ChatOps is an industry practice of conversation-driven development using instant messaging to automate common development or monitoring tasks. It makes team members communicate with each other efficiently to work, learn and create together. We expect ChatOps to have intelligent agents, which can perform predictive operations based on users' activities, in the near future. To enable such intelligent agents, we worked on building prediction models of users' next actions. We collected action histories and their objectives and then applied sequential pattern mining to predict next actions. The result shows that some rules that meet some conditions could apply in 99% rate. Even if data do not meet the conditions, excluding rules in unnecessary actions improved the quality of the rules. Then, we analyzed generality of rules. To relate different actions as the same action based on an interpretation, the quality of the rules was improved.