The List That Would Suit Anyone
Imagine opening an AI tool and typing: give me ten ideas for a travel channel.
The answer came in two seconds. Ten tidy, well-formatted ideas, each with a catchy title and a line of explanation.
You read them and felt something odd.
The ideas aren't bad. They're all reasonable, all logical. But they'd suit any travel channel in the world. They'd suit you, they'd suit your neighbour, they'd suit someone filming in a country you've never heard of.
Which means they don't suit you.
On the same day, you gave it two hours of an interview you'd recorded and asked for a transcript. The whole text came back in three minutes, accurate and searchable.
Same tool, two completely opposite outcomes.
The Question That Settles Every Tool
The conversation about AI runs between two camps: people who say it makes everything for you, and people who say it kills originality.
Both ignore the only distinction that matters:
Is this task repetitive work, or personal judgement?
Repetitive work has a correct answer everyone agrees on. Turning speech into text, organising files, translating a sentence. Delegate it without hesitating.
Personal judgement has no correct answer. It has your answer. Which story deserves making, which sentence sounds like you, where you stop sharing.
And every mistake in using AI is delegating personal judgement as though it were repetitive work.
Where It Genuinely Helps
Transcription and Translation
The clearest win with no competition. An hour of speech becomes text in minutes, and you can search it by word.
This changes how you work, not just how fast. You become able to read a two-hour interview in ten minutes and locate exactly the line you want.
Translation is the same, provided someone who knows the language reviews it if it's going in front of people.
Initial Research
It saves you the first hour, not the research.
It gives you a map of the subject: the terminology, the main names, the usual questions. Then you start from real sources.
And don't carry a number, a date, or a name from it into a video without verifying. That isn't a cautious suggestion, it's a condition.
Organising What You Wrote
From scattered thoughts into a clear order. You write your material the way it came out of your head, then ask for it to be arranged into sections.
The distinction here matters: the material is yours, the ordering is help. That's entirely different from asking for an idea that isn't yours.
Titles, Descriptions, and Tags
Twenty title options in a second, to pick from or take inspiration from.
There's no loss here, because you're the one choosing in the end.
And Where It Ruins the Work
In the Idea Itself
Which is the story we started with.
AI is average by nature. It gives you what's common, which is precisely what a hundred people did before you.
A good idea comes from something you saw or something that bothered you. And that's what it doesn't have, and never will.
In the Voice You Write In
A script written end to end by AI has a tone people now recognise.
Balanced sentences, polite transitions, enthusiasm with no cause. You read it and feel it's correct and cold at the same time.
And your audience notices. Not because they can identify tools, but because they know how you talk and can tell this isn't you.
In Judging What's Worth Making
"Does this story deserve a film?" is a question that can't be delegated.
The answer depends on your audience, on what you've published before, and on what you want your channel to be in two years. None of which anyone but you knows.
In Faces and Voices That Never Existed
Enter here with real caution.
A scene that looks real and never happened puts you in a position where you get asked a dangerous question: what else in this video is real?
And in documentary work specifically, that demolishes the foundation of the form itself.
A Quick Table
| Task | Delegate it? |
|---|---|
| Transcribing speech to text | Yes |
| Translation with review | Yes |
| Organising thoughts you wrote | Yes |
| Suggesting titles to choose from | Yes |
| Initial research with verification | Carefully |
| Phrasing sentences inside your text | Carefully |
| Writing the whole script | No |
| Choosing the video's subject | No |
| Deciding what gets published | No |
The Practical Rule
Use it on material you already have, not to generate material out of nothing.
Transcribe, organise, translate, summarise, and search what you've got. Starting with a blank page and asking it to fill it is the moment the content stops having an author.
And one simple test before publishing: read the sentence out loud. If you wouldn't say it in those words to a friend, it isn't your sentence.
In Short
Delegate what has one right answer. Keep what has your answer.
The tools will improve every month, but the line doesn't move: what saves your time is a gain, and what replaces your judgement is a loss that shows up late.