An employee nowadays is required to be able to adapt quickly to every job given, and to be able to learn from existing experience in order to compete with other workers. This research aims to determine the role of learning agility in improving the collaborative behavior of millennial generation employees in the Padang City Government. The independent variable in this research is learning agility and the dependent variable is collaboration behavior. The measuring instruments used in this research are the Learning Agility scale and the collaboration behavior scale. The sampling technique in this research is purposive sampling, namely sampling using certain considerations according to the desired criteria to determine the number of samples to be studied. The sample for this research was 266 employees in the Padang City Government. Validity and reliability testing in this research used the Cronbach Alpha technique. The results of the validity coefficient on the learning agility scale with corrected item-total correlation values​​ range from 0.316 up to 0.646, while the reliability coefficient is 0.831. The results of the validity coefficient on the Collaboration Behavior scale with corrected item-total correlation values​​ range from 0.302 up to 0.749, while the reliability coefficient is 0.853. Based on data analysis, the calculated t value of 17.309 is greater than the t table, which is 1.660. So it can be seen that the significance value is 0.001< 0.05, which means the hypothesis is accepted. This illustrates that learning agility has a significant effect on collaborative behavior. So it can be explained that H1 states that learning agility has a positive effect on collaborative behavior. The effective contribution of the learning agility variable to collaboration behaviour variable is 53%.