Dear Francisco, nice question. Let's firstly take into account that nearly all around the world, in each country large companies are a huge minority, and by and large, most companies are SMEs. Therefore, the question is to be framed accordingly.

Generally speaking, one culd answer affirmatively. However, as we take a more coarsed approach, the answer is to be nuanced.

Not to mention -going back to the justification of the question, a (in my own view) very serious problem, namely: even though close to each other to some extent, systems ascience is, at the end of the day, quite different from complexity science. From this standpoint I would dare to say: within the context of system science one can raher safely anser: yes.

However, as we focus on complexity science, the answer becomes certainly much more dubitable.

I mostly agree with regards to the firm being a middle-number system, however, just from a mere intuitive point of view. To my understanding the definition provided by the above cited authors is hard to sustain (scientifically speaking), since no evidence is provided upon which cases are those where an ecosystem components (or a firm's) fails to be manageable by statistical means (largenumber system) or one-by-one. "Middle-number" is at the root of the applicability of hierarchy theory to complex systems, thus this entails a weak justification.

As an economist who has done institutional research for over 30 years, I found this question to be interesting. Not knowing the background of complex theories, I'm reading between the lines that the proposition is that firms are institutions to economize information costs. Correct me if I'm wrong, a "middle-number system" is a way to summarize a larger unmanageable set of statistical variables?  

Can information as digital bits fit into this framework? Look forward to further enlightment.

could be also useful consider the supra-systems and sub-system inside

If we don't understand a middle system, we certainly cannot understand a supra, or a sub system