If Andy Burnham ever did use ChatGPT to write a speech, so what?
Government by ChatGPT is not, on the face of it, a slogan to be proud of. Last week, Andy Burnham’s No 10 team was pictured using the AI tool to improve a speech. Then came allegations that the written version of Alan Milburn’s Neet plan was doctored using AI.
While hasty denials were issued by Number 10, my advice would be to calm down. The truth is, everything is written with the help of AI these days, because nearly everything has been supplemented with material accessed via AI-fuelled search tools – even if you’re still using Google and not ChatGPT or Claude. It can be a fantastic resource for diving deeper into a subject, provided you always check the sources, of course.
Using AI, you see, isn’t the problem. Nor should people be embarrassed for using it. When used properly, it is a fantastic tool, and we would be fools to deny it. The important caveat to this is that we humans are in control and that an assumption isn’t made that AI can do the job itself.
This week, I won and lost a client because of AI. The win was from a company that has stopped investing in and relying on AI-generated material because they find the work so bland. They needed an expert. The second asked me to write a speech for an imminent PR campaign, the words for which were so obviously AI-generated it would have been embarrassing not to mention it. So I did. Normally people like my honesty. This one fired me.
These two stories encapsulate a moment. We are drowning in AI slop. Everything looks like the work of a chatbot. It’s not just bullet points and em-dashes (nor the sentences that begin with “It’s not just…”). It’s the lazy familiarity, the suspicious smoothness and superficiality, annoying word salads, lack of personality, and basic, soullessness.
Recent studies suggest that about half of everything produced every day is AI-generated. The rest is mostly AI-assisted. This has created a wonderful opportunity for us. If derivative “slop” is everywhere, humans have suddenly become essential.
There is a good reason “human” has just been named Word of the Year by Cambridge Dictionary because our skills have never been more valuable. The people behind this decision to hero the word “human” say it’s because “having the qualities, faults, and feelings that people have, as opposed to machines” is more vital than ever. Pope Leo XIV also wants us to remain “profoundly human” in the tech onslaught, and creatives are forming powerful alliances to stop machines copying their work.
Fear of being obsolete? Not so fast.
Four years ago, it was predicted people like me were destined for the scrap heap. Sam Altman’s newly launched ChatGPT could do anything I could but better, faster and cheaper. But faster and cheaper rarely means better.
I use AI all the time and so should you. AI-assisted, not AI-generated. It’s brilliant for research, for taking the “fat” out of my work that I can’t spot on my third edit, for helping me clarify an argument and summarise complex material.
Last week I ran a storytelling training session for a pharmaceutical company. I asked Claude (a far slicker partner than ChatGPT for long-document writing) to imagine a discussion between its CEO and a specific expert in the field. Material I couldn’t possibly have uncovered myself.
It helped us identify a working narrative because it pooled together existing data with both individuals’ strategic thinking and then imagined a conversation between the two. Rather than use bots to dredge up facts – some of which might be hallucinated so please do check – much better to openly partner them in clever ways to analyse vast amounts of material in seconds. Don’t let AI do the job for you and then deny it. Let it help you do the job and admit it.
Bad AI writing is easy to spot:
Using the bots isn’t a problem, but outsourcing our thinking to them is. That is what we should really be worried about. We do not sweat over a first draft to suddenly have that serendipitous moment of inspiration. We don’t cogitate on a point, walk through an argument, then make it better with an original thought that emerges after staring out of the window and daydreaming. We don’t go down dead-ends to find escape routes. If we’re not thinking critically of ourselves, stress-testing ideas, then we are not reasoning and have lost the ability to spot a flawed idea. If we are not pushing our brains to come up with better ideas and arguments. We’re not learning.
In my workshops, it’s obvious that people who love the difficult bits will be the stronger and more valuable colleagues than those who see AI as a shortcut to thinking and use it that way. They understand that utilising human skills like storytelling, creativity and critical thinking – and then supplementing their thoughts with AI – is a valuable superpower.
This is where leaders and their teams can get it so wrong. The best work is not just neat sentences. It’s also about real experience, anecdote, a connection with the audience and pacing things so that it doesn’t feel like a generic one-note speech or pitch. When one of Burnham’s aides was caught using ChatGPT to seek feedback on a draft of his speech to the Welsh Labour conference, it included a prompt asking how to make it more emotive. But you can’t just pretend to add emotion; you have to feel it. Like all the best speeches, it needs to be felt by a human if you want it to land.
I’m looking forward to the new film, Fatherland, about Thomas Mann’s post-war return to his native Germany, and his strained relationship with his family and birthplace. Mann once wrote something which is even more relevant today: “A writer is someone for whom writing is more difficult than it is for other people.” It’s also more difficult than it is for AI. Which is why we should all want to be writers. AI-supported ones, of course.
Grant Feller is the Founder of Every Rung and writes at The Storytelling Newsletter on Substack