My AI Agent Isn't Allowed to Decide Anything

说明与用途

Shot‑Delivery Guardian 使用 AI 分析后期制作流程，并推荐安全的任务延迟，但最终决策由固定规则做出，确保满足客户和导演的约束。

问题

When a film or TV production nears a delivery deadline, a bottleneck in any post‑production step can cause a missed deadline unless a human manually delays another task. The manual process is slow and error‑prone.

结果

In live tests, the system correctly identified safe tasks 60% of the time (3 of 5), matching the fixed rule’s choice. It reported issues in 44 seconds on average and never violated any safety rules.

工作流

• 1. Monitor pipeline stages via three monitoring tools (overall backlog, task metadata, and stage‑level timing).
• 2. AI queries each tool to gather current state.
• 3. AI applies the fixed rule to filter out client‑approved, director‑flagged, or dependent tasks and selects the one with the most slack.
• 4. AI writes the recommendation as a note on the existing dashboard.
• 5. Human on call reviews the note and confirms the delay.
• 6. System logs the decision and updates the pipeline state.

证据

Three out of five. That's how often Shot‑Delivery Guardian, the system I built for Google's Agentic Cinema hackathon, picks the right answer when it's the one making that call, in real tests against the actual running system, not just on my laptop.

来源

It never once picked a piece of work the client had already signed off on. It never once picked something other work was waiting on. It never once touched anything the director had flagged as important.

来源

About 44 seconds on average to look into a problem and respond, timed against the real live system.

来源