AITraining2U
Working demo

Daily Decision List — Logistics Operating Model

The executive view a Malaysian 3PL CEO opens every morning, rendered live from the case study's mock data. Each decision is justified by the agents that produced it, with the signal-window that triggered it shown inline.

The operating model behind this dashboard click to collapse Layer 1 — Signals POS · BasketsReal-time, per site Weather (MET)PM2.5 · Rainfall IoT (chillers, HVAC)Telemetry · Faults HRIS · WorkdayRosters · OT · Skills Competitor + Festive calPriceSpy · MET · KPDN circulars Layer 2 — 5 Specialist Agents Volume Agent Parcel forecasting Routing Agent Network optimisation Driver Agent Roster & licensing Fleet Agent Vehicle health Hub Agent Hub portfolio Layer 3 — Master Orchestrator Chief of Staff Synthesises the 5 specialists into 1 ranked list CEO · Daily Decision List at 06:00 MYT Shared Memory ↔ all agents Obsidian vault (markdown, git) Pinecone (per-agent namespaces) MCP tools: search · write · outcome Nightly sync: outcomes → playbooks (Build-It Steps + KG Steps 1-5) Approve → Learn loop CEO clicks Approve / Defer / Escalate → outcomes.csv updated → CSOA nightly retrain @ 02:00 MYT → agent playbooks refresh → tomorrow's list is sharper
5 specialists → 1 orchestrator → 1 ranked list · the same architecture shown on slide 7 of the logistics case study, instantiated for this demo. Click any decision below to see how each agent justified its part.

24-hour P&L scope · selected day

Approve / risk if not addressed

Portfolio tier snapshot

Over-performing
On-target
Under-performing
Open site scatter — point out the optimisation opportunities
Over-performing — replicate On-target — protect Under-performing — cull or fix
Each dot is one of the 60 hubs in entities.csv. X-axis: Daily parcels ('000). Y-axis: Vehicles. Quadrant labels are the Portfolio Agent's standing recommendation for hubs that land in each zone.

Ranked decisions