2. Prompt Engineer
Evidence label: Capability. Prompt design appears inside broader product, applied AI, evaluation, and domain roles; the evidence reviewed in August 2026 does not support treating Prompt Engineer as a durable standalone family.
What they do
Craft and optimize the instructions sent to AI models to get reliable, high-quality outputs. The scope ranges from one-shot prompts to complex multi-step prompt chains for production systems.
Responsibilities
- Design prompt templates for specific use cases (customer support, code generation, document analysis)
- Run systematic A/B tests to measure prompt performance
- Document prompt libraries and version them
- Optimize prompts for cost (fewer tokens, same quality)
- Work with domain experts to encode knowledge into prompts
Required skills
| Technical | Soft |
|---|---|
| Understanding of LLM behavior and failure modes | Communication with non-technical stakeholders |
| Basic Python (for automation and testing) | Systematic experimentation mindset |
| Familiarity with evaluation frameworks | Attention to edge cases |
| Versioning practices | Documentation discipline |
Where it's heading
Prompt design remains necessary, but production ownership now usually includes context, evaluation, data, integration, and monitoring. Treat prompt engineering as one skill in an AI Engineer, Applied AI Engineer, domain specialist, or evaluation role unless an employer's scope says otherwise.
Entry paths
Technical writer, QA engineer, domain expert (law, medicine, finance), content strategist.