What Live TV Taught Me About Automation

Ten seconds to air.

The set is lit, the graphics are loaded, the presenter is in position. In the control room, nobody talks anymore. Then the countdown hits zero, the red light comes on, and whatever happens next happens in front of everyone.

In live television, there is no second take. When something breaks, the audience sees it. You can't fix it in post, you can't quietly redo it tomorrow, you can't hide it. So you learn, very quickly, to build things that hold up.

For the past five years, I have worked in that environment at LN24, where I still run virtual production: designing the virtual sets and the workflows behind them. I do it alongside a small team of brilliant people who create a lot with very little, and who make it work live, every single day. Most of what I believe about automation, I learned with them.

This is the first post on this site, so it seems right to share how Sayeret came to be.

When days became hours

AI came into my work through the creative side first, the way it did for most people in production.

At first it felt like a cheat code. A 3D sequence that used to take a week of modelling and rendering became an afternoon. Subtitling a full episode, once an entire day of manual timing, became something that ran in the background while I worked on something else. Ideas that would have stayed on paper because there was never enough time suddenly became possible.

So I went further. I tested, built, broke things, and learned properly. Not to follow a trend, but because in a live environment, a tool that works only most of the time doesn't work at all.

Where the time actually goes

Then something surprised me.

Once the creative bottlenecks loosened, the remaining delays weren't creative at all. They were coordination. Files moving from one person to the next. The same information typed into three different systems. A whole process frozen because someone was waiting for a confirmation from someone else.

None of it required talent. None of it was anyone's favourite part of the job. And yet it was eating entire afternoons.

The question changed. It was no longer "how do we make this render faster?" It became "why does a five-minute task take an hour of someone's day?"

The AI is the easy part

Here is the lesson live TV taught me, and the one I keep coming back to.

Generating output is not the hard part anymore. A model can write an email, draft a schedule, summarise a meeting, produce an image. That capability is everywhere, and it gets better every month.

The hard part is the unwritten rules. Every team carries them in its head. What is actually available. How long things really take. Which step always has to happen before the other. What you never, ever do on a Friday afternoon. They exist nowhere in writing, and yet the whole operation depends on them.

Ignore those rules, and you get output that looks perfect and is completely unusable. Worse, you get it confidently, at scale, and someone has to clean up afterwards.

Respect them, and automation holds up. Even live.

At LN24, the people in the control room knew every one of those rules, by heart. We mapped them together, one by one, before automating anything. Today, a single director can run an entire live show, sets, graphics and camera moves, from one position. Not because the automation is clever, but because a whole team's knowledge is built into it.

That, to me, is what good automation looks like. It doesn't replace the people who know the work. It carries what they know.

Humans shouldn't be the connection between tools

Outside the studio, I started seeing the same thing everywhere.

A team spends hours on work that follows a pattern. Everyone knows the pattern, but nobody has ever written it down. The tools don't talk to each other, so a person becomes the connection between them: copying, retyping, forwarding, checking, waiting for a confirmation. Smart, skilled people, working as the glue between software.

And we've quietly gotten used to it. We call it admin. We call it the cost of doing business.

It doesn't have to be. For the first time, AI can take on that role. With access to the right context, it can handle the communication between tools, move information where it needs to go, and keep things moving, while people focus on the work that actually needs them.

Small wins, every week

That is why I started Sayeret: to take what live TV taught me, where every mistake is visible, and bring that way of working to other industries.

I don't believe in big-bang transformations that take a year and change nothing. I believe in small wins.

Understand how people actually work, not how the process is documented. Find the unwritten rules. Pick one workflow, automate it properly, make sure it holds up, then move to the next. Each one built on a clean, well-thought-out structure, so it can keep growing instead of having to be rebuilt in a year. And each one giving people real time back, week after week.

It is less spectacular than a grand AI strategy. It is also the only thing I've seen work.

The point of all of this

What motivates me is not efficiency charts. It is watching skilled people get their time back.

The journalist who can report instead of renaming files. The producer who can think about the show instead of chasing confirmations. The team that stops losing its afternoons to work that should never have been theirs.

The technology is ready. It has been for a while. What it needs is someone willing to sit down, understand the work, and build something that holds up when the red light comes on.