My AI Agent Isn't Allowed to Decide Anything

Developer Dann Waneri has introduced 'Shot-Delivery Guardian,' an AI-driven system designed to manage production deadlines in media workflows. Rather than granting the AI full autonomy, Waneri implemented a 'human-in-the-loop' architecture where the AI performs data analysis and reporting, while critical decisions are strictly governed by hard-coded Python rules. This approach ensures that the AI cannot override client approvals, director flags, or downstream dependencies. By separating the AI's explanatory capabilities from the deterministic logic of business rules, the system maintains high reliability in high-stakes environments. During testing, the agent successfully identified optimal tasks to delay without violating any constraints, achieving a 60% success rate in real-world scenarios. The project emphasizes that AI is most effective when used to process complex information and provide actionable insights, rather than being entrusted with final decision-making authority. The source code is available on GitHub under an MIT license.