How should AI-generated project tasks be reviewed and tracked?

One thing I'm curious about with AI-assisted project management is what happens after the AI generates the initial work items.

For example, an AI could take a requirement and generate:

• User stories
• Development tasks
• Test cases
• Risks
• BDD scenarios

But generating these artifacts is only the first step. The more difficult part is keeping them connected and making sure the team can see what was generated, what was modified by a human, and what has actually been completed.

I'd be interested in a workflow where AI handles the initial decomposition, while the team validates the output and maintains traceability throughout development and testing.

This is also why I'm interested in platforms like Sharkly, where AI agents can participate in project workflows alongside human teammates rather than operating separately.

https://sharkly.ai/

For those using SpiraTeam, how are you currently reviewing and tracking AI-generated tasks or requirements?

Do you prefer AI to generate the initial work items only, or would you trust it to keep them updated throughout the project?