The first task you automate sets the template for everything that follows. Start by automating work that’s already efficient — the parts that are fast, low-friction, and working — and you bake in the assumption that speed is the bottleneck. Start by automating work that’s repetitive but broken, and you scale the break. The order matters because automation doesn’t just save time. It rewrites your production logic.
I’ve watched teams automate caption placement before they fixed the script structure that makes captions necessary in the first place. I’ve seen shops build batch-rendering pipelines for templates that shouldn’t exist at all. The automation works. The output multiplies. But the underlying story problem — the reason the work felt heavy to begin with — never gets addressed. You just produce more of it, faster.
Automation Locks In Whatever Comes Before It
When you automate a task, you’re not just removing manual effort. You’re deciding that the task, as currently structured, is worth repeating at scale.
In After Effects, I can write an expression that automatically syncs lower-third animations to audio markers. That’s useful if the lower thirds are already serving the story — if they’re highlighting the right moments, in the right way, at the right density. But if the original edit over-relies on lower thirds because the script doesn’t establish context on its own, the expression just makes it easier to produce cluttered videos. The automation is technically successful. The content is still overdesigned.
The same pattern shows up in AI-assisted workflows. A rough-cut tool that auto-assembles B-roll based on transcript keywords will speed up editing. But if your shooting strategy doesn’t capture the right B-roll to begin with, the tool just surfaces bad options faster. You’re still making the same creative decisions — you’re just making them under time pressure instead of during pre-production, where they belong.
Automation is infrastructure. It makes the next instance of a task cheaper and faster. But it doesn’t make the task better. If the task isn’t load-bearing, you’ve just built infrastructure for waste.
The Repetition Rule: Automate What You’ve Already Proven
The tasks worth automating first are the ones you’ve done manually enough times to know exactly why they matter, where they break, and what good looks like. Not the tasks that feel tedious. The tasks that are tedious and stable.
In our production department, we automated the export and delivery pipeline for client review files long before we automated any part of the edit itself. Not because export is more important than editing — it’s not — but because we’d done it hundreds of times and knew exactly what the output needed to be. File naming conventions, resolution specs, codec settings, folder structure, notification triggers. All of that was settled. Automating it removed friction without introducing risk.
Editing, by contrast, is still a judgment call every time. We use AI tools to generate rough cuts and surface candidate moments, but the decision about what stays in the final cut is still manual. That’s not because we’re precious about craft. It’s because we haven’t yet done enough edits in a given format to know what a reliable automated decision would even look like. The pattern isn’t stable enough to scale.
If you automate too early — before you’ve repeated the task enough to understand its variance — you end up automating guesswork. The system runs, but it doesn’t know what it’s optimizing for, because you don’t either.
Broken Tasks Scale Faster Than Good Ones
The worst automation projects are the ones that succeed technically but fail strategically. The system works. The output increases. The problem gets worse.
I’ve seen this most clearly in data-driven animation work. A client wants to turn a spreadsheet into a video — say, quarterly performance metrics visualized as animated bar charts. The data is clean, the template is built, the script pulls numbers and renders frames. You can produce a hundred variations in the time it used to take to produce one.
But if the original decision to visualize that data as a bar chart was wrong — if the story actually needed a line chart, or a single headline number, or no chart at all — then automating the bar chart just means you’re producing a hundred videos that don’t land. The system is efficient. The content is still a mismatch for the audience.
This is where most automation advice breaks down. It focuses on volume and speed — how much you can produce, how fast — without asking whether the thing being produced should exist in that form at all. Efficiency is only useful if the task is already effective. Otherwise you’re just failing faster.
What to Automate First
Start with the tasks that meet three criteria: high repetition, low variability, and proven value. The work you’ve done so many times that you could write the checklist in your sleep. The work where the output is predictable and the quality threshold is clear. The work that, if it disappeared tomorrow, would immediately create a bottleneck.
For us, that’s been file management, delivery workflows, and template-based animation for formats we’ve proven over dozens of iterations. It’s not the creative core of the work. It’s the scaffolding that lets the creative core happen without administrative drag.
The tasks that feel like they need automation the most — the ones that are slow, frustrating, and inconsistent — are usually the ones that need redesign, not automation. If a task is painful because it’s poorly structured, automating it just moves the pain downstream. You’ll produce more output, but you’ll also produce more revisions, more client confusion, and more rework.
Automate the tasks you’ve already mastered, not the ones you’re still figuring out. The first task you automate teaches your system what success looks like. If you haven’t defined success manually, the system won’t define it for you.
| Kevin Baer is VP of Production at CGI Digital in Rochester, NY, with 26 years in video production and motion graphics. |