Remember when AI was going to give us all our afternoons back? It turns out many of us have simply swapped one job for another: fixing what the AI wrote.
A fresh wave of workplace surveys points the same way. AI is fast, but its first draft often needs a lot of tidying, and that tidying is eating a surprising amount of the time it was supposed to save.
The numbers behind the groan
- Almost half the time. A BambooHR survey of 1,600+ US workers, reported by HR Dive, found people spend about 1.5 hours a day using AI. Nearly half of that goes on troubleshooting errors and re-working prompts, roughly 20 working days a year.
- Half a working day a week. Zapier’s survey of 1,100+ US AI users put the cleanup at about 4.5 hours a week, and only 2% said they never have to revise AI output.
- Starting over. A new study by Use.AI, covered by Dimsum Daily and based on 8,316 AI-using workers, found 43% have had to redo an AI output entirely, and 39% said fixing it took longer than doing the job from scratch.
The “workslop” problem
Researchers have a name for the polished-looking but half-baked material AI can churn out: workslop. The worst part is that it travels. In the Use.AI study, 34% of workers admitted passing AI-made material on without checking it, and 58% said their workplace has no rules for reviewing it first. So one person’s lazy prompt becomes a colleague’s afternoon of repairs.
A fair reality check
Before you bin your subscriptions, note that the headline claim, that workers spend more time fixing AI than doing real work, is stronger than these surveys actually show. They mostly measure how much time goes on fixes, not whether it outweighs the gains. In fact, 92% of Zapier’s respondents said AI boosts their productivity, and 62% in the Use.AI study found it made drafting easier. AI is genuinely useful. It just isn’t free.
How to stop being your AI’s proofreader
The Zapier data holds the most useful clue: training makes a huge difference. Among trained workers, 94% say AI helps them, against 69% of untrained ones. Four habits worth adopting:
- Give it context. Feed it your style guide, templates and examples, not a one-line request.
- Choose the right jobs. Drafts and summaries suit it well. Numbers, analysis and anything customer-facing deserve extra scrutiny.
- Check before you forward. One minute of review saves your colleagues an hour.
- Write the rules down. If your team has no policy for reviewing AI-assisted work, you are in the 58%.
The AI Boom Team take
AI is like a brilliant intern who works at lightning speed and never admits to being unsure. You would never let that intern send a client email unsupervised, and the same applies here. The businesses that win will not be the ones using the most AI. They will be the ones who have taught their people to use it well.
Sources: BambooHR via HR Dive (2 September 2026), Zapier via IT Pro (January 2026), Use.AI via Dimsum Daily (20 October 2026). These are survey findings, not controlled experiments.
By the AI Boom Team. © aiboom.uk