A logistics company with growing volume needed a system to automatically triage incoming shipments, flag exceptions, and route decisions to the right team without human intervention on standard flows.
Pinnacle operates 12 distribution centres processing over 8,000 shipments daily. Each centre had its own manual triage process — email, spreadsheets, phone calls — and the central operations team was spending 30+ hours per week just routing standard shipments.
Exceptions (damaged goods, address changes, carrier issues) were handled reactively, often taking hours to surface because the triage process didn't differentiate between routine and exception flows.
We designed an AI triage layer that sits between their order management system and the centre-level dispatch teams. The system classifies each shipment into one of three paths: auto-route (standard), flag-for-review (ambiguous), or escalate (exception requiring human decision).
The routing logic was trained on 18 months of historical dispatch decisions. We built a feedback loop so that every human override re-trains the model, continuously improving auto-route accuracy.
A real-time dashboard gives the central team visibility into all 12 centres, with alerts when any centre's exception rate exceeds its normal threshold.
Standard shipments now route automatically in under 90 seconds — down from an average of 12 minutes. Exception handling time dropped by 55% because the system surfaces them immediately with context attached.
The central operations team reclaimed 25 hours per week, which they redirected to process improvement work. Six months in, auto-route accuracy reached 94%, and the model continues to improve with each override.
EXPERIENCE PREVIEW
12 distribution centres running on one unified triage system. The work connected customer experience, internal clarity and measurable commercial movement rather than treating them as separate design problems.