Checked work: who checks what AI produces

If someone has sent you this: they’d like to agree with you how work that AI helped produce gets checked before it goes out. Tick what you’ve done, and talk about the rest.

Last updated

Download the PDF
Email it

Tick it on your own, or each of you separately and then compare.

What AI helps produce

How it’s checked

When something gets through

Why these points are important

What AI produces reads as finished whether it’s right or not. In my experience that’s the main reason mistakes get through: nothing about the work signals that it needs a second look.

What most often goes out unchecked

  • Replies to customers.
  • Figures in a proposal.
  • Summaries of a meeting or a document.
  • Translations.
  • Job adverts and other public text.

These are the routine pieces, which is where most of the volume is.

Why the maker shouldn’t be the only checker

Whoever asked AI for something reads the result as what they meant to get. A second reader sees what’s actually there.

What I’d do first

I’d take ten pieces from last week that AI helped with and read them in one sitting. The pattern of what needed changing usually shows where the checks should go.

When you’re the one checking

It helps to say when there isn’t time to check properly. A check that’s known to be quick is safer than one everyone assumes was thorough.

Where coaching fits

People at this point often use coaching to decide how much checking the business can afford, and what they’d accept going out with a mistake in it. I won’t tell you how to set up your checks. I’ll ask the questions that show which mistakes you could live with and which you couldn’t.

If AI is already changing your business, AI Checkpoint is a free reflection on what’s been decided, and who’s checking: simonbowen.com/reflections/ai-checkpoint