Reporting archive
Evidence on AI and work
Reporting archive

San Francisco ended its teachers’ strike with a new rule of the road: AI can help, but it can’t cut headcount.

Microsoft’s AI chief just put a 12–18 month timer on white‑collar work, as agents take over tasks and careers get rewritten as policy files and guardrails.

The hottest signal for where AI comes online isn’t GPUs—it’s a 20% raise for pipefitters and a 112% surge in drywall hires.

An inside-the-firm study shows generative AI didn’t give time back—it tightened cycles, blurred roles, and quietly taxed focus.

The memo says “focus,” but the budget says you were traded for compute—here are the three doors and how to pick the one you can still walk through.

A central banker just turned ‘AI training’ into forward guidance, arguing that shrinking entry routes and rising AI task loads demand urgent, operational skilling as the economy’s pressure valve.

The shock isn’t layoffs but the entry-level offers that never print—and a two-track plan aims to reopen the on-ramp via AI fluency and upgraded human-centric work.

Goldman is moving Claude from helper to doer in reconciliations and KYC, and if exceptions don’t rise, the rungs of Wall Street’s apprenticeship start to disappear.

January’s ledger put AI-linked layoffs at 7% and exposed how much of the rest is de-layering and market-friendly storytelling, not robots at the gate.

IBM’s Arvind Krishna just turned “half” into a blueprint: automate the repeatable, prove it with audit trails and KPIs, and steer humans toward exceptions and relationship repair.