Yes, but it can allow you to match "J L Smith" against "Mr John L. Smith" or even "Jim Smithe" and similar variations. Usually that alone won't be enough, but combined with other data and perhaps human oversight it can allow you to detect/eliminate duplicates. (You do have to be careful of John Smith Senior and John Smith Junior who live at the same address.)

I implemented this years ago on a 6 million name dataset and it (helped) illiminate over 20,000 misreads, data entry errors and fraudulent attempts. The indexes were implemented as bitfields -- 1 bit per full name -- which allowed any entry to be rapidly compared against all the names and reduced to a tiny subset of the 6 million for further manual investigation.