It isn't. Of course, both are systems of some sort, but there are several differences:
1. You can simulate a systems engineering problem and if it doesn't give you an acceptable solution in the first go, change some parameters and re-run again. In human systems, you (usually) do not get such luxury.
2. Components in an engineered system behave predictably. Humans are free agents who often act in irrational, surprising ways. You cannot simulate and analyze that. This is why there are black swans.
We cannot bundle these two systems as one and the same thing.
2. Humans in general also behave predictably, though not rationally. Human engineering is surprisingly quite similar to debugging native crashes, but not, say civic engineering or software design. It is also similar to art, with its rule governed creativity.
3. Most engineers promoted into leaders were not great engineers, and especially not great systems engineers. Peter principle is hard at work here. Not even good managers.
4. Many of the problems posed to managers are completely ill-posed (heck, not even amenable to satisficing) and likely not possible to solve properly.
5. There is a difference between a leader and a manager. One does not necessarily excel at the other's job either.
6. It's a terrible mistake to not apply engineering methods to society. Author has it completely backwards. The problem is almost always not the method, but optimization algorithm and the metric used. Author presupposes changes that are immeasurable. Well, how does he know about them then? Guesses?
7. The standard solution to an open optimization problem is to use powerful methods for analysing infinite impulse response systems and nonlinear systems. And multivariate analysis. Since these are hard, they are often glossed over and preference is given to simpler ones.
8. Author provides no good concrete examples. Only suppositions.
9. Uses correlation with IQ ignoring the fact that social and mental problems are more pronounced at the extremes and it is a well known fact, this reversing causation. (Consider how sane you have to be to explain the obvious thing time and again.)
10. The unsaid rule about leadership is that you're not a good leader when you don't get elected for the next term or people you lead start to resent you. Even if you can convince them. The thing about convincing people, it never works in the long term because it does not solve any cause.
11. Many people touted as great leaders are actually pretty shielded from consequences and also employ or employed armies of lower rank leaders and managers.
Incentives.
Right now the road network has no information on where you are going. Because of this it is near impossible to route or plan traffic beforehand other than using historical estimates. It is everyone for themselves. In a completely driverless world people could submit their driving route the day before a major event like Thanksgiving so the network has more information on what traffic will look like then. The people that plan ahead could then be given a 'queue credit' or 'queue boost' that routes their car in to the faster moving lanes, where those that did not get stuck in the slower lanes for longer.
Systems level engineering breaks a problem-solving heuristic that people rely on to make analysis simple: that changing a variable does not feedback into the value of that variable, which allows you to break a big problem into a bunch of trivial small problems that can be solved independently. For systems level engineering problems, you have to consider the interaction of all components simultaneously, which is a very different type of analysis (you are designing for locally stable and hopefully approximately optimal systems level equilibria). It also has characteristics that go against the way humans want to design things: the general design pattern is that you can't design for "good" outcomes per se, you can only design systems that mitigate "bad" outcomes, with optimal "good" outcomes being the natural result of efficient pathology minimization. Most people tend to design for the outcome they want, rather than minimizing the outcomes they don't want, but only the latter works.
Bringing this back to human systems, in a naive analysis you can take any variable, label it "good", and dial it up to eleven with some direct trivial changes. I almost always see it approached this way at a policy level. In a systems level analysis, it never works that way and you have less control of the tradeoffs than you might like as a matter of policy. We suck at human systems because, thus far, we haven't come to terms with the reality that a systems level analysis doesn't let you dictate the absolute values of "good" and "bad" no matter how you define them and controlling the relative values is far from trivial in a systems level analysis. We have tools for approaching this, as some types of engineering require it, but we don't apply it because it doesn't lend itself to the kind of naive analysis people understand and the tools don't allow the kinds of outcomes in human systems we wish we could engineer.