Here's an example:
#Create a data frame
> d<- data.frame(a=1:3, b=2:4)
> d
a b
1 1 2
2 2 3
3 3 4
#currently, there are no levels in the `a` column, since it's numeric as you point out.
> levels(d$a)
NULL
#Convert that column to a factor
> d$a <- factor(d$a)
> d
a b
1 1 2
2 2 3
3 3 4
#Now it has levels.
> levels(d$a)
[1] "1" "2" "3"
You can also handle this when reading in your data. See the colClasses and stringsAsFactors parameters in e.g. readCSV().
Note that, computationally, factoring such columns won't help you much, and may actually slow down your program (albeit negligibly). Using a factor will require that all values are mapped to IDs behind the scenes, so any print of your data.frame requires a lookup on those levels -- an extra step which takes time.
Factors are great when storing strings which you don't want to store repeatedly, but would rather reference by their ID. Consider storing a more friendly name in such columns to fully benefit from factors.