Can AI help create user stories and estimates?
Note: This article is an excerpt from my book, P-AI-R Programming: How AI tools like GitHub Copilot and ChatGPT Can Radically Transform Your Development Workflow.
This is my eighth installment in my experiments with GitHub Copilot and ChatGPT. For today, I wanted to see how well ChatGPT would do with some agile planning tasks.
I’m going to make an admission here. I am not a fan of project management tasks. I don’t like creating stories, grooming stories, doing project estimates, etc. I want to build working code and do it as fast as possible.
However, I do recognize the value of project management tasks. That’s where AI comes into play. AI can help make the process a little simpler and more pleasant by automating certain aspects of the project management process. For example, AI-powered tools can assist with story grooming by creating comprehensive user acceptance criteria.
This gets these tasks off my plate as quickly as possible, so that I can get to the more interesting job of delivering software.
Note: GitHub Copilot is not really suitable for this type of usage, but I’ll get back to using it when it’s time to write some code.
Project Concept
I started by describing an application to ChatGPT — the demo application from many of my books and video courses (a classroom attendance app with roster, parents’ contact info, multi-platform support, and camera integration).
ChatGPT produced a marketing summary (“Attendance Manager”) suitable as a starting point for a market brochure and for building the rest of the project plan.
Epics and User Stories
ChatGPT generated epics for Student Information Management, Attendance Management, Multi-platform Support, and Photo Capturing — with user stories for home page, roster page, and student details page. This took only a few seconds and was more than I would have come up with on my own.
User Acceptance Criteria
ChatGPT generated UAC for the stories. It added a search feature I hadn’t requested — which I could ignore, adopt, or ask it to regenerate without. It didn’t always use strict “Given/When/Then” formatting, but the output was still useful.
Estimations
T-Shirt Sizing
ChatGPT assigned T-shirt sizes and attempted to justify its estimates — a useful starting point for management discussions.
Fibonacci Estimates
When asked to convert to Fibonacci values, ChatGPT adjusted relative complexity (e.g., search ranked simpler than roster view) showing contextual awareness.
Summary
ChatGPT can write marketing summaries, create agile epics and user stories, generate user acceptance criteria, and provide rough estimates. I’ve been using this strategy personally and enjoy the process more than ever. I find staring at a blank screen hard, but don’t mind rephrasing and refining ChatGPT’s output.
Get P-AI-R Programming on Amazon or DRM-free on Gumroad.