Strategists and creatives – AI can’t take your job. But it might make you suck at it

Mark Diamond, ‘hopeful saboteur’ at London-based brand agency Saboteur, offers a razor-sharp rebuttal to the idea that AI is set to replace brand designers anytime soon. But, as Diamond argues, that doesn’t mean the industry will remain unscathed.

When ChatGPT first woke us all up to the sheer power of generative AI, I was as frightened as anyone. “It’s like the Industrial Revolution,” I thought. “Except this time technology isn’t replacing horsepower, it’s replacing brainpower”.

Scrolling down my LinkedIn news feed, I realised I wasn’t alone. “The entire profession of copywriting may be on the edge of obsolescence,” wrote one brand industry leader whom I follow and admire. I’m a brand consultant and not a copywriter per se, but writing copy is a significant part of my job. I went for a lie down.

Later, in a Hollywood-style flashback, I recalled being dismissive to a friend – who knows a lot more about technology than I do – when he said, “AI will come for the creative jobs first”.

I started to panic. I remember my wife – herself a brand consultant – accusing me of catastrophising. I told her that I’d remind her of this once the robots take over and we have to “hunt squirrels for meat”. Then, I did something even more hysterical: I, an arts graduate (in history of art no less, probably the artsiest of all the arts), signed myself up for MIT’s six-week crash-course in Artificial Intelligence.

And that’s where the panic stopped. Because, it turns out, when it comes to complex human work - including strategy, creativity and almost everything else that brand consultants do best - generative AI is pretty rubbish, and probably always will be.

Take ChatGPT as an example. As a friend (himself an AI professional) put it to me, “Once you get over the amazement that a machine is writing copy, what you’re left with is some pretty average copy.”

“Sure,” you might retort, “it’s average right now, but give it six months won’t it be a lot better? It’s already faster than me, but I can compete with it on quality. Tomorrow, will I be competing with the AI equivalent of Bill Bernbach on speed?”.

Well, no, actually you won’t. The way ChatGPT works is (and I’m aware I’m radically simplifying things here) by predicting the next most likely word in a sentence and then writing it, then very occasionally predicting the second or third most likely word in a sentence and writing that instead. It’s essentially predictive text, but with a bigger dataset and just enough built-in randomness to make it look like it’s “thinking”.

That’s why - providing you’re any good - it can’t take your job. It can give you answers that are “likely”, or “the average”. It can’t give you answers that are “original and best”. Given as, for strategists and creatives, “originality” is pretty much in the job description, we shouldn’t be worried about AI taking our role. In fact, the only people that ought to be worried are hacks.

You might be reading this and thinking “I’m not a hack, but I’m still not reassured. AI may not be ready to take my job now, but I’ve no idea what’s coming next.” Fair enough. A new AI breakthrough seems to hit the news every week at the moment, and at the current rate of progress - and given all the hype - you’d be forgiven for thinking we could all be living on the set of Blade Runner by the end of August.

In reality, however, I suspect this feeling is the product of some false assumptions about how AI works, and some confusion around the term “AI” itself.

As Andrew Ng (one of the pioneers of machine learning) explains in this video, the phrase ‘Artificial Intelligence’ usually refers to two things: Artificial Narrow Intelligence (ANI) which is training machines to do very specific tasks, like winning a game of Go, and Artificial General Intelligence (AGI), which is creating a machine that can independently replicate any intellectual task a human can perform. There has been a tonne of progress recently in ANI, and practically diddly-squat in AGI. If you wanted a machine to replace most knowledge workers wholesale, you’d need AGI, and we’re not really any closer to that now than we were before artificial intelligence became the new black. Every successful AI system is a chronic mono-tasker, which can handle only one task and only if the rules of that task never change. Every successful brand consultant is quite the opposite.

So, providing your job involves more than one task, the robots aren’t coming for you yet. Good news. You can relax. Or can you?

I happen to think that generative AI does represent a threat, but not to our jobs; I think it’s a threat to the quality of our output.

Many people have attempted to sell generative AI to me as a means of “inspiration”. “Get Midjourney to generate a few visual suggestions,” they say, or “get ChatGPT to give a load of brand lines and you can just build on it”. The trouble is, I’m not convinced that great, original ideas are built on the foundation of pretty bloody average suggestions.

When Paul McCartney spoke about writing the song “Yesterday”, he described how it came to him in a dream. He said he woke up with the lyrics and the melody half-formed in his mind before working it all out at the piano. He didn’t say, “I went around the town centre with a clipboard asking people to describe heartbreak, worked out the average of everyone’s responses, and then just built on it.” Yet that’s the kind of “inspiration” ChatGPT has to offer.

Original thinking is part “magic” (if it wasn’t mysterious there wouldn’t be a job in it for us) and part process. And that process is one of writing, designing, attempting, struggling, despairing, rallying and ultimately triumphing. It’s in that very human creative tussle that we discover what we think, what we know, and what we want to say. We can’t ask a machine to do our writing for us because, until we’ve written it, we’ve no idea what we wanted to say.

Generative AI invites us to cut all of this out. It invites us to look at a list, pick one, then tweak it. This will be an irresistible temptation for already-stretched agencies on oh-so-tight budgets. “Think of the time saved! Think of the stress we’ll avoid!” - yet getting a machine to do this kind of work will be a false economy at spectacular scale.

By handing over the act of “generation” to a machine you avoid the act of ‘creating’, of which frustration and waste are natural components. And, if we avoid being creative for long enough, we might lose the skill altogether. Just like the pilot trained on autopilot or the cab driver who navigates solely by GPS, the next generation of brand consultants might never learn to sit, brow-knotted, at their desk for hours, sweating over the best visual or verbal articulation of a brand’s purpose, position or meaning. In this dystopian vision for our industry, they will have an important role to play in joining up all the many stages of the brand building process - from the brief to the execution - in the midst of all the chaos of a product launch or a business transformation. And their AI tools will be a spectacular ‘safety net’ when they’re asked a question they ought to know the answer to, and a great way to create some passable visual executions at short notice. But the big idea will be beyond them. Artificial Intelligence won’t have taken their job, but it will have made them suck at it.