In my experience as a conference program chair, shepherding is usually applied to borderline papers, where the chairs think the paper has valuable material but flaws too serious for publication in its current form.
Assigning a shepherd for the paper means that your acceptance is conditional on revising to address those flaws. It's still good news, however, in that the conference wouldn't be assigning you a shepherd unless they think that you can overcome the current flaws and produce an acceptable paper. Moreover, it is normally the case that you are invited to communicate back and forth with a shepherd to make sure that your revisions are on target---the conference wants you to succeed in your revision.
What you need to do now is to communicate with the shepherd about your revision plan and make sure that what you want to do matches what they will consider acceptable. If you can find agreement on a set of acceptable revisions and successfully execute them, then your paper should become finally accepted.