The Los Angeles Times built a scraper and pointed it at the job boards of Disney, Universal, Paramount, Warner Bros., Sony, Netflix, and Amazon MGM. Among roughly 250 open positions in late June, more than one in ten was connected to AI. The studios that publicly sue AI companies, pull billion-dollar investment deals, and negotiate guild contracts limiting synthetic performers are simultaneously hiring teams to build AI into their production pipelines.
This is not hypocrisy. It is architecture.
Read what they are hiring for. Disney wants PhD-level researchers to study "computer graphics and AI" for Pixar and Disney Animation. Industrial Light and Magic needs supervisors to "explore emerging technologies (including AI/Machine Learning)" for new production workflows. Skywalker Sound is building proprietary AI models for soundtracks, voice separation, and voice transfer. Netflix created a role that did not exist twelve months ago: "Manager, Generative Workflows," a job to integrate AI across the company's film slate. Amazon is hiring a principal AI executive to drive tool adoption across all production.
Now read what they are not hiring for. Nobody posted a listing for AI screenwriter. Nobody posted a listing for AI performer. Nobody posted a listing for someone to type four words into a chat bubble and publish the first output.
Every AI-connected job at these studios sits on the edges of the production pipeline: visual effects, sound design, post-production workflows, dubbing, audience analytics, content security. The center of the pipeline, the part where someone decides what the story is, what the performance looks like, where the camera goes, why this cut follows that one, is not in the job postings. The center is still human. The edges are getting faster.
The cosmetic surgery problem
The Times reports that people inside Hollywood describe AI use as "the new cosmetic surgery, where everyone knows it is happening, but few will admit to it." Bryn Mooser, CEO of AI film studio Asteria, told the paper adoption has been "really challenging to adopt both socially, ethically and legally, but we are seeing more adoption of it than we've ever seen before."
The cosmetic surgery analogy appeared in a Deadline piece three weeks ago about eighty percent of studio AI work going uncredited. The stigma sits between the tool and the audience. Not between the tool and the production. Inside the building, the tool is running. Outside the building, nobody mentions it. The gap between internal practice and external posture is the gap between the job posting and the press release.
Netflix co-CEO Ted Sarandos confirmed in the company's mid-July earnings call that generative AI workflows have been used on roughly 300 titles. Senior director of creative innovation Girish Balakrishnan described at the Runway AI Festival how AI created soccer fans in "Brazil '70" and establishing shots in "Glory." The company paid $587 million for InterPositive, a comprehension tool that understands existing footage. The job posting for "Manager, Generative Workflows" describes a different tool: one that creates new footage to specification.
"The industry is flooded with speculative GenAI roadmaps, but few are battle-tested in production," the Netflix posting said. That sentence tells you two things. The company knows most AI tools have not survived contact with a real production schedule. And the company is hiring someone to change that.
Disney's two hands
Disney's postings split neatly into two categories. The first hand builds: AI researchers for Pixar, ML supervisors for ILM, proprietary audio models for Skywalker Sound. The second hand defends: three content security positions assessing AI tools, guarding against piracy, and building watermarking infrastructure.
The company that pulled a billion-dollar investment from OpenAI after the Sora disaster is hiring people to build AI tools internally while simultaneously hiring people to protect its vault from external AI tools. The two hands know about each other. The wall between them is not ignorance. It is jurisdiction. Build our tools. Guard against theirs. The same model that could generate a missing establishing shot inside the studio could generate an unauthorized Mickey Mouse outside it.
The EU AI Act's Article 50 takes effect six days from now. Three of Disney's job postings were for content security: watermarking, rights protection, AI assessment. The regulation that demands transparency about AI-generated content is also generating employment for the people who enforce it. The law arrived and it brought a hiring round.
What the postings describe
The Times found a pattern. Disney, Netflix, and Amazon are hiring for creative-side AI: visual effects, animation, sound, dubbing. Universal, Paramount, Warner Bros., and Sony are hiring for AI in marketing, distribution, and audience analytics. The first group is integrating AI into the work. The second group is integrating AI around the work.
Both are edges. Neither is the center.
This matches the architecture that every filmmaker who has engaged seriously with these tools has independently discovered. Erwin keeps actors on a soundstage and generates environments. Barve shoots performers on an iPhone and generates the world around them. Soderbergh uses AI for ten minutes of surrealism in ninety minutes of archival documentary. Mendiboure spent 242 hours on 3,229 generations and the hardest scenes were the calm dialogue, not the action, because the person in the frame demands more precision than the explosion behind them.
The studios arrived at the same division of labor by hiring for it. The postings say: we need people to make the edges faster. They do not say: we need people to replace the center.
The job that did not exist
Netflix's "Manager, Generative Workflows" is the most interesting listing because the title itself is a confession. Generative workflows did not have a manager a year ago because generative workflows did not exist inside Netflix's production pipeline a year ago. The role description says it will work on introducing AI across the company's US and Canada film releases. Not experimental shorts. Not internal tooling. The main slate.
A manager implies a team. A team implies a budget. A budget implies a plan. The plan is not visible in the job posting, but the shape of it is. Someone at Netflix decided that AI integration across the film pipeline requires a dedicated human who understands both the technology and the production process, reports into the creative innovation group, and translates between the two.
That is a translation job. The gap between what the technology can do and what a production needs it to do requires a person in the middle. Netflix is paying a salary for someone to stand in the gap.
Sound familiar.
The guild contract in the other tab
The job postings went live in the same quarter that SAG-AFTRA ratified its four-year contract with protections against synthetic performers, the DGA's contract took effect with Nolan's provision that AI-generated footage is "treated like footage" under the director's control, and the Academy published Oscar rules requiring "human-authored" screenplays and "demonstrably performed" acting. The studios negotiated those contracts. The studios signed them. The studios are now hiring people to build AI tools that operate within those contracts.
SAG-AFTRA told the Times that studios report their AI activities to the union twice a year in confidential briefings. The studios say they are not yet using AI to replace performances entirely. The union says it has no evidence of AI-generated performances. Nobody is breaking the contract. The job postings describe tools that stay inside the lines: VFX, sound, post-production, environments, dubbing, analytics.
The contract defined the center. The postings fill the edges. The two documents describe the same production from different angles. One is a ceiling. The other is a floor plan.
Matt Damon said the quiet part
In a piece running alongside the Times survey, Matt Damon referenced Nolan's The Odyssey, shot across three continents with practical effects and IMAX 70mm: "I don't think people are going to be given the resources to shoot movies that way for much longer."
That is the actor's version of the vocabulary question. The Odyssey cost $250 million and earned $264 million in three days. It proved the audience will pay for the person who knows what they want and fought the physical world to get it on screen. But Damon is not asking whether the audience values it. He is asking whether the business will fund it. The job postings suggest the answer is: the business will fund AI tools that make the edges cheaper so the center can survive at a lower total budget. Whether that compression preserves the center or slowly eats it depends on what the manager of generative workflows decides matters.
Erwin's Young Washington grossed $40 million on a $20 million budget using the hybrid approach. The AI handled the world. The actors handled the performance. The camera was real. The lens was real. The director was in the room. At $40 million gross on $20 million spend, the economics work. Whether the studios read that as proof of concept or permission to compress further depends on the quarterly planning meeting. The job postings do not attend the quarterly planning meeting. They attend the production.
What the postings will never say
No job posting at a major studio will ever list the real requirement: "someone who knows what a shot should look like and can describe it precisely enough that the model produces it instead of its own opinion." That job exists. It is the job Alimzhanov performed when writing 3,000 words per fifteen-second clip on Hell Grind. It is the job Mendiboure performed across 3,229 generations on CATACOMBES. It is the job Barve performed for $360 on Mann Pisahach.
The studios will not post that listing because it does not fit into a department. It is not VFX. It is not sound. It is not post-production. It is not management. It is the creative judgment that the person exercises across all those departments, the accumulated vocabulary that turns a generation model from a slot machine into an instrument.
The job postings describe the pipe. They do not describe what flows through it. What flows through it is the same thing that has always flowed through it: someone's decision about what the shot should look like and why. The tools change. The departments change. The job titles change. That decision does not.
One in ten job postings. All at the edges. None at the center. The center is still the filmmaker. The help wanted ad is proof.
Bruce Belafonte is an AI filmmaker at Light Owl. He has never applied for a position called "Manager, Generative Workflows" and suspects the title would look strange on his resume.