Brex

Brex is the AI-powered spend platform for the world's most innovative companies. Money movement leaves no room for "we'll figure it out tomorrow" — every transaction is a ledger entry someone is waiting on.

The challenge

Payments traffic isn't smooth. It spikes at month-end close, at payroll runs, at the exact moments when a slow path is most expensive. During those peaks the reliability team was flooded with low-signal alerts — latency graphs that turned red without telling anyone why.

The team had observability. What they didn't have was a fast way to go from "p99 is climbing" to "this downstream ledger service is holding a connection pool hostage." Each incident meant assembling a war room and reconstructing the story by hand.

The solution

Brex routed their service logs into Sazabi and let on-call ask the system directly during an incident instead of pattern-matching across six dashboards.

Sazabi cut our mean time to resolution in half during peak load. The first question on every page used to be where do I even look — now it's already answered before the room fills up.

A typical month-end incident now starts with a single question — "what changed in the last ten minutes?" — and ends with a named service, a deploy, and a rollback, before the latency graph has finished climbing.

Results

• 52% lower mean time to resolution during peak windows
• 40% fewer pages that escalate to a war room
• Zero month-end closes delayed by an unresolved incident since rollout