AI in job descriptions – January 2026

LLM & tooling signals

Although mentions remain under 1% of job descriptions overall, generative AI (1.0%) and LLMs (0.7%) lead tooling language, showing employers increasingly expect baseline familiarity with these technologies. The presence of vendor names (OpenAI, ChatGPT, Copilot, Claude) alongside techniques like RAG and prompt engineering signals job specs are shifting to include building and integrating retrieval-augmented pipelines, prompt-design skills, and oversight of agentic assistants — effectively blending AI tool operation, workflow orchestration, and product/engineering responsibilities.

• Generative AI: 3,572 (1.0%)
• LLMs: 2,644 (0.7%)
• OpenAI: 971 (0.3%)
• RAG: 851 (0.2%)
• ChatGPT: 837 (0.2%)
• Copilot: 528 (0.1%)
• Prompt Engineering: 465 (0.1%)
• Claude: 185 (0.1%)

Agentic architectures

Although agentic concepts are infrequent in job descriptions (Tool Orchestration 362, Agentic Frameworks 301, Model Context Protocol 232, Agent-to-Agent 150, Agent Patterns 147 — all ~0.0–0.1%), they point to emerging expectations for orchestrating tools, building agent frameworks, and formalizing model context protocols. Employers are beginning to request skills in multi-agent coordination, LLM-to-tool integration, and agent pattern design — a small current footprint that’s likely to expand as agentic workflows become productionized.

• Tool Orchestration: 362 (0.1%)
• Agentic Frameworks: 301 (0.1%)
• Model Context Protocol: 232 (0.1%)
• Agent-to-Agent: 150 (0.0%)
• Agent Patterns: 147 (0.0%)

Category leaders

The data shows AI tooling, LLM workflows, and agentic patterns are now embedded across job descriptions — most intensely in machine-learning-and-ai-engineering (92%) and substantially in data, game, cloud, hardware, and even unexpected areas like venture capital (~48%), indicating broad cross-functional adoption. As a result, roles increasingly emphasize AI-tool integration, prompt engineering, agent orchestration, monitoring/evaluation, and governance, shifting work from manual execution toward higher-level orchestration, safety oversight, and human–AI collaboration.

• machine-learning-and-ai-engineering: 1,557/1,690 roles (92.1%)
• game-development-engineering: 84/146 roles (57.5%)
• data-engineering-and-infrastructure: 1,998/3,837 roles (52.1%)
• hardware-design-and-verification: 488/1,010 roles (48.3%)
• venture-capital-and-entrepreneurship: 63/132 roles (47.7%)
• compensation-and-benefits: 207/452 roles (45.8%)
• cloud-computing-engineering: 260/577 roles (45.1%)
• growth-marketing: 243/549 roles (44.3%)
• embedded-software-engineering: 480/1,155 roles (41.6%)

Low impact categories

These categories largely involve hands‑on, safety‑critical, or highly regulated work and often lack the digital interfaces or task modularity that make LLMs and agents useful, so employers haven’t added AI‑related responsibilities to those job descriptions. Where AI tooling does appear, descriptions increasingly emphasize AI literacy, human‑in‑the‑loop oversight, prompt‑engineering and governance — framing agents as assistants that require supervision, compliance awareness, and coordination rather than full automation of manual tasks.

• hospice-and-palliative-care: 0/1,178 roles (0.0%)
• surgical-and-sterilization-technicians: 0/1,154 roles (0.0%)
• janitorial-maintenance: 0/250 roles (0.0%)
• respiratory-and-pulmonary-care: 0/933 roles (0.0%)
• environmental-engineering: 0/451 roles (0.0%)
• home-health-aide: 0/2,627 roles (0.0%)
• healthcare-coordination: 0/1,643 roles (0.0%)
• electrical-and-electronics-technician: 0/357 roles (0.0%)

Top skills

Category leaderboard