That's a grander question than we can usually address here. More often, we are helping using reproducible examples, called a reprex to troubleshoot specific problems. The question might even be over-broad for the statnet.org mailing list.
I can only offer some general observations.
The first is that it appears that you have the same initial analysis to do 264 times--creating a social network object, called a graph (not to be confused with a chart; this one is a mathematical object). Even though you may eventually want to compare the companies on graph measures of centrality, connectedness, etc., this is where you start.
Employees are nodes in each graph. Interactions between employees are edges and the types of interactions are edge attributes. The graph objects may either be unidirectional, birectional or multi-directional.
For an illustration of how such graphs are constructed, my unpublished paper illustrates some of the tools involved and might get you started.