You know who roasts your coffee. Who writes your copy?

Cropped shot of a young male hipster working in a trendy coffee shop

Some people will queue twenty minutes for small-batch coffee. They’ll interrogate the barista about the origin of the beans and the roast profile. And they’ll pay a sizeable premium because a real human, with opinions and a slightly ridiculous moustache, roasted it under a railway arch.

Sourdough. Craft gin. Vinyl. Farmers’ markets. An entire economy built on the promise that craftsmanship makes a difference.

Yet, while culture has gone nuts about “artisanal”, we’ve seen content production industrialised.

Many marketers are weirdly relaxed about handing over the thing that is supposed to make their company stand out to a probabilistic algorithm. A very sophisticated algorithm to be sure, but still an algorithm. The “instant” of the content world.

The great punctuation panic

Here’s the surreal bit. For twenty years I’ve had clients tell me that no one cares about commas—and that grammar pedantry was something to be tolerated, like a colleague who likes to constantly remind you that they do triathlons.

Those same clients are now clutching their pearls over the abundance of em-dashes, squinting at perfectly good sentences and asking, “did a human write this?”

They were wrong then. And they’re wrong now, but in an interesting way. Because when every feed is flooded with the same frictionless, AI-scented prose, tiny human tells suddenly matter. The slightly odd sentence. The deliberate pause. The joke that doesn’t quite land, but is a hallmark of beats per minute over gigahertz.

Oddly, the growing scarcity of imperfection has made it a sign of quality. And scarcity, as any economist (or coffee roaster) will tell you, is where the value is.

The arithmetic of the average

There’s a structural reason why AI-generated prose is insipid and it’s not a bug. An LLM is, at heart, a consensus engine. It’s trained on everyone and so sounds like no one.

That’s fine—useful, even—when you want to conform. A privacy policy for example. But it’s fatal when the entire job of the words is to be distinguishable.

Brand voice exists for exactly one reason: so that people can tell you apart. Give that job to a machine optimised to reduce variation from the mean and you haven’t scaled your voice, you’ve erased it—quickly and with excellent grammar—but erased it nonetheless.

Aged 8, I read my dad’s copy of The Book of Heroic Failures. I remember many of its anecdotes to this day, including the one about buses that repeatedly flew past queues of wannabe passengers. When they complained, the local authority explained that if the buses stopped to collect passengers they would not be able to keep to the official timetable. Much of the content produced today, even that used in large expensive campaigns, suffers from the same problem. In the quest to save a few bucks, project owners have forgotten about trying to pick up customers along the way.

The ditch on the other side

But there’s an equal and opposite silliness where people react to any whiff of silicon like a Victorian seeing a naked ankle.

LinkedIn is full of people justifying their use of AI, as if it’s a sin that must be confessed. By that logic we should also give up spellcheck and turn off those grammar squiggles… even abandon Word processing apps entirely! Every writer you admire used tools. Hemingway famously swore by his Royal typewriter. Douglas Adams, a favourite of mine, was an Apple fanboy and the first person in the UK to own a Mac.

The question was never whether a machine touched the words, but rather whether a human meant them; whether somebody with skin in the game decided this is how we want to sound, this is the reaction we want to evoke.

I’m not anti-AI. I’m anti-homogeneity.

AI didn’t invent beige copy

Now for the uncomfortable part, and the connection to my recent article on agentic AI and corporate culture.

AI did not invent bland, interchangeable corporate speak; that dishonour goes to the twelve SMEs in the review group, the VP who “just wants a quick look”, and the lawyers who must check every word. AI just does it faster and at a fraction of the cost.

Which is why the voice problem, like the autonomy problem, is really a culture issue. A distinctive voice requires someone empowered to sound different—and a leadership willing to sign off on content that somebody somewhere, yes even an analyst or potential customer, might disagree with. Companies trapped in cultures of negative success, where the safest sentence is the best sentence, were always going to end up sounding like everyone else.

This isn’t what we were promised

Sci-fi promised us jetpacks and flying cars. Instead, we got the rapid industrialisation of discourse. Jeez, we don’t even have a machine to clean the bathroom yet!

But if we must live in a world of AI-flooded feeds, we might as well play to win. If you want to build a real competitive edge, ask yourself if you are the barista striking up a conversation or the bus driver not even making eye contact with your audience as you efficiently speed by.

Encourage your writers—hardware or wetware—to express a point of view and embrace a little imperfection. And don’t get in a tizz over the odd em-dash.

Also posted on LinkedIn.

Discover more from Gloo - content that delivers

Subscribe now to keep reading and get access to the full archive.

Continue reading