Prompt Engineering Is Not A Job Title. Stop Hiring For It.

Every LinkedIn recruiter post. Every "Prompt Engineer — $250K + equity" job listing. Every "AI Whisperer" Substack with a $400 cohort. It is the most embarrassing hiring fad I have watched this industry cook up since "Blockchain Lawyer" in 2017. We are advertising a job that does not exist, paying six figures for a skill set that is roughly eight months from being deprecated, and confusing writing English competently with a profession.

Prompt engineering is not a job title. It is a thing software engineers already do, badly, every day, in code review comments and commit messages. We did not invent a role for it.

The Job Description Is A Self-Own

Read any of the listings. "Design, test, and refine prompts to optimize LLM outputs across production systems." Translation: you will write sentences into a text box and pray. There is no formal method. There is no reproducible process. There is no body of theory. There is a vibes-based craft where someone rewrites the system prompt on Tuesday, the eval moves by three points, and Slack goes wild. Then they rewrite it again on Friday and the number drops back. We are paying $200,000 a year for someone to A/B test system prompts in a text file.

Every team I have worked with that hired a dedicated "prompt engineer" in 2024 has either laid them off, renamed the role "AI Engineer," or quietly absorbed the function into the existing engineering org by Q1 2026. The role is a rebrand of "the person on the team who was good at English." That person is now called a Staff Engineer, and they have done this job for thirty years without a special title.

The Skill Ceiling Is Depressing

I can teach a competent engineer to write a better system prompt in two days. So can you. So can any of the ten thousand blog posts with "47 prompt engineering techniques" in the title. The craft is shallow. There is no equivalent of Algorithms or Designing Data-Intensive Applications for prompt engineering because the surface area is small enough to fit in a Notion page.

The actual hard parts of shipping production LLM systems are retrieval pipeline design, eval harness construction, tool-calling schema discipline, latency budgeting, token cost modeling, and failure-mode analysis. None of those are "prompting." All of those are engineering. Hiring a "prompt engineer" tells me your org does not understand what is actually hard about shipping LLM products. You are staffing the easy layer and hoping it covers for the missing layers.

The Vendors Are Writing The Job Descriptions

OpenAI, Anthropic, and Google all have prompt-engineering-as-marketing departments whose entire job is to keep the role alive in your org chart. Every model release ships with a 90-page prompting guide. Every conference has a "PromptCon" with $1,200 tickets. The reason is structural: if prompting is a profession, model quality is the user's responsibility, not the lab's. Anthropic does not want to own the failure when your chatbot hallucinates — they want a "prompt engineer" in your org chart to blame. The vendor incentives and your incentives are misaligned by design, and you are paying the bill.

The Role Is Being Deprecated By Tools In Real Time

DSPy exists. Instructor exists. BAML exists. Guidance exists. Outlines exists. The tooling that replaces hand-tuned prompting is shipping every six weeks. By the end of 2026 you will not write a system prompt by hand for any non-trivial task — you will compile it from a typed schema and an optimizer that runs against your eval set. The job title is a countdown timer. Hiring into it in 2026 is like hiring a Photoshop retoucher in 2014 and expecting them to be relevant by 2018.

What To Do Instead

Hire an AI engineer. Hire someone who can build an eval harness, reason about retrieval, design a tool-calling schema, debug a streaming response, model token cost, and also write a decent system prompt. Pay them like a senior backend engineer. Stop splitting the function into a special role.

If your company has "Prompt Engineer" in its org chart, the next reorg is the right time to retire the title. Promote them to AI Engineer. Give them a real scope of work. Stop pretending that "writing English into a chat box" is a profession.

We tried it. It did not work. Move on.