Meggan WitteProject Manager / IT Operations P2| Leggett & Platt, Inc.Peculiar, Missouri, United States
Hi everyone! I’m new to the PMI community and excited to learn from others who are experimenting with AI in their project or Agile practices.
I’ve been added to an internal beta group at my organization to explore how AI can support our day‑to‑day work. We’re a mixed group—project manager, developers, mechanical engineers, network admins—each testing AI from our own angle.
My focus is on building an AI agent that can help draft user stories and related tasks. I’m being intentional about keeping this aligned with Agile values, so the goal isn’t to replace collaboration, just to reduce some of the repetitive admin work.
The idea is simple: during backlog refinement, I’d input the objectives we identify, and the agent would generate a draft user story with a clear objective, high‑level acceptance criteria, and a proposed owner. It would also create related tasks with high‑level descriptions. The team would still refine, estimate, and adjust everything as usual.
Since we’re a small team working across multiple projects, I’m hoping this could free up more time for deeper discussion and help us build out our backlog more efficiently. It’s an experiment, so it may or may not end up being a time‑saver — but I’m curious to try.
I’d love to hear from others: are you using AI agents in your Agile or project management work? What’s been helpful, and what hasn’t?
Is your agent one-shotting the entire flow? I find it helpful to assess whether I need automation, GenAI, or Agentic AI. A lot can be done with just automation or just GenAI; combining both into an agent can be powerful, but is sometimes overkill.
In our case, we're not creating stories/tasks in our PM tool without the stories and use cases being reviewed, first. I've played with an agent for requirements, but a prompt fits our flow. AI doesn't understand EVERYTHING about us, so it misses some stories and creates others that don't apply. I'm not sure AI actually speeds up the process, however. When I've done this, instead of waiting for people to create stories, I'm waiting for review and refinement of what AI has generated and hoping it doesn't get rubber-stamped because the person was too busy to give it more than a cursory glance.
If we were automating AI generated tasks into our PM tool an agent would make more sense. Overall, we've identified a few solid use cases for GenAI and are still figuring out the best use cases for agentic AI.
...
1 reply by Meggan Witte
Mar 10, 2026 10:25 PM
Meggan Witte
...
Hi Aaron! I probably didn’t give enough context in my original post, but to answer your question — no, I don’t want the AI agent to one‑shot the flow, nor would I ever allow that. I’ve worked hard to get my team into Agile, and collaboration is the most important part of our process. My goal for the AI agent is actually very minimal. I’d only use it during grooming. We’d still do everything we do today: collaborate as a team, break down features, discuss objectives, risks, business value, testing needs, and so on. As we break a feature into workable user stories, I want to be able to plug in the objectives we identify during discussion. Once those objectives are clear, I’d like the AI agent to trigger and create the draft user story — assign the owner, input the objectives, and generate high‑level acceptance criteria. That gives my developer something to refine as the work evolves. I’d also like it to create the related task or tasks based on what we discussed. Then we move on to the next user story while we’re still in the flow. After all the stories are drafted, we’d still move into voting. That part can’t be automated; it has to be a team conversation as well. It’s also our chance to review complexity and make sure the stories make sense. I really see this as an experiment — a way to see if AI can support my Agile practices without replacing the values and principles that matter. I don’t expect it to catch everything, and I’ve already seen it miss things while testing. My team genuinely enjoys our grooming, planning, and meetings. They’re very communicative, and I don’t want to take away anything that works well. But if I can trigger something to create the user story drafts while we keep collaborating, it might help… and it might not. That’s what I’m trying to find out. Thanks for sharing - I appreciate the feedback!
Meggan, this is a thoughtful experiment and a very practical use case.
One thing I’ve been observing as teams experiment with AI in delivery work is that the biggest value often shows up not in replacing thinking, but in reducing the friction around structuring information.
Drafting user stories, organizing acceptance criteria, and outlining tasks are good examples of that. Those activities are necessary, but they’re often more about structuring ideas clearly than generating the ideas themselves. AI can be surprisingly helpful there.
Aaron’s point about prompts vs. agents is also an important one. In many cases a well-designed prompt inside the team’s existing workflow may provide most of the benefit without introducing additional system complexity.
Where I’ve seen AI help the most so far is when it acts more like a structured drafting assistant rather than an autonomous generator. For example:
• helping convert discussion notes from backlog refinement into a first draft of stories • suggesting acceptance criteria based on the objective being discussed • identifying missing elements (owner, dependencies, definition of done)
The team still reviews and reshapes everything, but it reduces the blank-page problem and frees up more time for the conversation that actually matters.
Many of the newer backlog and collaboration tools are beginning to experiment with this kind of structured drafting capability as well, which suggests this may become a common pattern in delivery workflows.
Your approach of keeping the team in the loop for refinement and estimation is key. If AI starts generating artifacts that people stop actively reviewing, it can create more problems than it solves.
I’d be interested to hear, as you test this, whether the biggest time savings come from draft generation, or from helping the team identify gaps and clarify stories earlier in refinement. Saving Changes...
Hi Meggan, this is a great idea and I think it would be beneficial in the long term, but training the AI agent to be able to create acceptable user stories and acceptance criteria can take some time in the beginning. I'd use an AI prompt tool to help generate good prompts and instructions for your AI agent, that way you don't need to use tokens inside of the AI agent to iterate.
For ourselves, we currently use AI agents to help us generate lessons learned and synthesize that data to generate insights for us. We've actually built a tool leveraging this in Jira called WorkshopIQ so if this is something that you'd be interested in, feel free to reach out or check us out here: https://marketplace.atlassian.com/apps/123...mp;tab=overview
...
1 reply by Meggan Witte
Mar 10, 2026 10:31 PM
Meggan Witte
...
Jacob — hello! Thank you for the feedback. You’re right; this will take time, and I want to make sure whatever I build is actually beneficial to my process and to my project team. I’m working with a co‑worker right now to figure out the best approach. I’m still a little hands‑on with what I want the agent to do, but I’m being cautious because I don’t want to disrupt a process I’ve worked hard to get my team to adopt—Agile. I’m also really interested in what you’re doing with lessons learned. I’ll check out your link, and if I have any questions, I’ll definitely reach out. Thanks again
Program Manager| HARPER SRLSanto Domingo / Distrito Nacional, Dominican Republic
We’ve experimented with AI in a similar way, mainly to draft user stories, summarize meetings, and prepare first versions of backlog items. It works well as a starting point that the team can refine together.
The key is treating it as a drafting assistant, not a decision-maker. It reduces repetitive work, but the real value still comes from team discussion during refinement.
...
1 reply by Meggan Witte
Mar 10, 2026 10:36 PM
Meggan Witte
...
Lissette, I couldn’t agree more. That’s exactly my goal — an assistant that supports us, but doesn’t make decisions for us. Team collaboration is a must, and making those decisions together, along with our stakeholders, is what really matters. Like you said, if it can reduce some of the repetitive work, that would be helpful. Thanks for sharing your feedback — I really appreciate it!
Project Manager| AWR Development (BD) Ltd. Cox's Bazer , Bangladesh
Welcome to the community! Your approach sounds thoughtful—using AI to reduce repetitive work while keeping collaboration and team refinement at the center. Generating draft user stories and tasks could definitely help speed up backlog preparation, especially for small teams working across multiple projects. I’m curious to see how it works in practice. It will be interesting to learn whether the AI drafts truly save time during refinement or if the team ends up rewriting most of them. Either way, experiments like this are a great way to understand where AI really adds value in Agile workflows. Saving Changes...
Meggan WitteProject Manager / IT Operations P2| Leggett & Platt, Inc.Peculiar, Missouri, United States
Mar 09, 2026 9:46 AM
Replying to Aaron Porter
...
Is your agent one-shotting the entire flow? I find it helpful to assess whether I need automation, GenAI, or Agentic AI. A lot can be done with just automation or just GenAI; combining both into an agent can be powerful, but is sometimes overkill.
In our case, we're not creating stories/tasks in our PM tool without the stories and use cases being reviewed, first. I've played with an agent for requirements, but a prompt fits our flow. AI doesn't understand EVERYTHING about us, so it misses some stories and creates others that don't apply. I'm not sure AI actually speeds up the process, however. When I've done this, instead of waiting for people to create stories, I'm waiting for review and refinement of what AI has generated and hoping it doesn't get rubber-stamped because the person was too busy to give it more than a cursory glance.
If we were automating AI generated tasks into our PM tool an agent would make more sense. Overall, we've identified a few solid use cases for GenAI and are still figuring out the best use cases for agentic AI.
Hi Aaron! I probably didn’t give enough context in my original post, but to answer your question — no, I don’t want the AI agent to one‑shot the flow, nor would I ever allow that. I’ve worked hard to get my team into Agile, and collaboration is the most important part of our process. My goal for the AI agent is actually very minimal. I’d only use it during grooming. We’d still do everything we do today: collaborate as a team, break down features, discuss objectives, risks, business value, testing needs, and so on. As we break a feature into workable user stories, I want to be able to plug in the objectives we identify during discussion. Once those objectives are clear, I’d like the AI agent to trigger and create the draft user story — assign the owner, input the objectives, and generate high‑level acceptance criteria. That gives my developer something to refine as the work evolves. I’d also like it to create the related task or tasks based on what we discussed. Then we move on to the next user story while we’re still in the flow. After all the stories are drafted, we’d still move into voting. That part can’t be automated; it has to be a team conversation as well. It’s also our chance to review complexity and make sure the stories make sense. I really see this as an experiment — a way to see if AI can support my Agile practices without replacing the values and principles that matter. I don’t expect it to catch everything, and I’ve already seen it miss things while testing. My team genuinely enjoys our grooming, planning, and meetings. They’re very communicative, and I don’t want to take away anything that works well. But if I can trigger something to create the user story drafts while we keep collaborating, it might help… and it might not. That’s what I’m trying to find out. Thanks for sharing - I appreciate the feedback!
...
1 reply by Aaron Porter
Mar 11, 2026 12:40 PM
Aaron Porter
...
That could be interesting. Hopefully you come back and share what you learn. My only suggestion would be to restate what I said about prompts and agents. Evaluate your use cases and determine when you need an agent and automation, and when you just need a prompt.
Saving Changes...
Meggan WitteProject Manager / IT Operations P2| Leggett & Platt, Inc.Peculiar, Missouri, United States
Mar 09, 2026 8:14 PM
Replying to Jacob Vu
...
Hi Meggan, this is a great idea and I think it would be beneficial in the long term, but training the AI agent to be able to create acceptable user stories and acceptance criteria can take some time in the beginning. I'd use an AI prompt tool to help generate good prompts and instructions for your AI agent, that way you don't need to use tokens inside of the AI agent to iterate.
For ourselves, we currently use AI agents to help us generate lessons learned and synthesize that data to generate insights for us. We've actually built a tool leveraging this in Jira called WorkshopIQ so if this is something that you'd be interested in, feel free to reach out or check us out here: https://marketplace.atlassian.com/apps/123...mp;tab=overview
Jacob — hello! Thank you for the feedback. You’re right; this will take time, and I want to make sure whatever I build is actually beneficial to my process and to my project team. I’m working with a co‑worker right now to figure out the best approach. I’m still a little hands‑on with what I want the agent to do, but I’m being cautious because I don’t want to disrupt a process I’ve worked hard to get my team to adopt—Agile. I’m also really interested in what you’re doing with lessons learned. I’ll check out your link, and if I have any questions, I’ll definitely reach out. Thanks again Saving Changes...
Meggan WitteProject Manager / IT Operations P2| Leggett & Platt, Inc.Peculiar, Missouri, United States
Mar 10, 2026 10:00 AM
Replying to Lissette Indhira Pimentel Sosa
...
We’ve experimented with AI in a similar way, mainly to draft user stories, summarize meetings, and prepare first versions of backlog items. It works well as a starting point that the team can refine together.
The key is treating it as a drafting assistant, not a decision-maker. It reduces repetitive work, but the real value still comes from team discussion during refinement.
Lissette, I couldn’t agree more. That’s exactly my goal — an assistant that supports us, but doesn’t make decisions for us. Team collaboration is a must, and making those decisions together, along with our stakeholders, is what really matters. Like you said, if it can reduce some of the repetitive work, that would be helpful. Thanks for sharing your feedback — I really appreciate it! Saving Changes...
Great initiative, Meggan Witte. Using AI agents to draft user stories during backlog refinement can reduce repetitive work while preserving Agile collaboration. The key value will be how the team refines and validates the AI-generated outputs to maintain clarity and business value. Looking forward to the insights from your experiment. Saving Changes...
Hi Aaron! I probably didn’t give enough context in my original post, but to answer your question — no, I don’t want the AI agent to one‑shot the flow, nor would I ever allow that. I’ve worked hard to get my team into Agile, and collaboration is the most important part of our process. My goal for the AI agent is actually very minimal. I’d only use it during grooming. We’d still do everything we do today: collaborate as a team, break down features, discuss objectives, risks, business value, testing needs, and so on. As we break a feature into workable user stories, I want to be able to plug in the objectives we identify during discussion. Once those objectives are clear, I’d like the AI agent to trigger and create the draft user story — assign the owner, input the objectives, and generate high‑level acceptance criteria. That gives my developer something to refine as the work evolves. I’d also like it to create the related task or tasks based on what we discussed. Then we move on to the next user story while we’re still in the flow. After all the stories are drafted, we’d still move into voting. That part can’t be automated; it has to be a team conversation as well. It’s also our chance to review complexity and make sure the stories make sense. I really see this as an experiment — a way to see if AI can support my Agile practices without replacing the values and principles that matter. I don’t expect it to catch everything, and I’ve already seen it miss things while testing. My team genuinely enjoys our grooming, planning, and meetings. They’re very communicative, and I don’t want to take away anything that works well. But if I can trigger something to create the user story drafts while we keep collaborating, it might help… and it might not. That’s what I’m trying to find out. Thanks for sharing - I appreciate the feedback!
That could be interesting. Hopefully you come back and share what you learn. My only suggestion would be to restate what I said about prompts and agents. Evaluate your use cases and determine when you need an agent and automation, and when you just need a prompt. Saving Changes...
Community Champion
