Boss wants us to add more AI to our workflow
I’m working on a Django project running in Docker with multiple services (PostgreSQL, Redis, Celery). The codebase is fairly large and not strictly modularized.
Currently, I use GitHub Copilot (ask mode) to generate code snippets based on a described approach, which I review and adapt manually. This works well for me.
I now have to move toward semi-autonomous AI agents for my boss that can:
implement isolated features or refactors
run tests
submit changes for review (via pull requests)
Before introducing agents, I want to improve the structure and workflow to make this safe and maintainable.
What architectural or workflow changes would make a Django project more suitable for AI-assisted or agent-driven development?
focus on modular apps, service layers, isolated APIs, strong tests, CI/CD pipelines, and safe AI execution. This ensures that semi-autonomous agents can propose and validate changes easily without introducing instability.
What worked best for me and my team was not trying to fully redesign the project first, but making the implicit project knowledge explicit. We created an AGENTS.md file explaining the architecture, main Django apps, domain concepts, common patterns, and things to avoid. We also added project-specific skills for running Docker, running tests, reviewing migrations, working with Celery, and doing code review.
But if you really need that, I would suggest to use Django as a wrapper of your business logic. Create you logic in services, and the Django views or celery tasks just call those services to apply the business logic. I saw people doing it on separated classes, functions, or even adding the responsibility to model methods...
My advise, write a prompt to ask you AI agent to scan your codebase and measure how big and risky would be that migration. And if possible, go in rounds, with a few modules and making sure you have tests to keep the same funcionality.