# MENA Dealership Service-Centre Scheduler

After-sales service centres run the same play every day. Receive job cards. Decompose them into tasks. Assign tasks to bays and technicians under capacity, parts, and skill constraints. Push status updates to customers. We replaced the manual dispatch board with an agentic orchestration layer that reads each job card, decomposes it into tasks, runs a proprietary scheduling optimisation engine for the day's timesheet, auto-assigns work, and broadcasts customer-facing status updates as the day unfolds.

## Key facts

- **Industry:** AUTOMOTIVE
- **Client (abstracted):** A global automotive OEM operating dealership service centres across MENA
- **Timeline:** PoC, multi-month
- **Role:** Agentic orchestration over an operational workflow
- **Year:** 2024

## The challenge

Manual dispatch boards in busy service centres produce cascading delays. Bays sit idle while job cards queue. Technicians switch context. Customers chase updates because nobody owns the inflight view.

A scheduler that survives the floor needs to respect a long tail of constraints: bay capacity, technician skill, parts availability, and the realities of a working day. Off-the-shelf optimisers typically need handholding to encode these well.

## The solution

Built an orchestrator agent that ingests each job card, decomposes it into tasks, and routes the request to a proprietary scheduling optimisation engine. The optimiser returns the day's timesheet under the full constraint set; the agent assigns tasks back to the operational system and triggers customer notifications.

Packaged the optimiser as a containerised service on Google Cloud Run. The agent runtime calls it as a tool, keeping the orchestration layer thin and the optimisation logic isolated and independently versionable.

## Outcomes

- A working PoC that closes the loop from job card receipt to assigned timesheet to customer status, without manual board work.
- A reusable orchestration pattern: agent + optimisation engine + customer-facing notifications, ready to extend to other after-sales workflows.
- A foundation for a production rollout across the network when the OEM moves out of PoC.

## Impact

- Real-time — Day-of timesheet generation
- Multi-bay — Capacity-aware assignment
- Auto — Customer status broadcasts
- PoC — Engagement type

## Tech stack

- FastAPI
- Python
- Scheduling Optimisation Engine
- Google Cloud Run
- Docker

## Related

- [All work](https://levent.ai/work/)
- [Next: Real Estate 3D Visualisation Platform](https://levent.ai/work/terravia/)

Client names and real outcome metrics are not published. See https://levent.ai/ai-content-policy/ for the abstraction policy used across this site.

---

**Canonical URL:** https://levent.ai/work/pitlane/