For a quick primer on GPCRs, you can try to decipher this picture:
Or just check out Wikipedia.
Anyway, GPCRs, in response to hormones or drugs, activate a three-component G-protein, which in turn switches on a signaling pathway. The key is that all three-part G-proteins are basically the same - they need each of the three components, and those components hang out at the plasma membrane of the cell.
Or do they? This interesting paper shows that a completely different protein, a phosphatidylinositol-3-kinase, can act as the beta subunit of a heterotrimeric G-protein. This allows GPCRs to signal from inside the cell, away from plasma membrane - inside of endosomes. The paper shows that the phosphatidylinositol-3-kinase (named Vps15), is structurally very similar to the beta subunit - an impressive piece of convergent evolution. A protein, Vps15 with one function (a PI-3 kinase), is tweaked to perform a new one.