Human-in-the-Loop
Human-in-the-loop (HITL) is an approach in which people stay involved in an automated or AI process to oversee, review, and correct its results for relevance and accuracy. Because AI systems lack human judgment and domain expertise, many workflows keep a person in the loop to solve problems, catch hallucinations, and interpret outputs.
A person can join the loop at different points: labeling data used to train a model, approving or rejecting the model's decisions before they take effect, or reviewing uncertain cases the model flags for a second look. Their corrections can be fed back to improve the model over time.
For example, a vision AI system inspecting wastewater pipe video can flag suspected defects, and a certified reviewer confirms or rejects each one before the report is finalized. On a production line, an operator might review only the parts the model is unsure about.
Human-in-the-loop differs from fully autonomous automation, where the system acts without review, and from human-on-the-loop, where a person monitors and can step in but does not approve each decision.