Three things we need to stop pretending about Gen AI

I've been in enough rooms with marketers, filmmakers and brand leaders to notice something.

Most of the loud opinions about Gen AI (generative artificial intelligence) right now are wrong. Not a little wrong. Confidently wrong. LinkedIn carousel wrong. And the longer we let them sit there unchallenged, the more we slow down the work that actually matters.

1. “Human craft and taste” only matters to the people who have it.

Every other week someone publishes a manifesto about how AI will never replicate the soul of human storytelling. Fine. I agree with the sentiment more than I let on. However, here's the part nobody wants to say out loud: That argument only holds for the small group of people who actually have craft and taste in the first place.

The honest version is this: Most people, including most professionals being paid to make content, will not come up with better storytelling than a competent AI model on a first pass. Their structure is weaker. Their instinct for tension is duller. Their references are thinner. Put a generic brief in front of a generic writer and the same brief in front of a generic AI, and the AI wins more often than the industry will admit.

That doesn't mean craft is dead. It means craft was always rarer than we pretended. The “human soul” argument has become a comfort blanket for people who don't want to compete. The ones who actually have taste aren't threatened by AI; they're using it as a medium to push their storytelling further. Painters didn't die when photography arrived. They just had to be better painters.

If you're a craftsperson, AI is a new instrument in your hands. If you're not, AI is the thing that finally exposes you. Both are true at the same time, and the industry needs to stop pretending only the flattering one is.

2. Most brands are looking at AI purely as a cost saving, then wondering why their work doesn't stand out.

Walk into almost any brand conversation about Gen AI right now and the opening line is the same: How much can this save us? Almost never: What can we now make that we couldn't make before?

This is the quietest tragedy in marketing right now. We've been handed a technology that genuinely expands the canvas of marketing. New aesthetics, new pacing, worlds that weren't possible on a commercial budget, characters that can live across formats in ways traditional production never allowed. And most brands are using it to shave a line item off a quote.

Worse, the same brands that want to save money with AI will tell you in the next breath that they want to do something new. They want the case study. They want the award. They want the press release. Then when the work shows up, they retreat to the safest possible version—the one that looks like every other ad in their category, just made cheaper. New ambition, old instincts. It doesn't work.

Obviously not all brands are like this, but you get the point.

3. Almost nobody in the room knows the real cost of commercial grade AI work.

There's a kind of LinkedIn post I've come to dread. You've seen it: “I made a billion dollar budget film in one day for a hundred bucks.” 40k likes, comments section full of marketing directors quietly recalibrating their next production budget downward.

Let me put real numbers to this. Every step across the pipeline burns credits. Image gen, video gen, upscales, re-rolls. And the latest models cost a multiple of the previous generation. Seedance, for example, is around five times the cost of the last wave of video models. Some complex shots take up to 30 generations to get right.

A commercial grade 30-second piece is roughly 12 to 15 shots. To deliver at the quality clients actually sign off on, you're looking at 250 to 500 image gen iterations and 200 to 400 video gen iterations on average. That's before client revisions, upscaling, refinement, and everything else in between.

Credits are only half the cost. The other half is the team. A competent team knows what to use AI for and what to use other methods for. They know how to problem-solve a shot with a real toolkit: Model selection, workflow building, technique selection, character, world, sound, and prop reference stacking, prompt building, and iteration. All of it tuned together.

A good team also knows how to prepare for client revisions before they happen. One of the techniques is to lock as much context and scene as you can in the image gen phase, so that when revisions land you're not regenerating a whole shot from scratch. Without that kind of foresight, every “can we just change the colour of the jacket” becomes a brand new shot. Because AI rolls the dice every single time you regenerate.

The grown-up conversation we still haven't had

If I sound impatient, it's because I am. The Gen AI conversation in marketing has been stuck in two cartoon positions for over a year now: The doomers who think it kills creativity, and the hypers who think it replaces it. Neither group is doing the harder work of figuring out what this technology is actually for.

What it's for, in my view, is this: It lets people with real taste make things that weren't possible before, at a scale and speed that finally matches the imagination behind them. That's the prize. Not cheaper ads. Not faster decks. A wider creative canvas.

But you only get there if you stop romanticising craft you don't have, stop pretending cost cutting is a strategy, and stop believing the LinkedIn fantasy about price. The brands that get this right over the next couple of years won't be the ones who used AI to save money. They'll be the ones who used it to mean something.

The rest will keep making the same work, slightly cheaper, and wondering why nobody's talking about it.

This article was written by Donavan Ratnasingam, CEO and chief creative of Vision Machina.