Give the machine the paperwork. Keep the judgement

Professional writers should use AI. There, we said it. Half the industry may now denounce us as traitors; the other half will ask the machine to write a celebratory LinkedIn post.

Use it to compare versions, summarise background material, turn a transcript into searchable notes, test whether an explanation survives contact with a lay reader and find the unanswered questions hiding in a pile of documents. Use it to do the clerical work around thought.

Do not confuse that with thinking.

A proposal answer can contain every required term and still fail because it does not understand the buyer’s anxiety. A case study can be factually accurate and still be useless because the result has no baseline. A thought-leadership article can be polished and still have no thought in it. The machine cannot rescue a writer who has not decided what the reader should believe, why they should believe it and what they should do next.

This gives us a workable division of labour. The machine handles recurrence: sorting, comparing, formatting, checking, compressing and generating possibilities. The writer handles consequence: deciding what matters, testing claims, choosing evidence, recognising risk, establishing tone and accepting that their name belongs on the result.

The fashionable phrase is “human in the loop”. It makes the human sound like a safety catch clipped onto the important machinery. Reverse it. Put the machine in the human loop. Let it work inside a process owned by someone with context, standards and skin in the game.

Otherwise you have not automated writing. You have automated the production of things somebody still needs to read.