Car delivery dispatchers waste critical time jumping between fragmented tabs to track down drivers, update customers, and resolve minor transit delays. When everything is manual, small routing issues quickly cascade into missed delivery promises and overwhelmed support desks, causing a chain reaction of delays.
We built a unified operations dashboard featuring a priority-sorted queue and an email-style inbox. Instead of replacing the dispatcher, the system acts as a highly capable assistant, gathering context and teeing up actionable resolutions for humans to approve.
The backend runs an autonomous ReAct loop powered by a local Llama 3.2 model. When a delivery issue arises, the AI agent evaluates the case facts and uses strict JSON tool schemas to fetch real-time routing margins from the Postgres database. It can autonomously trigger a driver simulator to request status updates, synthesize the replies, and formulate a concrete action plan (like drafting a delay notification email) all before handing the case over to the human operator for final approval.
Our immediate next steps would be to swap the simulated driver communications for a live SMS gateway, integrate real-time vehicle telemetry, and pilot the dashboard with a small team of active dispatchers.
