The prompt engineer role in 2026 has evolved from a standalone job into a competency inside broader roles: AI engineers, eval engineers and domain experts now own prompting as part of context engineering, designing what the model sees, not just what it's told. Hiring a pure prompt-writer is rarely the right move anymore; hiring people strong at this layer absolutely is.
Then vs now
| 2023 | 2026 | |
|---|---|---|
| Job form | Standalone 'prompt engineer' title | Competency inside AI engineer, eval and domain roles |
| Core work | Wording tricks, clever phrasing | Context engineering: retrieval, structure, tools, system design |
| Validation | Eyeballing outputs | Eval suites gating every prompt change |
| Who does it | Generalist prompt-writers | Engineers plus domain experts closest to the problem |
Who to hire for this layer now
- AI engineers who treat prompts as versioned, tested artifacts in a pipeline.
- Eval-minded people who can prove a prompt change helped.
- Domain experts (legal, medical, support) who encode expertise into context and rubrics.
- Skeptics of magic: the best candidates say 'it depends on the eval' a lot.
