Ah, but only some 10-20% of patients seeing a PCP for a new problem end up with a referral to a specialist. And the initial care rendered by the PCP will be much less aggressive than that of a specialist. Consultants must, somewhat necessarily, do definitive, "buck stops here" evaluations much of the time.
Come to your internist with a headache and he or she will check for signs of serious causes; if there are none, maybe some basic tests, a bit of reassurance and careful follow-up with some symptomatic measures. If it seems to be a migraine, no problem, specific treatment can be given. Recheck in 2 weeks and if all is well, that's that. If things are not improving, maybe then consider a referral (maybe 10% of the time).
OTOH, if you skip the PCP and refer yourself to a neurologist, there is a much greater likelhood you will get an MRI or CT scan, receive expensive blood work for vanishingly improbable (although possible) causes.
I may be exaggerating a bit to make my point, but it's like that. PCPs are trained and experienced in evaluating the undifferentiated patient where the various diseases are present with the same probability of their incidence in the general public. They do things stepwise and probabilistically. Specialists see a subpopulation already seen and initially evaluated by a PCP with little yield; because the chance of rarer diseases is higher and the liability for "missing something" is greater, they take a more exhaustive (and expensive) approach.
Not saying one is right and the other wrong - just that each does what is appropriate for the population they are trained to see. Primary care by a team of subspecialists is arguably the worst of all worlds: expensive, excessive, aggressive and filled with incidental findings, and likely to overlook preventive measures, psychophysiologic issues, etc.
My first stop for care is my internist for almost any problem.