A recent McKinsey discussion estimates that currently available technology could automate more than half of US working hours. This is an estimate of technical capability—not a prediction that half of all jobs will disappear.
The more important issue is how work changes.
AI can produce reports, recommendations, communications, and other materials much faster than people can review them. McKinsey offers a useful example: an AI agent might generate 5,000 reports overnight, while the organization lacks the human capacity to check their accuracy or decide what should happen next.
For public services, this creates a new workforce bottleneck.
An agency may save time producing information but create additional work validating it, correcting errors, explaining decisions, and responding when something goes wrong. If review capacity is not included in the design, “human oversight” can become little more than a promise.
The International Labour Organization similarly finds that AI is more likely to transform many jobs than eliminate them entirely. OECD research finds that results depend on the task, worker experience, and quality of human–AI collaboration. The NIST AI Risk Management Framework emphasizes defined responsibilities, testing, monitoring, and risk management.
The practical control point is straightforward: before introducing AI, map who will review its work, how much time review requires, what authority reviewers have, and how uncertain or harmful outputs will be escalated.
AI does not remove the need for human judgment. It can increase the volume and speed of work that requires judgment.
This link was selected because it moves the workforce discussion beyond simple predictions of job loss. Readers should notice that AI capacity and organizational capacity are not the same thing.
Primary source: McKinsey, “The rise of the human–AI workforce”
GPT-assisted draft developed and reviewed for Public Services Alliance.