Autonomous systems have been deployed to perform difficult and risky tasks on humans' behalf. Software agent and multi-agent technologies have been used as a means to implement the autonomous capabilities of systems. Definitely, there is a challenge of building a model that enables agents and humans to interact and efficiently operate a system. This paper presents a Flexible Human-Agent Interaction (FHAI) model for a supervised autonomous system. The FHAI model provides a flexible mechanism that facilitates human-agent interactions. It enables agents to decide on the interactions with a human to intervene specific actions of the agents. The aim of the model is to improve autonomous systems' performance, reduce humans' workload and reduce interactions disturbances. The FHAI model is intended to be used in unmanned aerial systems.