Our Scrum Master is overwhelmed. Can we use AutoGen to create a "Scrum Crew" where agents look at our Jira backlog, estimate points, and suggest a Sprint goal? How would this integrate with our current Agile and Scrum ceremonies?
3 answers
We’ve been experimenting with an "AI Scrum Assistant" using AutoGen for the last three sprints. We have three agents: a "Product Owner" agent that prioritizes based on business value, a "Developer" agent that looks at code complexity, and a "Scrum Master" agent that ensures no one is over-allocated. They have a "discussion" about the backlog. The Developer agent might say a task is a 5-pointer, while the Product Owner thinks it’s a 2. They "talk" it out until they reach a consensus. It doesn't replace the planning meeting, but it gives the human team a "Draft Plan" to start with, which has cut our planning time by 40%.
Does the "Developer" agent have access to the actual codebase to make those complexity estimates accurate?
It’s great for summarizing the daily standup notes into a weekly report for stakeholders. Saves so much manual work.
Absolutely, Angela. The natural language processing in this framework makes it perfect for those kinds of "soft" Agile and Scrum tasks.
Yes, Jim! We use a tool that lets the AutoGen agent pull specific snippets or file structures from GitHub. It’s not reading the whole repo (that would be too many tokens), but it can see the files related to the ticket. This makes the estimates far more grounded in reality than just a generic guess. It’s been surprisingly accurate for boilerplate tasks, though it still struggles with high-level architectural changes.