Abraham Heward

What if a work item could move itself from review to dev to QA to release to production monitoring? I think that's where this goes next. Not better code generation. Work items driving the whole SDLC. Right now, my workflow still starts with me. I take a work item, clarify the intent, shape the scenarios, and then open the agent. That works well. But I don't think that should be the end state. The more interesting future is a standard CI/CD flow where agents take over more handoffs. Something like this: Work item is written in English. It moves to Ready for Review. That status change triggers a product owner agent. It adds acceptance criteria, happy-path BDD scenarios, and an implementation plan. Then it moves it to Ready for Dev. That activates the implementation agent. It moves the work item to In Development, implements the feature, adds the tests, and keeps going until the slice passes. Then the work item moves to Ready for QA. That activates a QA agent which pushes beyond the happy path: broader scenario coverage, more edge cases, more regression protection. Next comes Ready to Deploy. That activates the release agent. It runs any remaining validations, then starts the CI/CD pipeline to deploy to staging. After staging passes, the same flow promotes to production and hands off to a production monitoring agent. In my case, that would mean watching Application Insights for feature-specific errors and warnings, looking at traces and logs, and reacting quickly if something is off. If production reveals a defect, the monitoring agent creates a bug work item. That triggers a bug-fixing agent, which works the same way as the original feature flow: update the work item, implement the fix, validate it, and move it back through delivery. That's not a new software lifecycle. It's the same SDLC most teams already understand. Same CI/CD. Same work items. Same environments. Same quality gates. The change is that more of the flow is executed by specialised agents instead of being manually pushed from one step to the next. And because the agents keep updating the work items as they go, the process stays visible and traceable for the humans around it. I don't mean removing people from software delivery. I mean reducing the dead time between steps, making handoffs explicit, and keeping the process moving with better traceability. And yes, I said story and feature in earlier drafts, but work item is really the better term here. That's where I think this goes next. Not one giant autonomous coder. A delivery system of coordinated agents working inside a normal engineering process. The next question is what that does to the human role. If you're experimenting with anything similar, I'd be interested in how you're thinking about agent handoffs across review, development, QA, release, and production monitoring. #GitHubCopilot #AgentMode #AzureDevOps #CICD #BDD #ApplicationInsights #AIEngineering #DeveloperProductivity

3