(as published on Mumbrella)
Last week, a panel at the University of Sydney’s Sydney Policy Lab debated AI. Some of the panellists make their living from teaching about writing. They promoted the craft, and they defended it well.
They were defending the wrong part of the job.
An article, a board paper, and a crisis statement are each built in four stages.
Thinking comes first: the high-level idea of what you’re trying to say and why. Research follows developing the concept, then the writing of the first draft. Then comes the polish, which takes judgement and experience.
AI is coming for that first draft.
In 2023, two MIT economics researchers gave 453 college-educated professionals a set of workplace writing tasks and gave half of them ChatGPT. The AI group took 40 per cent less time, and assessors rated their work 18 per cent higher. That was three years and several model generations ago.
The writing is becoming the cheapest part of the work. The thinking, and the judgment that decides what survives the polish, are becoming the most valuable.
I think this is a fundamental shift for everyone in the written communications business.
The Author – Peter Wilkinson
Horses didn’t disappear. They became a hobby
In 1915, over 90 million acres of American farmland went to feeding horses and mules.
Nobody rides to work. Plenty ride at pony clubs, or at the track.
That’s where writing is heading. It won’t die. It will become a pastime for the people who love it, while the work moves somewhere else.
Jobs won’t disappear either
Another point raised in passing was that AI will cost a lot of jobs. I don’t accept that. The horses left the farm, and many of the people who worked with them moved into new trades.
In a speech to a communications conference in March, I predicted that a communications firm of 20 people today will be six to eight by 2030. The other 12 won’t stop working. Their drive to work will take them into jobs built on what AI can’t supply: trust, earned through truth and transparency.
The drive to work didn’t go anywhere. The work changed.
Students need to learn to think. Writing is one way
One argument at the Policy Lab was that writing is essential to a student’s development. You don’t know what you think until you’ve written it and rewritten it.
There’s truth in that, but what develops is the thinking. Writing is one path to it, not the only one.
Another is reading how an AI interprets your argument. Give it a half-formed idea and it hands back its version of what you meant. Sometimes it’s right. Often it’s wrong in a way that shows you where your own thinking was loose.
That gap is where the learning happens.
What would your graduates learn if they spent less time drafting and more time arguing with the draft?
Last year is the wrong benchmark
Early last year, I sat down to design what a company would look like in 2030. The finished organisation: who does what, what clients pay for, and which tasks no person touches.
Then I worked backwards to the present.
Futurists call it backcasting. The Canadian academic John Robinson set out the method in 1990, in a paper subtitled a recipe for people who hate to predict. The point is not to forecast the likely future, but to decide on a desirable one and work out how to reach it. It’s now a standard tool in government foresight. The UK Government Office for Science describes it as a way of overcoming “present bias” in planning.
Most organisations do the opposite and measure AI against last year. That comparison is flattering. It tells you how much faster you’re doing yesterday’s job, and nothing about whether that job will still exist.
Rear-view thinking is flawed thinking.
The 2030 view changes the questions
In crisis communications, a holding statement that once took an hour will take a minute, or as long as it takes to speak the thought to an AI agent. The time saved doesn’t go to the writer. It goes to the decision: what’s true, what the organisation will admit, and who it faces first. Those were always the hard parts. Soon they’ll be the only parts.
Reputation management goes the same way. When anyone can produce fluent copy, fluency stops signalling competence. Readers, and the AI answer engines that increasingly brief them, will look at the record instead: what you said, what you did, and whether the two match.
If you mapped your own organisation in 2030, how many of today’s roles would turn out to be drafting roles? I think that’s the question leaders should be thinking about now.
In the meantime, the panellists will keep writing, and good on them. So will I, on Saturdays.
