AI is not what makes our clubs sound the same
2026-07-30
“At 6:30 this morning, Nomsa carried 40 litres of water home before her children left for school.”
That sentence, and the reason it works, sat at the centre of a district Public Image session on ethical storytelling in the age of AI, presented by Sarah van Heerden.
The argument is blunt. AI is not making organisations sound alike — people are. A model learns patterns, so when every club types the same instruction, “write a Facebook post about…”, every club gets the same paragraph back. The sameness is not in the software. It arrives when nobody does the thinking first.
Set Nomsa against the opening most of us have already written: “Today we celebrated another incredible milestone as our amazing volunteers came together…” Everyone has seen it. Nobody remembers it. The difference is not vocabulary. It is that somebody was there, watched what happened, and wrote down the part that was true.
Four questions, then the tool
Before opening an AI tool at all, the session put four questions on the table: What happened? Why does it matter? Who changed? Why should anyone care? Great prompts cannot rescue poor thinking — answer those four badly and no amount of instruction will save the result.
Where AI does earn its place is in the shaping. A prompt that sets a word count, bans the tired phrases, demands an opening line that cannot be skipped and asks for a closing question produces something usable, because a person decided all of that.
The part that does not transfer
If AI invents something, the software is not accountable. The organisation is. The writer is. The communicator is. That responsibility never moves, which is why the five rules from the session start with never publishing what you cannot verify, and why the second is simply that AI cannot witness. It was not at the handover. It did not meet Nomsa.
There is a practical half too: stop writing everything everywhere. A magazine article carries the whole story in 350 to 450 words. Facebook gets the first paragraph and no more than 200. The website carries the rest. People stop reading because we ask them to read 800 words in the wrong place.
The three working guides from the session are now in the knowledge base: the thinking that comes first, the five rules, and writing for the channel.
AI can predict words. Only people can recognise what is true. So the question the session leaves every club with is a plain one: when somebody reads your next post, will they be able to tell it was your club that wrote it?
