A few years ago I started telling people that we are all product managers now. It felt true, and mostly it is. The work of the modern knowledge worker has drifted away from doing the thing and towards orchestrating the thing: setting direction, making the trade-offs, keeping a dozen moving parts heading towards something you can actually ship. AI has only poured petrol on that drift. But the longer I sat with "product manager" the more it nagged at me, like a word that captures the mechanics of the job while quietly missing its soul. A product manager makes sure the thing gets built. It says nothing about whether the thing is any good — whether it carries anyone's stamp, whether it was worth making at all.

The better word, I've come to think, is director.

Consider what a film director actually does. They hold the vision. They decide what the story is, how it's paced, what it looks and feels like, which performances stay and which end up on the cutting-room floor. They almost never operate the camera, write the score, or build the set. A film is the work of hundreds of people, and yet we still call it the director's film. That isn't an accident of credits. It's a recognition that one person's sensibility ran through every decision and unified it into a single thing. The crew executed. The director authored.

That distinction is the whole game now, because AI has just handed every one of us a crew.

Think about what's really changed. The expensive, scarce thing used to be production: the ability to write the competent proposal, build the working feature, draft the clean report. That capability took years to acquire and it's what separated the people who could from the people who couldn't. AI has made it close to free. The crew it gives you is fast, tireless, endlessly willing, and competent at almost anything you point it at. And the moment everyone has a crew, the bottleneck stops being production and becomes direction: having something worth making, and the taste to know when it's right.

So the question I keep coming back to is this: can the film director be a role model for the rest of us? Not as a vibe, not as "be more visionary," but as a specific set of attributes you can actually practise. I think it can, and I think there are four of them.

Bring a point of view the tools serve. A director knows what the film is about before a single frame is shot. An AI crew, for all its capability, has no point of view of its own; it produces the average of every point of view, which is the precise opposite of authorship. The person who types "write me a strategy" has no vision and gets the average back. The person who knows what they're trying to say uses the same tool to render it. So the practice is simple and surprisingly hard: show up with a thesis before you ask the machine anything.

Learn to say no to competent work. This is the most underrated thing a director does. They kill good footage — well-shot, technically perfect, expensive footage — because it isn't right for this film. When competent output is free and infinite, the scarce act is rejection. Looking at something fluent and plausible and perfectly fine and saying "no, not this," because you can feel it's wrong even when you can't fault it. That instinct becomes the job.

Treat taste as retained judgment, not active craft. A director can overrule the cinematographer about the lighting without ever touching the camera, because they once learned how light works and carried that knowledge up a level. This is the part that matters most. You don't survive AI by out-executing it; you can't, and you shouldn't try. You survive by being the person who can evaluate what the crew produced, and you can only evaluate what you once understood how to make.

Own the coherence of the whole. A film is ten thousand small decisions, and the director's real job is making sure they add up to one thing. AI will happily make each local choice well and leave you with a heap of locally-good, globally-incoherent parts — a deck where every slide is sharp and the argument doesn't track, a codebase where every function works and the architecture is built on sand. Holding the whole in your head, so the parts cohere, is irreducibly directorial.

Notice what every one of those four has in common. They aren't things the crew can supply. The crew gives you everything except the point of view, the judgment, the no, and the coherence. Which is exactly why the director is the right role model and the actor isn't. An actor's gift is bound to their body: it can't be delegated, abstracted, or run over different material. A director's gift is bound to their judgment, which means it can be imposed on any production at all, including an AI one. A Guy Ritchie film is recognisably a Guy Ritchie film no matter who's in it. That portable signature — the thing that survives being handed to any crew — is precisely the creative value that survives being handed to this crew.

Now for the part I can't write around, because pretending it isn't there would make everything above a comfortable lie.

All four of those attributes are retained production knowledge. The director can do them because they came up through the craft: they cut bad footage, sat through dailies, worked as an AD, made three mediocre films before they made a good one. You acquire taste by doing a great deal of production and slowly learning to tell good from bad against honest feedback. And AI is busy removing exactly the production work that the next generation would have learned on.

I see it already. A junior can now produce a genuinely good proposal with AI, but if they've never seen one land, never sold one, never watched one die in a room, the proposal is hollow in a way neither they nor their manager can easily see. The same is true of code. Architecture matters far more than syntax, and AI has eaten the syntax, but a junior who can generate working code without the architectural experience to see that the data model won't survive the second feature is building on quicksand. The cruelty is that none of it shows up at the time. It looks competent. It runs. The quicksand only reveals itself three months later, when the thing has to change and the person who'd have caught it at the whiteboard was never in the room.

This is task masking, but a stranger and more troubling version than the old one. The old kind was a worker pretending to be busy, a discipline problem, and an ancient one. This new kind isn't pretended at all. The worker genuinely believes the work is good, because every signal they can read says so. The masking is being performed by the output itself, on their behalf, without their knowledge. The harder they work, the more convincingly they're masked. We used to suffer from imposter syndrome, where you feel less capable than your work. This is the inversion: work that's more capable than the worker, with no felt signal that anything is wrong.

So if directorial taste is built through production, and AI is removing the production, where does the next generation's taste come from? It would be easy to end on that cliff. But I don't think it's actually a cliff, and the proof has been in front of us the whole time.

Watch a nineteen-year-old make short-form video. They have directorial judgment — real taste, earned, not borrowed, and they got it young. They didn't skip the reps; they did more of them than any film director in history. Hundreds of videos, thousands of hours watched, calibrated against the most brutal and honest feedback loop ever built. Completion rate doesn't flatter you. The view count doesn't care about your feelings. They earned their eye the same way a director does — make something, find out fast whether it landed, adjust — except the loop runs in days instead of decades, and costs nothing.

That's the hopeful part, and it's worth being precise about, because it isn't wishful thinking. The path to taste hasn't closed. It's changed shape. AI removed the slow, expensive, ambient version — the one where you absorbed judgment over fifteen years just by being on the set — and it rewards those willing to build the fast version deliberately. The thing that decides whether AI lifts you or hollows you out isn't your age or your talent. It's whether your feedback is fast and honest. Short-form video is both. Software architecture and sales proposals are slow, and in the AI era increasingly fake, because the artifact looks so competent that the absence of pushback reads as success. Slow feedback is a handicap. Fake feedback is a trap. The work, then, is to manufacture the TikTok loop inside the domains that don't have one: to expose your output to honest reality fast enough that you actually learn from it. I'm not certain that loop transfers cleanly to every kind of work, but the principle holds. Make the prediction before the decision, then go back and check whether you were right. Show the rough version to the one colleague who'll tell you the truth, before you let AI polish it into something too smooth to criticise.

And here's the cheering part. The old path to directorial judgment was gatekept. You needed access to the set, the apprenticeship, someone willing to let you near the equipment, and a great many years. AI hands everyone the crew on the first day. The barrier was never vision: the urge to present something, to tell a story, is about as universal as anything human gets. The barrier was access to production, and that barrier has just collapsed. Everyone has the crew now. The only question left is whether you'll do the work of becoming the director — and that work, for the first time, is open to anyone willing to bring a point of view and build themselves an honest mirror.