Meridian — Replacing fourteen internal tools with one operations platform
A custom-built operations console for a freight forwarder, consolidating dispatch, exception handling, and customer comms into a single deterministic interface.
Meridian moves shipping containers between forty-three ports. Their operations team was working across fourteen separate tools — three TMS systems, two ETA trackers, a homegrown PHP exception queue, four spreadsheets, and a Slack bot held together by a contractor who'd left in 2022. Average exception resolution time was four hours and forty-two minutes. Most of that was tab-switching.
We built a single console — Next.js on Vercel, Postgres on Neon, Inngest for the long-running workflows — that presents one container at a time, with every relevant action surfaced in context: rebook, reroute, notify, escalate. AI agents handle the routine paperwork in the background: generating customs documents, drafting customer updates, reconciling carrier invoices against contracted rates. Humans handle the judgment calls.
The design discipline was reductive throughout. Every screen, every field, every notification had to justify its existence against a single question: would removing this make the operator slower? Most of them couldn't, and got cut.
Resolution time is now thirty-eight minutes. The ops team reorganized around exception severity instead of which tool a problem happened to live in. The replaced systems are decommissioned; the contractor's PHP queue was archived to a single S3 bucket on the day we shipped.