
The Problem: Modern engineering teams spend too much time on "work about work"—triaging tickets, assigning tasks, and manually updating project boards. Project Managers (PMs) are often bottlenecks in the flow of information. The Solution: We built the AI PMO Agent, an intelligent assistant that lives inside IBM watsonx Orchestrate. It acts as an autonomous Project Management Officer. Instead of clicking through menus, a user simply says, "Create a high-priority bug ticket for the Login issue in the UniUnity project," and the agent handles the rest. Technical Innovation (The "Secret Sauce"): While building this solution, we went beyond the standard implementation. We engineered a custom, lightweight runtime architecture: Optimized Performance: Instead of using the resource-heavy standard Docker ADK (which requires 14GB+ RAM), we built a custom FastAPI server. Structured Intelligence: We implemented comprehensive Pydantic models to define our API schemas. This allows the AI (Llama 3.2 90b) to understand exactly what data to extract (e.g., Project IDs, Priorities, Member Names) with zero hallucinations. Hybrid Deployment: The solution utilizes a secure Ngrok tunnel to link the IBM Cloud SaaS interface directly to our local high-performance Python execution environment. Key Features: Seamless Integration: Deep integration with Plane.so API to manage projects, issues, and members. Context Awareness: The agent understands project context, priority levels (High/Urgent), and assignment logic. Zero-Code Interface: Users interact via the intuitive watsonx Orchestrate chat, while the complex logic happens in our custom backend. Impact: This project demonstrates how Agentic AI can bridge the gap between unstructured human conversation and structured business systems, reducing administrative overhead by over 40%.
23 Nov 2025