AI Task Management System 2025-01-08
Twitter
nishio When I showed my friend the conversation log of the AI task management system I was working on today, he was super amused and said, "It seems to have direct synergy with Cybozu's core business! You've been doing your job! I'm going to introduce it here as well. First, "What do you want me to do today?" and then the VM will stand up.
https://pbs.twimg.com/media/Ggxvio5aEAA8aeF?format=jpg&name=large#.png
nishio AI read the document and understood its task management job. https://pbs.twimg.com/media/GgxwCt6a8AAb7tW?format=jpg&name=large#.png
nishio As this document says, tasks are managed in backlog.yaml, so AI then goes to read it and figure out my tasks. It asks, "What should I do today?" and it will give me a recommended task. https://pbs.twimg.com/media/Ggxwg-Ma4AAPzmu?format=jpg&name=large#.png
nishio and they will break down the task and recommend specific steps to do this. They explain why you should do it now, what will be possible next if you do this, etc. https://pbs.twimg.com/media/Ggxw5lZboAEA1KS?format=jpg&name=large#.png
nishio and they will break down the task and recommend specific steps to do this. They explain why you should do it now, what will be possible next if you do this, etc. https://pbs.twimg.com/media/Ggxw5lZboAEA1KS?format=jpg&name=large#.png
nishio So I look at it, and when I tell him the condition I didn't mention at the beginning, "I can't sit down and develop today," he takes that into account and says, "Well, how about this?" I didn't have time to do the tasks I had already done, so he suggested something that I had already done. I didn't put "done" on the task I had already done, so he suggests something I had already done, and I say, "That's done. https://pbs.twimg.com/media/GgxxjFnboAAdf3O?format=jpg&name=large#.png
nishio While exchanging something like that, some tasks can be entrusted to another AI instance, and others can be put on hold until the conditions are met. Some tasks are delegated to other AI instances, while others are put on hold until the conditions are met. https://pbs.twimg.com/media/GgxygCpbwAA7zDz?format=jpg&name=large#.png
nishio I did the task proposed here because I thought "Oh, that's a task I can do in the remaining hour, so I'll do it" (AI had 90% completed and only the UI placement was not good, so I fixed it and did PullReq). (The AI was 90% complete and only the UI layout was not so good, so I fixed it and did PullReq) nishio Talk with ChatGPT's Advanced Voice on the way out and on the way back about the task you came up with, then throw the whole conversation log to the task management AI for management. and let it be the target. https://pbs.twimg.com/media/Ggx0KB4a8AAlGs_?format=jpg&name=large#.png
nishio The AI will look at that log and say, "You have three tasks, this is the one that should be your top priority right now. https://pbs.twimg.com/media/Ggx0j8SbkAAFsGB?format=jpg&name=large#.png
nishio What's interesting after this is that I said, "I'm going to leave the house in 12 minutes and walk to the train station to catch the train, tell me how to use my time efficiently," and then I said, "I have a task that can be performed in the time it takes to walk to the station. I said, "I'm going to leave my house in 12 minutes, walk to the station, and take the train. If a human boss instructs a human subordinate to "work during the time it takes to walk to the station," it could be considered power harassment. https://pbs.twimg.com/media/Ggx1TCCaQAAnx5Q?format=jpg&name=medium#.png
nishio "Think while you walk, use this framework on the train" or something like that. https://pbs.twimg.com/media/Ggx1t3iaYAEGhij?format=jpg&name=medium#.png
nishio So I took the train while doing my tasks, met a friend and said, "I made this today," and that's how we got to the beginning of the story. nishio When you are working, there are urgent tasks that need to be done with the highest priority, and there are tasks that are "I should do it, but it's not urgent", this latter task The people who don't record the "I should do it, but it's not urgent" tasks when they think of them will forget most of them while doing the urgent tasks in the first place. Then how about those who record nishio Many people who keep records tend to do so in the form of "one-line memos" or "TODO lists. Then, about two weeks later, when they look at the list and think, "I have time to spare today, what should I do? In this way, important but non-urgent tasks are being overlooked. nishio On the other hand, I can't take the time to describe it well. Because we are in a situation where we are competing for time with urgent tasks. And even if it were written well, the listability would now deteriorate and the cost of looking through the whole thing would go up, making it useless in the end.
However, if it is AI that reads and writes this list, this problem is solved.
nishio AI is good at creating solid context-supplemented descriptions based on text and reference material that humans have written in a messy way. It can also look at a lot of information and perform it more widely and faster than humans. So, from a human's crudely written or voice spoken task, AI can do the cleanup, and AI can read the cleanup. nishio My friend who was excited to read this is in the CEO business, and as you can see, this is a system that is beneficial when used by one individual, but even more so when used by a team. nishio The "points to be improved and ideas for improvement" that until now have disappeared, remained in the minds of individuals, or taken the form of fragmented descriptions with so little context that even the individuals themselves often lose the meaning, will now be stored in the organization in a form that can be shared with others in a form of a clean document. This is how "points to be improved and ideas for improvement" that until now have remained in the minds of individuals, or in the form of fragmented descriptions with too little context for even the individual to understand their meaning, are now stored in the organization in a form that can be shared with others. nishio If multiple people think something is an issue, it must be of high importance, we can use it to decide whether to take the time to explicitly address it, etc. If solutions are gathered from multiple perspectives on a single issue, it will be easier to arrive at a better solution. nishio On a simpler note, when a newcomer has a problem and wants to talk to a senior, rarely does the senior look busy. All competent seniors look busy. This can be reduced by simply saying "consult the AI, senior staff will look at the logs when they are available" to reduce the inefficiency of issues staying within the individual.
nishio The more efficient way of working brought about by the "AI in the team" situation is already so much more feasible as of January 2025. The situation makes one wonder how far progress will be made in the next 12 months. "It's like science fiction, but it's not science fiction, it's the present.
nishio A specific example of the story I was writing about here >nishio: I had AI create a "system to let AI manage tasks" because AI was asked to create it, but in short, there was a trade-off: a TODO list with one-line tasks loses too much context, and if you write detailed information to avoid losing it, the list becomes less comprehensive this time. But if we stop listing tasks and let AI do it, we can destroy the equilibrium.
nishio Part of the task description generated by AI yesterday https://pbs.twimg.com/media/Gg0drptaMAALpF1?format=jpg&name=medium#.png
nishio Today's recommended task I'm torn between an ambiguous task with an urgent flag and preparation for a presentation with a subtask due next Friday, which should be prioritized (a move that shows both distressed items and then recommends one over the other would be helpful...)
https://pbs.twimg.com/media/Gg0lJZqaMAQK3yf?format=jpg&name=medium#.png
---
This page is auto-translated from /nishio/AIタスク管理システム2025-01-08 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.