I'm not exactly the one best suited for this specifically though, since this has a lot more to do with lore really. But mechanics-wise...
If I may though, I'd like to propose that we use a tiered multi-perspective approach towards this. Specifically, we identify individual provinces/areas as "centres of excellence" in specific fields (the most broadest approach would be military/economic/civil, but we can further subdivide that into something like tactical military tech, strategic military tech, fundamental economic tech, practical economic tech, ideological civil tech, humanist civil tech). All provinces that are immediately connected to that province have a lower-level that "slopes down" further away from the "center of excellence" - the degree of this "slope" fully customizable as lore dictates.
Let's use Rome as a hypothetical example, and say we use a six-tier six-field approach. So let's arbitrarily (for sake of example) identify Rome as a level 6 CoE in terms of strategic military tech, fundamental economic tech, practical economic tech, and ideological civil tech, then a level 5 CoE in tactical military tech and humanist civil tech. So the province of Rome could then hypothetically have a 1.5 or so investment in the level 6 techs, a 1.2 level on the level 5 techs, and so on (so a level 1 gets 0.0 tech)
Then we define the surrounding provinces. Let's forget the LI setting first and hypothetically pretend the province directly north of Rome are barbarians, and to the south are "fellow-civilized pacifist cultures". So the "barbarian province" could hypothetically have a level 6 in tactical military tech, level 4 in strategic miltary tech, and level 3 or lower else where, while the "fellow-civilized pacifist culture" could have a level 6 in the economic and civil techs while a lower level in military techs. This way, we can be very specific in our designations of how exactly each region/area specialized, and we can make the levels like a topographic map of sorts with certain areas as the "peaks" sloping downwards into the outlier regions.
The benefit of this systematic approach is that it can also be used to tie into the building redesign attempt I'll also be making. If we specifically identify which buildings are tied to which specific CoE levels, which can be more specific about which buildings are pre-built instead of a random motley of buildings wtih a few static and "strongholds"