The industry has a phrase for human oversight of AI: human in the loop. It treats you as a checkpoint at the end. The agent does the work. You approve the output. The loop closes.
That works fine for small, occasional decisions. It collapses at scale. Three agents shipping in parallel, twenty clarifications open at once, fifty diffs to review, and the loop becomes the bottleneck. The human becomes a rubber stamp on decisions they never saw being made.
A factory doesn't work that way. A factory has a director.
Human in Control (HIC) means you direct the work, not just approve it. Lit Factory is built on that distinction.
What HIC looks like in Lit Factory
You choose the responsibility before delegating. Set the autonomy mode for agent work: who drives, when the agent asks for input, and where a person reviews completion. Keep tool access and workflow instructions aligned with that responsibility.
You watch the work in flight, not after. The Floor brings recorded work activity, status, and clarification requests together. You can inspect connected evidence and intervene while the work is moving, within the visibility supported by each integration.
Agents ask for help when nuance is at stake. Models can't capture what users haven't yet put into words. Agents can request clarification or review, with the question attached to the work. Your team can respond with the context in view and redirect the next step.
The result is a factory you direct, not a queue you approve. That's the difference between supervising slop and directing craft. It's also the difference between autonomy that stalls at the human bottleneck and autonomy that scales with the team.