Glossary · Making it with AI
Human in the loop
Human in the loop means a person reviews and approves an AI system's work at set points before it moves on, instead of letting it run from start to finish on its own.
Why it matters
AI can make confident mistakes, and an automatic process can repeat them quickly. A person at the right checkpoints catches errors, and someone is answerable for what gets published.
Where the checkpoints sit matters more than how many there are. The most useful ones come just before something is hard to undo: before a plan is turned into finished footage, and before a video goes out to the public.
A good checkpoint shows the reviewer exactly what needs a decision and what it is based on. A person asked to approve everything at once, with no context, tends to approve without really looking.
For documentaries that name real people or make factual claims, human review is also about responsibility. Someone should be able to say why each claim is in the video and why the video was fit to publish.
Before any motion is made for a short about an early airline, an editor reviews the storyboard and sends back one frame that shows the wrong kind of plane for the year. Once the video is finished, it is watched in full again before it is published.
How we use it
A person approves each video's storyboard before any motion is made, and approves the video again before it is published.
Last reviewed 15 October 2026.