Austin · Lisbon · KarachiTrust centredev@binarybreach.tech
BinaryBreach
Case study — Global logistics operator

An agent fleet that negotiates delays across carrier APIs, email and EDI, escalating to humans only when confidence drops.

SectorLogistics & supply chainDuration14 weeksStackClaude · MCP
Freight lorries loading at a distribution hub
71%Auto-resolved
−19%Detention cost
24/7Coverage
ReplayableEvery action audited
The challenge

Where it started

Exception volume had outgrown the operations team. Each delayed consignment required chasing a carrier across whichever channel that carrier used — REST, EDI, or a shared inbox — then making a commercial judgement about rebooking. Night-shift coverage was thin and detention charges were climbing.

The approach

  1. 01

    Normalised carrier communications into a single event model spanning API, EDI and parsed email, so an agent reasons over one representation.

  2. 02

    Gave agents a narrow catalogue of idempotent, reversible actions with hard commercial limits encoded outside the prompt.

  3. 03

    Calibrated a confidence threshold for escalation against six weeks of historical decisions, tuned to the client's tolerance for a wrong rebooking.

  4. 04

    Recorded every action, input and rationale in a replayable audit log that operations reviews each morning.

Stack
  • Claude
  • MCP
  • Go
  • Temporal
  • Postgres
  • GCP
The outcome

Where it landed

Seventy-one percent of exceptions now resolve without human involvement, with round-the-clock coverage. Detention costs fell nineteen percent in the first two quarters, and the operations team moved to reviewing escalations rather than chasing carriers.

The only vendor who pushed back on our roadmap — and they were right. We shipped something smaller and it actually worked.
Chief Product OfficerGlobal logistics operator
Capabilities used

More work

A short conversation with an engineer, not a sales qualification call. If we're the wrong people for it, we'll say so and point you somewhere better.