For workloads with a well-known process pattern (e.g. db instances), you can do the following:
extract command and full command records from UDR using printudr,
filter out instance strings,
assemble an Analyze commands file by using command name and instance for specifying TRAN-CLASS and PROCESSES,
add all generated TCLASS records to a workload,
reference this Analyze file using the READ command.
That's a rough description of a workflow which can be scripted. It worked fine in several implementation projects. Good candidates for workload instances are Oracle, UDB, WebSphere or web (http) processes.
Hope this helps,