Everyone is debating whether AI will replace post-production jobs. Meanwhile, inside production pipelines, people are dealing with a much more practical question: how do you reliably ship AI-assisted work to clients?

Having worked in production operations at a generative AI filmmaking company serving major Hollywood studios, and having observed how AI is being adopted more broadly across the industry, I want to share what this actually looks like in practice not the version from the keynote. What AI is genuinely doing well right now.

The most mature application right now is lip sync and dubbing. In that role, AI wasn't replacing traditional ADR workflows so much as adding a new option that didn't exist before. In-vision ADR opened up localisation possibilities: dialogue changes, censorship versions, multi-language delivery that would have been too expensive or slow to attempt. It felt less like disruption and more like the toolkit getting bigger.

Rotoscoping and cleanup are other areas where the shift is real. A lot of the repetitive frame-by-frame masking work can now be automated. Junior artists aren't disappearing, experienced people are still essential for quality control (QC) and final decisions, but the volume of mechanical work landing on their desks is changing.

Colour matching is a genuine win. Getting a consistent baseline across footage from different cameras, locations and lighting conditions used to eat a lot of time. AI can get you there faster, which means the colourist spends more time on the decisions that actually matter.

What changes for producers…

Here's the thing nobody really talks about: for producers, AI makes the job more important, not easier.

More of the work now sits around checking AI outputs, building approval processes, coordinating between teams and making sure everything still meets broadcast standards. Producers are becoming translators between AI systems, creative teams and clients. The technology can generate outputs in minutes, but someone still has to manage approvals, version control, legal requirements, client expectations and delivery schedules.

In practice, generating output was rarely the challenge. Building reliable systems to evaluate it consistently technically, creatively, and against client expectations was much harder.

There's also something I've noticed in interviews over the past couple of years. Employers are increasingly expecting producers to operate AI tools directly, not just manage workflows around them. Sometimes that expectation doesn't appear in the job description until the interview. For people coming from traditional post backgrounds, that's a real shift, and the industry hasn't quite settled on where the line sits between operating a tool and managing a pipeline.

What AI still cannot do…

Every AI tool I've worked with needs an experienced human at key decision points. AI can identify clipping, silence, frame mismatches or inconsistent colour. It still can't decide whether a performance feels emotionally believable or whether a grade supports the story. That judgment still comes from people who've spent years doing it.

Why operational complexity is increasing…

When content becomes cheaper and faster to generate, teams don't just do the same work in less time they end up making more of it. Think about what happens to a single approved cut:

Director version. Broadcast version. Airline version. Localised version for each territory. Social versions cut for each platform. Censored version. Marketing version.

Every version needs QC. That's something people outside post rarely appreciate, and it's only getting more complex as AI lowers the barrier to generating new outputs.

AI doesn't remove friction so much as shift it from creation into coordination and decision-making.

That operational complexity is easy to underestimate unless you've actually worked inside it.

What skills become more valuable…

From what I've seen, AI is mostly being used to support artists rather than replace them. The repetitive technical work is getting faster, which frees up creative people for the decisions that actually need judgment.

The skills that matter in all of this haven't really changed. Knowing broadcast delivery standards inside out. Strong QC instincts. Experience managing complex multi-format, multi-version deliveries. The ability to design workflows that hold up as AI gets introduced into the mix.

Conclusion

AI is changing how content gets made. It isn't changing the need for experienced people who know how to deliver it. If anything, the technology is making judgment, workflow design and production coordination more valuable than ever.