I'm working on a stable coin project right now (unsurprisingly almost all of the time and costs are process and regulation).
And yeah, it relies on centralized trust for issuance, so it's hard to call it a "coin" in the same way bitcoin is, but I do believe it solves real problems.
There's a lot you can do with a collateralized-cryptographic-asset-on-a-distributed-ledger (or stablecoin) that you can't do with a bank account. For one thing, you can write your own wallet app that works however you want it. That would be like if all banks implemented the same standard, and you could use any app to control your accounts.
You can use it to do some interesting things, like write apps that require two-to-sign for certain types of transactions, or write a generic two-factor auth that works with any of these apps. You could write an arbitration service for settling disputes, or do actual (less than 10 cents) micro-transactions since the processing fees are fractions of a cent.
There's a lot of potential from letting programmers interact with the finance system more concretely, and that's the problem at least some of these stablecoins are trying to solve. Not Gemini, but over on the stellar network that's what we're working on.