We propose a robotic system for assistive humanrobot micromanipulation. Using high-speed visual feedback operating at 1 kHz, the robot tracks the workpiece held by a human operator and using this information, it aligns the workpiece to the target which is mounted to the robot's actuator. The system is able to track and compensate for errors with an accuracy of less than one micrometer. As an example application, we perform experiments on the peg-in-hole task, where the robot continuously aligns a 70 μm hole, attached to the robot, to a 50 μm peg, held by the operator, in order to facilitate insertion. The results show that the proposed system outperforms the operators performing the same task with magnified visual feedback in terms of both completion time and number of successful insertions. In addition, by using the proposed system the test subjects experienced that they were subject to significantly less mental strain compared to using magnified visual feedback.