Heh, I can definitely imagine hooking up an LLM to our task board's API and having it suggest actions (say, marking something as ready for QA) based on my Git log, then execute those if approved. Probably save me a shitload of time. Hell, if I just let it run loose and do whatever it decided to do, daily, with no intervention, it'd likely do a better job than the median developer (and better than I do). Maybe suggest next-tasks for me. I'd worry about it damaging something, but I worry about that a ton every time I have to open up any of those confusing-UI shared-workspace sorts of things anyway.
That this would be a significant time savings mostly has to do with most task tracking systems being so very miserable and slow to work in for the majority of the people expected to use them, though. If we used something lighter and closer to where the work is happening (the code) it wouldn't really be that helpful.
Is marking a jira ticket ready for testing that complicated? That doesn't seem like something that I both spend a ton of time on while being non productive to want to outsource to AI
Various administrative junk probably takes up about 20% of my actual wall-time as far as the work itself, but contributes enormously more than that to my mental fatigue, task-switching time loss, and general stress level. It subjectively feels like 50% of my job. Interacting with Jira or Azure Devops or Asana or what have you is far from everything, but is a decent chunk of that.
This does tend to be a much bigger problem at bigcos than smaller shops, though.
I am at one of the biggest of bigcos. My entire jira interaction is move the ticket to in progress, comment on it maybe once a day when i make serious progress or get blocked, and then upload a screenshot when I've finished and move to done. I probably spend more time fixing merge conflicts because force-with-lease is blocked at my co than I do with jira
That this would be a significant time savings mostly has to do with most task tracking systems being so very miserable and slow to work in for the majority of the people expected to use them, though. If we used something lighter and closer to where the work is happening (the code) it wouldn't really be that helpful.