AI is giving marketing leaders a headache

Our good enough mentality isn't helping

An image of one of your machines crosses your feed. Good lighting, right angle, and a guard that isn't on the real unit. Nobody on your team made it. Someone in a regional office needed a slide by Thursday and what came back was close enough to ship.

It was close enough for almost everyone who saw it. Not for the engineer on the buying committee, who caught it in about a second, said nothing, and trusted the next thing he saw from you a little less. No ticket, no angry email. Just a quiet discount applied to everything you publish afterwards.

I don't know whether that's happened at your company yet, and neither does anyone else. There's no survey of your channel, and a wrong image doesn't look wrong, so nothing you've seen so far settles it either way. What I do know is that the tools are already on the same laptops your dealers and your regional offices have.

What it looks like when someone who knows the machine sees it

Here's the version of it I watched happen this week. I was sitting with the marketing lead at a manufacturer we've worked with for years, showing them an image we'd generated of one of their own machines. They looked at it for about three seconds and said, I see a few mistakes already. The tiller arm was straight where it shouldn't be. The operator was too tall next to the safety gates, and the gates were turned the wrong way round.

Nothing about the picture looked wrong. The lighting was right, the environment was right, it read as a photograph. You'd have to know the machine to catch any of it, and they knew the machine.

Then they told me how their product images get made today. A background, then the truck, then finding a driver to drop in, all composited by hand. Their words: and then it takes time.

That's the whole thing in one conversation. The care is real, and that slow, exact process is what the care produces. So when something faster shows up, the question was never whether people would use it.

AI shouldn't be the bad guy

AI is in the room in most of my conversations now, whether or not it's on the agenda, and the same marketing leaders getting real speed out of it are quietly getting a headache they can't name yet.

Let me get one thing out of the way, because it eats every conversation on this subject and it's the least interesting part of it. I don't care that a computer made the image. If you've ever used Photoshop you've altered reality, and nobody treated the pen tool as a moral question. And anyone who thinks prompting is typing a sentence and taking what falls out hasn't tried to get one specific thing, at a standard, more than twice. It takes knowledge and it takes stubbornness. It's work.

So that's not my problem with any of this.

Two standards, same building

My problem is a contradiction, and it's usually sitting inside the same building, often inside the same week.

On one side, companies want the exact image of their product. There's a reason someone will fly a photographer to a plant, block a line for half a day, argue about the angle, and then spend another week on a reflection nobody will consciously notice. There's a reason the brand guidelines run to sixty pages and somebody owns them. That care isn't vanity. It's a belief that the product deserves to be shown as it actually is.

On the other side, in the same company, there's good enough.

Good enough is Thursday afternoon with a deck due Friday. Good enough is the regional office that needed a visual and got one. It's the distributor who rebuilt the sales sheet themselves years ago and never came back, the dealer with a cousin who does design, the whole "I can do it better" club that marketing departments have been quietly managing for thirty years. None of those people are doing anything they were told not to do. They're hitting a deadline with what they had.

What changed is how much ground they can cover on their own. Doing it yourself used to be slow, and slow is the only reason most people gave up and used the template. Now the person who used to give up gets something usable in four minutes and it looks good. The barrier was never a rule. It was effort, and the effort is gone.

So the two sides of that coin can't sit quietly next to each other anymore. The company that won't sign off a product shot with a wrong shadow is now publishing images under its own logo that nobody looked at, because they were close enough and the deadline was real.

And if you make a physical product, close enough isn't a compromise on taste. Every guard, every hose route, every panel is a decision somebody made and can defend, so an image that moves one of them isn't a looser interpretation of your machine. It's a different machine, and in the worst version it's one you don't sell, in a configuration you'd never build, sitting next to your logo.

Where the slop actually comes from

We're all complaining about AI slop. But slop isn't something the machine does to us. It's what you get when the standard drops and nobody's bothered enough to raise it back, which makes it an acceptance problem wearing a technology costume. A tool that produces whatever you point it at can only ever reflect the standard of the person pointing it, and good enough is a standard.

Working with human designers doesn't always result in immediate success either. There are multiple versions going back and forth, and it's not uncommon to find files named v22 or v28 when you look at the details. Why should AI be different? Too often the images are generated by the same teams that put the designer through 28 rounds, and when it's AI everything is good enough after v2. That's a double standard, and it doesn't make sense.

Why banning and reviewing both run out

Which is why I think it's time to part with that mentality. Not by banning the tools, because a ban lasts until the first deadline that won't move and then the image gets made anyway on somebody's personal account, outside your line of sight rather than outside your channel. And not by reviewing everything, which genuinely works right up until the volume outruns the people doing the reviewing, after which the queue becomes the thing people route around and the images that skip it are the ones made under the most time pressure. Another reviewer buys you time. It doesn't change the direction of travel.

Both of those read the problem as a discipline problem, and it isn't one. Almost nobody chooses good enough over exact. They choose good enough over nothing, because nothing is what they had to work from.

What your channel is generating from

Right now, for most manufacturers, that's the honest answer. A photo library covers the machines that existed on the day somebody hired a photographer, never the newest variant and never the configuration a specific buyer asked about. Brand guidelines have a lot to say about color and layout, and very little to say about whether a guard is in the right place. Neither one gives a dealer on a Friday afternoon anything correct to start from, so they start from the internet's average idea of a machine like yours, and the generator obliges.

Meanwhile, for a lot of what you sell, engineering already has a 3D model that had to be good enough to build the machine from. That's a much harder test than looking right, and it's why that model is the one version of your product that can't be wrong. In most companies it's sitting in a folder marketing has never been able to open and the dealer network has never heard of.

It won't be the whole catalog. Machines older than 3D CAD are a gap, configured variants are a gap, and your own list runs longer than mine. That's worth finding out rather than a reason not to look.

Because raising the standard back up isn't a matter of asking people to care more. It's making the exact image the fastest one to make. When the models and the brand definition sit where your team, your regional offices and your dealers can generate from them, the correct version becomes the path of least resistance, and that's the only kind of compliance that has ever survived a deadline.

Find out what yours are made from

Take the last ten product images your company published, anywhere, including the ones from regional offices and dealers. Ask where each one came from and what it was made from. If more than a couple of them trace back to nothing, you don't have an AI problem to solve. You have a source problem, and that one you can fix.

It's the part my team has been quietly working on for a while now. There's a version of this where correctness and consistency aren't something marketing has to defend every week, they're just how the setup behaves, and the people running it get to spend their attention on the next move instead. More on that soon.

If you do run those ten, I'd be curious to hear what you found. Especially if the answer surprised you.