Hollywood’s Long View on Disruption
AI is forcing a lot of jobs to change faster than people expected. In some fields, the skills that got someone hired five years ago may not be the ones that matter by the end of the decade. That can sound a little grim if you stare at it from a spreadsheet. In practice, it mostly means the same thing Hollywood’s dealt with for years: the tools change, the job changes, and the people who stay useful learn the new version without acting like the sky has fallen.
That’s why Hollywood’s such a useful story for AI-era leadership. Film and TV have spent decades getting pushed around by new technology. Digital cameras changed how crews shot, edited and lit scenes. CGI altered what could be built on set. Virtual production put actors in front of giant LED walls instead of some of the physical sets that used to eat up time and money. Even work that used to depend on a very specialized hand can look different now. Electronic stabilizers have lowered the skill barrier for some Steadicam-style shots, and digital scouting tools let location teams check out streets, buildings and landscapes remotely before anyone burns a tank of gas.
That kind of change can feel dramatic from the outside, but on a working production it often feels annoyingly ordinary. A crew adapts, the process shifts and people figure out where their judgment still matters. No one gets a medal for missing the memo on the new software.
The people who last in a changing industry usually aren’t the ones who cling hardest to the old tools. They’re the ones who learn the new ones without making a funeral out of it.
My own route through that world made the lesson pretty hard to miss. I spent more than a decade as a director’s assistant, which meant living inside the messy, practical side of making projects happen: schedules, call sheets, location issues, script changes, the whole lovely circus. Then I spent roughly six years building AI tools meant to make preproduction less grueling. Simple as that. That combination was useful because it showed me both sides of the same problem. On set, the question’s how to keep production moving when plans change. In software, the question becomes how to cut down the repetitive work without stripping out the decisions that still need a person with taste and context.
The overlap’s clearer than people sometimes admit. Hollywood doesn’t treat reinvention like a philosophical exercise. It treats it like Tuesday. If a better camera rig appears, crews learn it. And if a location can be found faster through remote tools, teams use them. If a new workflow saves half a day in prep, nobody spends a week mourning the old one.
That’s the broader lesson for leaders now staring at AI and wondering what, exactly, gets left for people. The answer probably isn’t “nothing,” and it definitely isn’t “do business the old way and hope for the best.” It is closer to this: learn which parts of the work machines can take on, which parts still need judgment, and which skills deserve to be rebuilt instead of defended out of habit. Hollywood has been living with that pressure for years, and it has plenty to say before the next section gets into how hiring itself starts to change.
Hire for the Role in Front of You
Once the tools change this fast, the hiring question changes too. A résumé can tell you what someone did in the last chapter. It says less about how they’ll perform in the scene you’re casting right now.
Hollywood figured this out a long time ago. Casting directors don’t hand a role to the person with the longest IMDb page and call it wisdom. They run auditions. They watch how someone reads the lines, how they carry the part, how they fit with the rest of the cast. A famous name can help get people in the room, sure, but fit still decides plenty of jobs. If the part calls for a nervous, sharp, teenage Peter Parker, Tom Holland makes more sense than a veteran action star who has spent decades playing a very different kind of lead. Tom Cruise can sell a mission impossible with the best of them, but he would be a strange choice for Spider-Man. That’s not a knock on Cruise. It’s just a reminder that the right person depends on the role, not the résumé wallpaper behind it.
So Hiring works the same way more often than managers like to admit. If you’re building for an AI-heavy workflow, the person with twenty years of experience in pre-AI processes may not be your best bet for the next six months of work. In some cases, the stronger candidate might’ve only a short track record with the tools but a much sharper instinct for how to use them. They may write better prompts, move faster through drafts and spot where the machine starts to drift off course. The veteran, meanwhile, can be brilliant and still get stuck trying to force new tools into an old rhythm.
That’s the awkward part of hiring for AI. The old model rewarded accumulated experience, stable routines, and a long list of familiar software. Adapt without fuss and figure out where the machine’s useful versus where it’s just producing confident nonsense in a nice font, given the new model rewards people who can test quickly. A candidate who has spent less time in the field might still be the better hire if their current skills match the direction the business’s heading. That may feel a little rude to the résumé, but the résumé will survive.
A polished career history can describe the past. It can’t answer the only question that matters in the interview: can this person do the job we need today?
This is where leaders often get tripped up. They hire for comfort because comfort looks a lot like competence on paper. Someone who has done the same kind of work for years feels safer than someone who learned the new tools last quarter and is already using them well. Yet in an artificial intelligence workplace, the comfort choice can slow the team down. If the job now depends on rapid iteration, prompt judgment and a willingness to throw out a bad draft without getting precious about it, then a veteran of the older system might be less prepared than a newer person who came in after the ground moved.
There’s a useful way to think about this in uncertain conditions: hire for the work in front of you, not the work that used to exist. That idea gets at why winning in an uncertain world often depends on judging people by how they operate when the rules are still shifting. A team full of polished résumés can still stall if nobody on it knows how to work with the tools that are actually in use.
That doesn’t mean experience’s gone out the window. It hasn’t. Experience still matters when the work calls for judgment, taste, memory and the ability to catch a bad assumption before it becomes expensive. But experience only helps if it maps to the task at hand. A person who built a career on manual processes may be excellent at the old game and clumsy at the new one. Someone newer to the field may have fewer years behind them and a better feel for the systems you need now. That tradeoff’s real, and ignoring it’s how teams end up hiring for nostalgia.
Leaders who get this right tend to ask better questions in interviews. Not “How long have you been doing this?” but “What have you done with these tools lately?” Not “Where were you five years ago?” but “How do you work now, and how quickly do you learn?” Those questions sound plain for a reason. They cut through the polite fog around titles and tenure. They also make room for candidates who learned in public, adapted fast, and picked up AI tools before they became a line item on every job description.
So yes, the casting model’s useful. Hollywood has arguably always understood that a role is a role. The person who fits it best may not be the most decorated one in the room. In the same way, hiring for AI means judging people by what they can do now, whether their current skills match where the business’s headed, and whether they can keep pace when the tools change again next quarter. A smart leader won’t ask for the longest résumé. They’ll ask for the right performance.
Lead AI Like a Film Set
On a film set, a director rarely gets to sit back and hope the day sorts itself out. They change the blocking, if a scene feels flat. If a prop reads wrong, they swap it. They rewrite it, if a line lands awkwardly. If the shot needs more tension or more clarity, they ask for a different camera angle and keep the crew moving until the scene works. The job is active, and it has to be.
That habit matters in AI work, too. A manager who treats the model like a contractor on autopilot tends to get exactly what they asked for and nothing more. The task gets assigned, the output comes back, and the manager gives it a quick thumbs-up because it arrived on time. That approach feels efficient right up until the draft misses the brief, cites the wrong source, or sounds confident about something that never happened. By then, the cleanup takes longer than the original assignment would’ve taken.
AI work still needs a director, not a passenger.
Directors on set know that a good result usually comes from a chain of small interventions. They don’t wait for the actors, camera team and set dressers to guess what the scene should feel like. Notice what’s distracting, and make adjustments in real time, they watch the frame. Sometimes the fix is tiny. A chair’s moved six inches. A character turns slightly toward the window. A line’s shortened because the rhythm’s off. Nobody puts that on a motivational poster, but it’s how the work gets better.
AI asks for the same kind of attention. The first output often looks serviceable at a glance, which is part of the trap. It can sound polished while still missing the assignment. A summary might answer a different question than the one you asked. Makes sense. A research draft might quote a source that doesn’t support the claim. A product description might invent details with a straight face. That’s the hallucination problem in plain English. The model fills gaps with plausible nonsense, and unless somebody checks the result carefully, that nonsense can slip into a deck, a memo, or a customer-facing page.
This is where human judgment earns its keep. A leader cannot simply ask, “Did the tool return something?” The better question is, “Did it answer the brief I actually gave it?” Then comes the second pass: “Are the facts real? Are the sources trustworthy? Did the model drift away from the use case?” Those checks sound basic, but they are where a lot of AI projects wobble. Speed is useful only if the result can survive contact with reality.
Prompt writing follows the same logic. The first prompt’s often rough, and the output mirrors that roughness. A vague instruction leads to a vague draft, and an overpacked prompt can confuse the model just as badly. So the prompt gets revised. The output gets trimmed, re-ordered, or rewritten. Sometimes the useful part’s buried in the third or fourth version, not the first one. That isn’t failure. It’s the normal work of directing something that can generate text quickly but can’t reliably judge whether the text fits the moment.
Harvard Business Review’s written about how leaders create clarity in uncertain times, and that idea carries over neatly here. People do better when they know what good looks like, where the boundaries are, and who’s making the final call. The APA has also noted that many workers find meaning when their work feels clear and useful, which makes sense. No one enjoys producing five drafts of something and then discovering that the target moved three times without warning. A plain brief, a defined standard, and a real review sequence keep that chaos from swallowing the day.
That style of management’s less about control than about attention. A director doesn’t micromanage every second on set, but they do stay close enough to catch what matters. AI calls for the same balance. Let the system do the tedious first pass. Or condense material fast, let it sort options, draft variants. Then step back in and decide whether the result actually holds up. If it doesn’t, revise the prompt, rerun the draft and tighten the output again. A useful AI workflow usually looks less like one clean handoff and more like a series of small corrections.
Even teams that build practical tools for speed run into the same pattern. In that sense, the job isn’t all that mysterious. Whether the task is a scene on a soundstage or a draft from a model, someone has to decide what good looks like, test the result against that standard and ask for another take when the first one misses. That’s where leadership strategy turns into actual work.
Keep the Human Final Cut
As AI skills change across offices, studios and production teams, a funny thing happens to old hiring habits. The ability to perform one narrow task very well stops being the cleanest reason to bring someone in. If a model can draft, sort, summarize, or rough out a first pass in seconds, then the person who only knows how to do that one thing manually starts to look less rare. Speed matters. Cost matters too. In the way they used to be, but those two advantages are no longer attached to humans.
That leaves a different set of qualities doing more of the work. Judgment’s one of them. So is creativity, especially the kind that can spot when a technically correct answer still feels dead on arrival. Taste may sound squishy, but in practice it decides which version gets kept, which gets tossed and which gets another round of work. A leader with taste can look at ten decent outputs and notice that nine of them have the emotional temperature of a parking receipt. The tenth may not be perfect, but it’s a point of view.
When machines can do the busywork, the people who choose what survives the cut matter more.
Recent film debates make that pretty easy to see. Christopher Nolan’s The Odyssey drew praise for using practical builds for the Cyclops and the ships, a choice that felt deliberate rather than decorative. The discussion around that film leaned toward craftsmanship and restraint. Viewers could tell where physical work had been done, and that mattered to them. By contrast, Young Washington drew criticism for visible generative AI in its battle scenes. The problem wasn’t that technology appeared at all. Plenty of films use digital tools. The complaint was that the result looked too obvious, too eager to call attention to itself. People noticed, and once they did, the spell broke a little.
That reaction has shown up before the finish line, too. Killing Satoshi, later renamed Bitcoin, faced early backlash before the production was complete. Misaligned ran into the same sort of skepticism. In both cases, audiences and critics were reacting to what they could already see, not to some finished, polished version that might’ve softened the edges later. Quick aside. That’s useful for leaders to remember. Customers don’t wait patiently for the internal logic of your process. They judge what reaches them, sometimes fast, sometimes unfairly, often with very little interest in your excuse about timelines, tools, or pipeline issues.
The same pattern shows up in business. A company can automate a pile of production work and still lose the room if the final result feels hollow, off-brand, or strangely generic. AI can help write the first draft, sort the footage, generate options, and speed up repetitive work. It can also miss the thing a person would catch in one glance: a tone that feels wrong, a joke that lands flat, a scene that looks fake, or a product message that sounds like it was assembled by committee and a toaster.
Creative leadership matters more in that setting, not less. The people in charge need to decide what the machine should do, what it shouldn’t touch, and where the final call belongs. That may sound obvious, but a lot of teams get dazzled by throughput and forget the finish. Once that happens, the work starts to feel efficient in the worst possible way.
Companies should absolutely automate what AI does well. Let it chew through repetition. Let it move faster than a human can. But keep people in charge of the choices customers will notice and judge. That’s where trust lives, and that’s where leadership still earns its keep.
