Operational clarity under load
Freight forwarding · internal operations · fourteen-week build, then continuous operation with us on call.
The problem
Dispatch was the bottleneck, and not because the team was slow. Every routing decision required reading a transport management system, a customer portal, two carrier feeds, and a shared mailbox — then holding all of it in one head long enough to act. The knowledge lived in the people who had been there longest. Peak weeks were survived rather than managed, and a single person on holiday changed how well the operation ran.
Nobody wanted a new system to learn. They wanted fewer things to look at.
What we did
We spent the first three weeks in the operation, watching decisions get made and writing down the rules that were never written down. That produced a specification of what an operator actually needs to see, when, and what they should never have to look up.
Then we built one operational surface on top of the systems already in place. Nothing was migrated and nothing was replaced. The existing tools remained the systems of record; our layer read from them, reconciled them, and pushed forward only the cases that needed a decision. Where the rules were unambiguous, the system acted and logged what it did. Where they were not, it escalated with the context attached.
What we shipped
- An exception-first dispatch surface: one queue, ranked, with the reason each item surfaced.
- Continuous reconciliation across the transport system, the customer portal, and the carrier feeds, with mismatches raised rather than silently resolved.
- Automated handling of the routine paths, each action written to an audit trail an operator can read.
- Escalation rules authored with the dispatch team, changeable by them without a release.
- Operational health signals and alerting that page a human when the feeds go quiet.
The result
Dispatch now starts the day in one place. The routine cases clear themselves and the operators spend their attention on the handful that genuinely need judgement, with the reasoning for each escalation already assembled. Onboarding a new dispatcher no longer means an apprenticeship next to the longest-serving person in the room, because the rules are in the system rather than in their memory.
The measure the client cared about was simpler than any dashboard metric: peak weeks stopped being an event. The system has been running in production since handover, and we still operate it with them.
Client identity withheld by agreement.