When building AI systems, sprint planning can get interesting once agents become part of the development workflow.
A sprint might involve a researcher testing models, a developer integrating an API, an AI agent handling implementation, and another agent running tests or preparing documentation.
Traditional sprint planning doesn’t necessarily account for this kind of mixed team.
I’ve been exploring Sharkly.ai because it approaches sprint planning around a workflow where humans and AI agents can work on project tasks together.
The interesting question for me is whether AI agents should simply execute tasks that humans assign—or whether future sprint planning tools should help decide which tasks agents can take on, how much work they can handle, and when humans need to review the results.
For people building with Hugging Face models and agents:
How would you design an AI-native sprint?