We have the technology to build vastly superior replacements right now. It's mostly network effect requirements that make this extremely challenging/slow to implement.

An example of something we could do is cryptographically authenticated web-of-trust creditworthiness estimation, with techniques like proof of burn and selective trust anchoring used to establish terminal nodes in the unrolled trust DAG. This sort of thing would allow for pseudonymous, automated determination of trust without the extreme security and privacy risks posed by centralized identity stores like the credit bureaus.