There’s a common assumption circulating in corporate comms that AI will reduce the need for editing.
It sounds logical. If a tool can produce clean, clear copy in seconds, surely that removes some of the work around writing and reviewing it.
In reality, it tends to do the opposite.
Because while AI is very good at producing sentences, it’s not very good at knowing whether those sentences are the right ones or are achieving an organisation’s goals.
1. AI still makes things up
AI is very good at producing text that looks polished and plausible.
That doesn’t mean the information is always correct.
AI systems can hallucinate information in the process of trying to generate what they think you’re looking for.
We’ve already seen multiple legal cases where lawyers submitted filings containing fake case citations generated by AI. The cases had names, dates and references. They just didn’t exist.
Those mistakes were caught because someone had to check.
Inside organisations the errors are usually less serious, but they still matter.
AI is not infallible.
It still misses things. It still invents things. It still makes suggestions that are simply wrong.
2. It can’t exercise judgement
Even when the facts are technically correct, another issue remains.
Writing inside organisations isn’t just about producing sentences that read well. It’s about judgement.
Knowing when a word matters. When a definition carries weight. When a stylistic choice strengthens the point you’re trying, even if it doesn’t follow every English language rule.
Yes, it can cut words, tidy sentences and correct grammar. It can’t exercise judgement about what the organisation wants and means to say.
3. AI increases the volume of text
There’s also a more obvious effect that shows up inside organisations.
When producing text becomes easier, people produce more of it.
This means more drafts, more documents – and more versions of the same idea start circulating.
Something that previously took hours can now be produced in minutes.
Which means more material that needs to be checked for consistency and accuracy.
Someone has to step back and ask questions that AI can’t answer:
- Are we describing this in the same way across our content?
- Does this contradict something we published elsewhere?
- Is this wording in this context implying something we didn’t intend?
- Does this content convey the right tone and fit with the organisation’s values/brand/messaging?
When definitions matter and language carries weight, those questions cannot be automated.
The editing work doesn’t disappear just because the first draft came from a machine.
AI is a powerful tool for generating text… but the hardest part of writing was rarely typing the words.
It was deciding what those words should mean, and making sure they say it clearly.
