How does an LLM help when you can't get anyone to sign off on a change, or when QA is understaffed, or when there's too many stakeholders making conflicting and sometimes nonsensical requirements?

That kind of stuff can't be solved by technology.

In short, an LLM helps in creating enough shit disguised as work to keep the people spouting nonsensical requirements busy enough to not dwell on the important parts of the system.

LLMs are phenomenal busywork generators. And as a natural consequence, if you want to stay productive, you should be suspicious of anybody that uses one.

It's the opposite. LLMs are good after you make your way through the bureaucracy and end up with well defined specs and a ton of boilerplate code & tests to churn out. I'd say most devs are spending 90%+ of their time on the former.

Idk, ChatGPT has been great boost to my efficiency filling out all those "5 min" pre-review alignment standup meeting intake documents. Now it really DOES just take 5 mins!

Standup meeting intake documents?

I think you're doing standups wrong.

Do they need a cover sheet as well? Did you get the memo?

You fix what you can fix.

Wait until Atlassian starts adding AI generated tickets.