This is the best solution when you are building a whole system from scratch. If you have to build a feature or whatever it is then it already got implicit permission from the jira sponsor or the product owner or whatever stakeholders are in charge. And for a feature I don't have to lose days writing long documents. I just think how it should work, I discuss it with the senior members of the team and with the stakeholders if needed. From that point on I just speak with the users to understand exactly what they want showing them whatever I'm building to have an early feedback. I think that is much more productive spending several days in a tight feedback cycle with your users than spending the same amount of times drafting the architecture in a document.

How do you get from "I just think how it should work" to "discuss it with the senior members"? You've got notes, at the very least, that you've jotted down about the proposed changes. Guess what: that's the draft of a design document! If you've taken any reasonably thoughtful approach you'll have just a little more work to make it formal, invite formal comments and feedback, etc. Not weeks "drafting an architecture" document, just days (or possibly hours, if it's a small change). That's a small price in time to pay for such a big dividend down the road.

I don't think this is an argument for requiring plans, though: you can always encourage them, but in the end it should be up to the person doing the design. Otherwise you end up with the absurdity of having to write a detailed plan and wait a month for approval just to change the positioning of a button slightly.

If the implementation outlives the implementer it most certainly shouldn't be up to him whether he does the design or not.

What I'm saying is that depends on the nature of the code that's being written. Many small improvements can be made with a simple code change that has clear meaning on its own -- a long approval process would just make these changes less likely to happen.

The point of having an approval process of any length or complexity is precisely to make changes less likely to happen ad hoc. That's a Good Thing (tm)!

Sometimes "small improvements" have ramifications outside of the context of the change (potentially significant ones). The person implementing these improvements has to understand the scope of the change (including side effects on external systems, if any), and that effort should be documented somewhere or at the very least reviewed by peers for correctness, especially for truly mission-critical or sensitive systems.

It doesn't have to be a "fill out these forms in triplicate and don't forget the new cover on your TPS report mmmkay" kind of documented effort: it's not one extreme or another.