You’re spot on about the difficulty of modifying every cell in an adult.
For research purposes, usually there wouldn’t be any need to modify Bill. It’s much easier to modify germ line cells, where less than 100% success isn’t a problem. Typically, the CRISPR mix will be injected into oocytes or very early stage embryos. Just implant the embryos into Bill’s mate, and pick out any red-haired Bill Jr’s from the offspring.
But to get back from fighting the OP, there are ways to target and modify adult cells, with varying efficiency and specificity:
Remove, modify, screen, and re-implant the cells. This approach has been successfully used to treat genetic disorders of blood cell precursors, and to introduce modified immune cells that will target a patient’s cancer. However, it probably isn’t a good way to modify every hair follicle on Bill’s body, but we might be able to give him a wicked head of red hair plugs…
Use genetic elements that express only in hair follicle cells. After all, there has to be a set of instructions that direct only the hair cells to secrete all that Keratin. If we identify the right genetic regulatory elements, we can create a CRISPR construct that only expresses critical elements of the system in hair follicle cells.
Use a viral vector to deliver the constructs to a specific cell type. Natural viruses very efficiently target specific cells: HIV only infects certain immune cells, influenza only infects certain cell types in the respiratory tract, etc. And we can also engineer viral coat proteins that could bind to any cell surface proteins that are only expressed in a desired cell type. I don’t know whether there are hair follicle cell specific viruses, but in principle it should be possible to engineer one.
Physically deliver your constructs to target cells. This is pretty efficient when you have a large, isolated compartment that you want to modify: just inject everything into the eye, or the central nervous system. There’s nothing stopping you from micro-injecting the CRISPR constructs into every hair follicle (but I sure ain’t doing it…)
Finally, individual approaches often aren’t completely specific or efficient. Biology is messy and contrary. We might have genetic elements that express in hair follicles and intestinal cells, but not skin cells. We might also have a viral vector that targets all cells in the dermis and nasal epithelium. Finally, we might have a gel that will deliver the vector through outer layers of the skin. By combining all three approaches, we might be able to efficiently and specifically target the hair follicle cells for modification.