The Difference Between AI as Assistant and AI as Author | Catch Lyt

AI and Innovation           5 min read             

The Difference Between AI as Assistant and AI as Author | Catch Lyt

​

If nobody on the team can explain why a shot is in the film, the tool has started writing it.

​

The industry talk often treats “using AI” as one activity. It is not. There is a practical line between a system that helps you make the work and a system that is making the work. Crossing that line without noticing is how films go generic while everyone still feels busy.

​

1. Assistant is the professional default

​

Transcription, grouping rushes, first assemblies, caption drafts, format versions, cleanup, option-generation for a title card. The human team still defines intent, chooses direction and takes the call at delivery. The tool makes that process less wasteful. Clients are usually fine with this once you describe it in plain language.

​

2. Author is a different job

Primary performances, generative B-roll presented as observed reality, a voice that stands in for a real person, a visual system nobody art-directed. The creative questions get heavier, and so do the ethical ones: likeness, training data, disclosure, whether the organisation still recognises itself. Education and institutional clients feel this quickly. Their films are often asking to be believed.

​

3. Authorship creeps

It rarely arrives as a decision in a meeting. It arrives as “just fill that gap,” then another gap, then a cut where the point of view has thinned and no one can defend why this corridor, this sky, this sentence. The work starts to look like everyone else’s work that used the same models that week.

​

4. A useful test

Can you say, in one sentence, why this element is in the film? If the honest answer is “because it was easy to generate,” the assistant has been promoted without a conversation. Put the line back: either shoot it, style it so the artifice is honest, or cut it.

​

5. Name the line on the job

Before post begins, agree what AI may touch and what it may not. That is not bureaucracy. It is how you keep responsibility visible — internally and with the client.

​

​

​

​

​

Learning summary

1. Assistant and author are different jobs.

2. Keep intent, selection and sign-off human.

3. Watch for gaps filled “just this once.”

4. Generic is what default authorship looks like.

5. If you cannot defend the choice, reopen who made it.

6. Draw the line before the timeline is full of easy options.

​

​

Useful reading

1. Adobe — generative AI: adobe.com/ai

2. Frame.io Insider: blog.frame.io

3. Instagram for Creators: creators.instagram.com