You load the xml representation of the processes and then you can execute them using the APIs. As in any BPMS framework. I think that it's all explained in the Drools Flow Documentation and in the jBPM5 doc too -> https://hudson.jboss.org/hudson/job/jBPM5/lastSuccessfulBuild/artifact/target/jbpm-5.0-SNAPSHOT-docs-build/jbpm-docs/html/index.html
Greetings