An AI agent goes beyond answering a single prompt: it takes a goal, breaks it into steps, queries data sources or devices, acts, checks the result and continues. Modern agents typically use a large language model for the reasoning loop and a set of tools, such as a device API or a rule engine, for the actions.
In IoT the appeal is natural-language control over complex systems: “keep the workshop cool when someone is in it” becomes a plan that reads occupancy sensors and adjusts a thermostat. The risk is that a probabilistic planner is now connected to physical actuators, so guardrails, permission scopes and audit logs are essential. See AI agents for the smart home for a worked example, and the companion guide on factory agents.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.