Turning Hours of Several Sprint/Task Setup Into Minutes of Work

Last Monday I came into work with five new Jira sprints to build from scratch. Each one had 50-plus tasks that needed due dates, proper assignees, and dependencies linked correctly. Before I built this workflow, that kind of morning would have wiped out most of my day. I finished all five in 45 minutes.

This is how the system works.

The Part Nobody Talks About

If you manage a team inside a project management tool, you know how much invisible time goes into sprint setup. You create the sprint, you create the tasks, and then you spend an hour clicking into each one to set a date, assign it to the right person, and manually link it to whatever it's waiting on. None of that requires any real thinking. It just requires your time, and it happens every single cycle.

I got tired of it taking up space in my week, so I built something to handle it for me.

Two Tools Doing Two Different Things

The workflow runs on Jira automation and Claude Code, but they're not interchangeable. Each one does what the other can't.

Jira automation handles the scaffolding. I create a Story with a label that tells Jira what type of release I'm working on, automation fires, and a full child-task list gets generated from a template. The naming is consistent, the structure is predictable, and it happens fast. What comes out of that, though, is basically a pile of tasks with no dates, no dependencies, and assignees that aren't adjusted for which product I'm working on. That's where most of the manual work used to live.

Claude Code handles everything that requires judgment. I've built out a knowledge base for it that includes my team structure, who owns which product, how far in advance each type of deliverable needs to land before a release, and the marketing calendar. With that context in place, it can look at a fresh sprint and do in two minutes what used to take me close to an hour.

What the Workflow Actually Looks Like

Before I build anything, I have Claude Code cross-reference the calendar against what already exists in Jira. It tells me which upcoming beats are missing a sprint, which ones have date drift, and whether any task titles are out of sync with what's actually on the calendar. That step alone keeps me from building on a bad foundation.

Once Jira automation has generated the task tree, I hand it off to Claude Code. It works backward from each release date and proposes due dates for every task based on the sequencing rules I've given it, avoiding weekends and accounting for which deliverables have to come before others. Before it changes anything in Jira, it shows me a full table of proposed dates organized by beat so I can review and approve. Then it goes and updates everything.

In the same session it links dependencies between tasks based on what logically blocks what, and adjusts assignees by product. If it's uncertain about something, it flags it and asks before making the change.

The first time I ran this on a real sprint, I had 30-plus tasks sitting there with no dates. Claude Code had a full proposed schedule in front of me in under a minute and updated everything in Jira in about two minutes after I approved it. I remember thinking that I had just saved myself at least 30 to 45 minutes on that sprint alone, and I hadn't done anything except review a table and say go ahead.

What I realized after that session was that I had stopped managing individual tasks. I was managing the logic instead. I defined the rules once, and the system applies them every time I need a new sprint built.

https://video.wixstatic.com/video/f2809f_a1f2989177f94cd88958e09922ec98c3/720p/mp4/file.mp4

What You Need to Build This

You don't need an engineering background to set this up. You need a Jira automation rule that spawns tasks from a template based on a label, which is a one-time configuration that Jira's native automation handles well. You need a knowledge base for Claude Code that clearly spells out your team structure, your sequencing rules, and anything else it needs to make consistent decisions without you walking it through each one. And you need a calendar in a format it can read, which in my case is just screenshots of the month view.

The part that actually takes thought is deciding what your rules are and writing them down precisely enough that an AI can apply them on its own. Once that's done, the setup work for each new sprint is mostly just reviewing what it proposes and approving it.

---

I'm a Senior Marketing Strategist with 16 years in the PC gaming industry. I write about the AI workflows I've built to get the busywork out of senior marketing roles.