From the Little Stream Software blog
They consolidate each pair of RFM scores into a specific segment (like R5F3). Similar segments are then grouped to create the final Automatic Segments (R5F3 becomes the Potential Loyal segment).
The end result is that instead of 125 segments from RFM, there are about 30 Automatic Segments. Each customer is also assigned 3 segments based on their different behaviors.
The segment grouping also gives you a range of different size groups. The smaller groups are roughly 1/25th of your customers but the larger groups are up to 7/25ths.
These larger group sizes (but smaller than the customer grades) gives smaller and mid-sized stores more segmenting behavior to work with but without getting too fine-grained like RFM. Like I mentioned yesterday, you really want at least 200 customers in each segment. 400+ is much better.
With the Automatic Segments, even the smallest segment (the 1/25th one) should be large enough to start optimizing if you have 5,000 customers.
P.S. The Automatic Segments can work with even large stores since it's built on RFM. You'll just be working with larger groups of customers at a time.