Jonas Ackermann — An Assistant Director's Verdict on AI Script Breakdown

What Jonas Ackermann's Research Found

Jonas Ackermann began in film studies at the University of Zurich and quit — too theoretical. He went looking for sets instead, worked a year as a production assistant intern at two Swiss production companies, and used that experience to get into the National Film and Television School in the UK, where he graduated in 2026 from the MA in Assistant Directing and Floor Managing.

In May 2026 he was in Cannes for the La Cinef screening of Left Behind, Still Standing, directed by Vida Skerk — one of two NFTS titles selected that year and the only UK entries in the section. Jonas is credited as 1st Assistant Director on it.

He is not a power user, and this is not a story about a tool that changed his life. It is something more useful: an AD who made AI script breakdown the subject of his master's research, ran a controlled comparison, and then told us in precise detail what it did for him and what it did not.

Two nights in a steel factory, and why prep decides everything

The clearest argument for spending less time on paperwork is a schedule with no slack in it. Jonas's proudest credit is a two-night shoot inside a working steel plant — the United Cast Bar factory in Chesterfield — where the unit could only film during the plant's routine summer maintenance shutdown. Maintenance work ran through the day, so the crew had the building at night, with no pre-calls and nothing pre-rigged.

Since the location didn't allow any pre-calls and had works going on during the day, everything had to be fully setup during our shooting time on both days while we had to make sure everything is safe. — Jonas Ackermann

That constraint landed hardest on lighting and camera. A plant powered down for maintenance had to read on screen as alive and running, and cinematographer Kaysar Kasim's team had to build that impression from scratch inside the same window. Jonas still talks about walking into the fully lit factory: it looked like it was working, and all of it was light.

Vida Skerk, who directed the film, has since credited the AD team with making the shoot possible, describing Jonas as leading the set "with precision and kindness."

Nothing about those two nights is automatable. But they were only survivable because of what happened weeks earlier, on paper — which is exactly where his research points.

Why an assistant director wrote his master's about the breakdown

Jonas chose the breakdown because it is the one part of an AD's job that is pure extraction, and therefore the one part a machine has a real claim on. For his MA portfolio he set out to find and compare different AI-powered methods of breaking down a script, and Filmustage came out of that search.

The breakdown is a mechanical task where you are mainly extracting information for a written document. While it sounds simple, it is crucial for it to be accurate and takes a fair amount of time. — Jonas Ackermann

His hypothesis is a production argument, not a technology one. Compress the extraction without losing precision, and the hours go somewhere better:

I think if you were to speed-up this process while keeping the needed precision, it could give the 1st Assistant Director more time in pre-production to talk to the Heads of Departments and preparing an even greater schedule, which could also benefit the production in achieving more within the available resources. — Jonas Ackermann

That is the whole case for AI in pre-production, stated by someone with no commercial reason to state it: the saved time is not the point. What you buy with it is.

The old workflow: two quiet read-throughs and a night's distance

Before AI entered the process, Jonas built every breakdown by hand — and built deliberate slowness into it. Two read-throughs with no notes at all, then a manual breakdown, then a night away from the document before checking it again.

Without Filmustage, when I received the script, I would first give it 1-2 quiet read-throughs without making any notes. After that, I would manually breakdown the script in Movie Magic Scheduling. Normally, after finishing the breakdown I would put it aside for a night and then look over it again. Especially on complicated scripts, it is easy to miss something, especially when you work on the breakdown all day. — Jonas Ackermann

The night's distance is not a quirk. It is a correction mechanism for a known failure mode: after eight hours of tagging, an AD stops seeing the page.

The new workflow: same read-throughs, different starting point

The AI pass replaced exactly one layer — the first transcription of the obvious — and left the rest of his process intact. He still reads the script twice without notes. He still goes through it scene by scene at the end. What changed is what sits on screen when he starts.

For breaking down the script, I would upload it to the platform and let AI give me a solid structure to work of. Most explicitly mentioned elements in the script would already be allocated to the correct category, so I would save some time in getting a foundation. — Jonas Ackermann

Filmustage reads an uploaded PDF or FDX and tags production elements — cast, props, wardrobe, locations, vehicles — into their categories. From there Jonas finishes the breakdown himself and exports it onward to the scheduling software he builds the board in.

Where the AI stops and the assistant director starts

The boundary Jonas found is precise: AI handles what the script states, and the AD handles what the script implies. Continuity elements are the clearest example. A costume introduced in scene 1 and carried silently through the scenes that follow is not written down anywhere, so it cannot be extracted. It has to be known.

Afterwards, I would still be going through the script, scene by scene, finishing the breakdown by adding elements that didn't get recognised/continuity elements that aren't explicitly mentioned. — Jonas Ackermann

This is not a limitation anyone invented recently. Production teaching has always framed the AD's job as covering what the writer assumed rather than typed — one long-running workshop tells trainee ADs to "indicate everything that the writer intended not actually listed." ADs who distrust automation make the same point from the other direction: a widely read MovieMaker guide to breaking down a page argues you should "do your breakdown by hand", because no parser knows that two characters named on the page belong to a group who will all be in the lobby.

Jonas's workflow is not a rebuttal of that objection. It is an accommodation of it. The machine drafts what is written; he reads every scene anyway.

He is equally exact about the results, and this is worth quoting in full rather than paraphrasing into something more flattering:

I see great potential for Filmustage making to a point where it can achieve efficiency gains. In my experiment for my MA portfolio, I realised that getting this foundation to work off can provide time savings. I shared some suggestions for improvement of the interface with the Filmustage team since some functions didn't work as seamlessly, as I would have loved to, adding some extra time when it came to manual corrections. I think, once some interface changes have been done, Filmustage has the potential to reduce the time needed to break down a script. — Jonas Ackermann

That is a conditional endorsement, and it is more useful than an unconditional one. The foundation works. The corrections cost time that the foundation saves. His detailed interface notes went to the product team and are on the roadmap — which is the version of this story we would rather publish than a rounder one.

The Bottom Line

Jonas Ackermann's position on AI is the one most working ADs land on once they have actually tested it: skeptical about the creative side, interested in the administrative one. The industry's money is moving along the same line. Deloitte's TMT Predictions put the largest studios at under 3% of production budgets going into generative AI for content creation, while roughly 7% of operational spending shifts toward AI-enabled tools. Caution in the writers' room, appetite in the production office.

While I am not convinced that generative AI should have a creative input in a film, I think in administrative tasks (especially in the Production and Assistant Directing department) it could be of huge help. — Jonas Ackermann

He has just graduated and is stepping into runner roles on features and high-end TV while continuing to 1st AD smaller projects — which is to say he is at the exact point in a career where workflow habits get set. His advice to other ADs weighing it up is short:

Give it a try. New workflows can be scary and maybe it's not for you, but I think the platform is working on something great that can make your workflow more efficient in the future. — Jonas Ackermann