The paper demonstrates how AI and fuzzy logic techniques can be combined to implement real-time, regulatory and servo control of a simple process unit. The knowledge base for the regulatory controller was designed to provide smooth control near the set point, ‘typical’ feedback control in the intermediate region and as much control as necessary (available) to prevent the controlled variable from exceeding user-specified constraints. The servo controller implemented fast open-loop set-point changes by using two-level bang-bang control, and the knowledge base included rules to learn the switching parameters online. The controllers were implemented, in Prolog, on an IBM PC and interfaced to an actual process using an industrial PID controller as an interface. Experimental results showed that control performance met the objectives better than a conventional fixed-gain PID controller.