Half of all US employees use AI at work now - and waste almost 8 hours a week doing it

Read the full articleHalf of all US employees use AI at work now - and waste almost 8 hours a week doing it on ZDNet

What Happened

Companies on the front lines of AI adoption are also hiring and laying off more employees on average than those that aren't, Gallup found.

Our Take

Half of U.S. employees now use AI at work, often with tools like GPT-4 or Claude, yet waste nearly 8 hours weekly on redundant tasks AI was meant to eliminate. The pattern holds across tech, finance, and customer support roles.

Wasted time stems from duct-taped AI workflows—RAG systems with no evals, agents looping endlessly, or fine-tuned models running at 3x inference cost due to poor prompt hygiene. Developers still treat prompts like scripts, not code. Stop writing one-off prompts; version and test them like any other critical logic in production.

Teams shipping AI features at scale must enforce prompt testing with Haiku-level latency budgets and track cost-per-decision. Everyone else—using AI for one-offs or exploration—can ignore this. The gap between sloppy users and disciplined builders is now a productivity chasm.

What To Do

Do enforce prompt versioning and evaluation in CI/CD instead of ad-hoc prompts because untested prompts cost 3x in latency and ops

Builder's Brief

Who

teams running RAG in production

What changes

prompt management and testing workflow

When

now

Watch for

adoption of prompt evaluation tools in CI/CD pipelines

What Skeptics Say

Most AI use at work is performative—companies hire and fire more not because AI fails, but because they lack metrics to measure real impact.

Cited By

ZDNet Half of all US employees use AI at work now - and waste almost 8 hours a week doing it