The first step is to eliminate daytimers. They may have been useful and viable long ago, but not now. If they can be replaced by the equivalent of protected translators, they can serve small to medium markets entirely and also provide useful service to groups of immigrants who tend to concentrate in smaller geographic areas. And, of course, they can be general market stations for a particular part of a big market, too.
Those Robin Mathis 50 kw day and 250 watt night stations and their ilk qualify as "daytimer" as well
Then, look at solutions for highly directional stations that just don't cover much of the market at night or have been grown out of by post-WW II urban sprawl. Those are, in many ways, daytimers too.
Mexico solved the problem and moved over 600 out of about 800 AMs to FM by changing the separation between stations; they realized that the current NARBA standard was based on the technology of the 40's and 50's and tighter assignments can be made. That standard is almost universal in much of the rest of the world. It would, though, require many stations to move to different frequencies to optimally use the band.
The question is whether there are investors willing to back such a plan. We are so far into being a smartphone based society that major changes in FM allocations may be, like AM stereo, too little and too late.