The best structural engineers do not need complicated models.   Be skeptical of computer results and don't over complicate analysis models.   It is commonly said, that computer software can be a valuable and reliable tool only to those who otherwise do not need it.   This is true.   In your work, make this true.

In the October 2011 article in Structural Engineer, the world renowned structural engineer Bill Baker of SOM is quoted discussing the importance of simple hand calculations when designing the world's tallest building (Burj Khalifa):

Simplicity is not easy.  Always returning to the intrinsic idea of the building gives the design process lucidity and direction.   It also helps one make essential decisions when confronted with the unique situation that arise when creating a building of such great size [Baker: 2011, 12]

He required his team perform the preliminary designs using conjugate beam theorey, which are simple hand calculations. Our software writers that work on integrating BIM with FEM/Analysis Models completely do not understand this (for example Revit/Robot or XSTEEL/RAM).   They think it is actually useful to model the entire building, every floor slope or offset, every little filler beam around slab openings, etc.  They believe this is how we do our work!  I tried to help reduce this misunderstanding when writing BIM and the Structural Engineering Community.     We can model everything, it would be stupid, but we can.

Computers should be used as a tool to make a design decision, it shouldn't make the decision.   If it does, you are an idiot.  Stop what you are doing and talk to someone in the office that knows better and can be your mentor - you need one.  We can model base plates and foundations as shell elements, or we can do a 3-second hand calculation or quick spreadsheet.   You choose.  This is not about trying to take short cuts.   This is about knowing what the software can provide and what it can't.   If you already know the software cannot come close to mimicking reality, where do you draw the line?  We are further along than when Nervi wrote his book Structures in 1956 but his comments still resonate today:

Theoretical results are a vague and approximate image of physical reality... masonry and concrete are far from being isotropic and elastic.  Theory of structures considers our building as being out of time, in a kind of eternal stability and invariability.  But the simple and commonplace fact is that all structures decay...thus this second assumption is also unrealistic.  No soil is perfectly stable nor settles uniformly as time goes by.  All building materials flow viciously...cements and limes keep hardening for decades... theory of structures may be compared to the physiology of perfect organisms, which are permanently youthful and untouched by disease or functional deficiencies.  This kind of physiology is certainly indispensable in a school of medicine but such a school would graduate very poor doctors... the preeminence given to mathematics in our schools of engineering, persuade the young student that there is limitless potency in the theoretical calculations, and give blind faith in their results.  [Nervi: 1956, 15-16]

Is the concrete you are modeling Hookean (linear-elastic)?   Do plane sections really remain plane?  Is that foundation a true pin, a fixed point, or somewhere in between?   My point with these question is to convince those who rely to heavily on computers that these models, no matter how complex, still fail at mimicking reality.  I am not suggesting that we don't need to know about the state of the art in analytical modeling, I am just pressing the point that they will never achieve reality.   Sometime complex FEM modeling is unnecessary and does not contribute to a good design decision (for example, the modeling of a simple spread footing - don't do that).

In the book "Structural Engineering:  The Nature and Theory of Design" William Addis is concerned about how our educators may also be misleading our students when stating:

Students are now told mainly about the mathematics and engineering science relevant to engineering works, but not how to use this knowledge in design.  Nor are they taught the importance of other types of engineering knowledge in design, such as a qualitative understanding of structural behavior, precedent, empirical data and rules ('rules of thumb').  And lastly, they are poorly educated as to the limitations of theory, how and when its efficacy in design might be suspect, and when it might need to be supported, for instance, by tests or physical models.  [Addis, 1990]

David Krakauer of the Santa Fe Institute describes the m^cubed phenomenon or m^cubed mayham as confusions that arise in people minds between mathematics, mathematical models, and metaphors.  I would simplify this and call it m^squared and lump mathematics and mathematical models together as models.  In the past, I have described the FEM models we use in engineering practice differ from the real world and highlight how our models should never be assumed to mimic reality, but are simply tools for us to exercise our engineering (or moral) judgement (in my case, to design safe structures).  We should never mix up the model for what is actually happening in the real world.   This is analogous to Krakaur’s models and metaphors. 

Krakauer says in Harris’s book Making Sense

“you can talk about spring and levers and these are physical artifacts…and then there are mathematical models of spring and levers” and “there is this tendency to be epistemologically narcissistic.  We tend to take whatever current model we’re using and project that onto the natural world as the best-fitting template for how the natural world operates…for many reasons the model is imperfect, computers are not robust”.  

It is not that computers are not robust, they can be robust, it is that they don’t need to be – we humans need to be robust. We need to better understand computer models (and output) and not be subjects to its authority.   This is a actually a question of dominion – we need never forget that we rule over it. Also, to Krakaur’s point, we need recognize when we are being epistemologically narcissistic - it happens all the time and it is really lazy thinking.  This will make us better engineers.  We need to constantly question our models. Again models serve us, not the other way around.