Full automation is rarely the goal.
Most enterprise processes have decision points where a human needs to be involved - not because the technology can't proceed, but because the business requires it. Approvals, reviews, compliance checkpoints: these aren't obstacles to automation, they're part of how the work is supposed to run.Human-in-the-Loop makes those checkpoints a first-class part of any agent workflow.
Insert a human wherever the process calls for it
Builders can place a human checkpoint at any step in a workflow. The agent reaches that step, surfaces what it's done and what it needs, and waits. It doesn't proceed on its own. It holds until the right person acts. The checkpoint can take several forms: a simple approval or rejection, a structured form submission, a document review, or an inline decision embedded directly in the agent's output.
Who acts, and what happens next, is configurable
Each checkpoint defines who is authorized to act - a specific user, a business role, or an escalation path if the first person doesn't respond. The outcome routes back into the workflow, and everything is logged: who acted, what they decided, and when. Automation handles what it should. Humans handle what they must. And there's a complete record of every decision made along the way.
