problem begins when a course implies that a short certificate automatically qualifies someone for a highly paid profession.

Employers ultimately pay for useful outcomes. Someone who can demonstrate an AI workflow that saves a business time, improves research or solves a customer problem has stronger evidence of capability than someone whose portfolio consists primarily of prompt templates.

Learn prompting. But connect it to something harder to commoditise.

Our Recommendation

Prompt engineering was neither completely fake nor the guaranteed career revolution that early excitement sometimes made it appear to be. It became something more ordinary and arguably more important.

It became part of how people work with AI.

For students and professionals, that changes the strategy. Do not spend all your effort trying to earn the title Prompt Engineer. Learn how models respond to instructions. Learn structured prompting, evaluation and AI limitations. Then combine those capabilities with programming, cybersecurity, data, design, finance, law, marketing or another field where you can solve real problems.

The future may contain some dedicated prompt engineers, but for far more people, prompt engineering will simply be one valuable skill inside a bigger job.

Verification Links

• OpenAI — Best Practices for Prompt Engineering
• OpenAI — Prompt Engineering Best Practices for ChatGPT
• Research Preprint — Prompt Engineer: Analyzing Hard and Soft Skill Requirements in the AI Job Market
• Microsoft — 2025 Work Trend Index
• Microsoft — 2026 Work Trend Index
• Microsoft Africa — 2025 Work Trend Index and Emerging AI Roles