My AI Agent Isn't Allowed to Decide Anything

Описание и применение

Система Shot‑Delivery Guardian автоматически определяет, какой элемент пост‑продакшн‑потока можно отложить, соблюдая правила и не принимая окончательное решение, тем самым ускоряя работу людей.

Задача

When a film or TV show’s post‑production pipeline is near a deadline, a human must decide which work to push back to keep the schedule, but this decision is time‑consuming and error‑prone.

Результат

In real tests, 3 of 5 matches with the fixed rule alone, 0 broken rules, and an average response time of 44 seconds.

Рабочий процесс

• 1. Intake, rendering, color, quality checks, delivery, scheduling, test generator, and AI modules run independently.
• 2. Each module reports status to a shared connector.
• 3. AI queries three monitoring tools: overall backlog, work metadata, and pipeline stage timing.
• 4. AI applies the rule to filter out client‑approved, director‑flagged, and downstream‑dependent shots.
• 5. AI selects the shot with the most slack and writes the recommendation to the dashboard.
• 6. Human on call reviews the note and acts.
• 7. System logs the decision automatically without manual login.

Доказательства

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.

Источник

Zero broken rules. 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.

Источник

The AI never got an important decision wrong, because I never let it make one.

Источник