Work and skills intelligence / United States

The AI job question is really a skills question.

Counting “AI jobs” captures only a narrow slice of the change. The more consequential signal is how often ordinary roles begin to require AI literacy—and which existing skills are being rewritten around it.

Published 22 August 2026 Source accessed 22 August 2026 Confidence: moderate; platform dataset

01 / What changed

Fact
≈7%Share of technical job postings represented by AI engineering roles in LinkedIn's report.
+70%Year-over-year growth reported for postings requiring AI-literacy skills.
85%Share of U.S. professionals that LinkedIn estimates could see at least a quarter of their skills transformed by AI.

02 / Why it matters

Inference

A workforce plan based only on new AI titles is likely too narrow.

The decision unit should be the task and skill bundle inside each role: what is automated, what is augmented, what still requires judgment, and what training changes the economics. That framing is more testable than a single forecast of jobs “created” or “lost.”

03 / Boundary

Limitation

LinkedIn data reflects its members and postings, not the entire U.S. labour market. A requested skill does not prove that a worker uses it, that productivity improves, or that a job will disappear. The estimates should be corroborated with occupational, employer, wage, and task-level data before a workforce decision.

04 / Evidence

05 / Next test

Progressive next step

Build a role-by-skill transition map, not a headline forecast.

For one industry, join current postings to occupational task definitions, wages, training time, and employer filings. Then distinguish observed demand, analyst inference, and scenarios.

Review the research workspace trial