# How Blitz Cut Root Cause Analysis From Days to Minutes

## Metadata

- **Canonical HTML**: https://ctrlb.ai/customers/blitz
- **Customer**: Blitz (logistics and supply chain)
- **Use case**: Agent-led incident response and AI agent observability

## Behind Blitz's Agent-Led Incident Response with CtrlB

Blitz is a quick commerce fulfilment and last mile delivery platform. D2C and e-commerce brands use it to promise delivery in 10 minutes, 60 minutes, 2 hours, same day or next day, and Blitz runs the dark stores, fulfilment centres and rider network behind that promise.

The system behind it is highly distributed, and more than 25 agents run on top of it, most of them facing customers directly. Observability has two jobs: explain what broke in the infrastructure, and show how the agents are behaving.

**This is where CtrlB came in.**

## The challenge: The observability system was the first thing to go down

Datadog did not fit the way Blitz's ecosystem is built, so the team ran its own observability system, with logs and metrics but no traces. During one incident that system was down as well. No data existed for the incident window, and although service returned in reasonable time, the final RCA took days.

Without traces, every incident starts by asking which service is the bottleneck and answering by elimination.

## The solution: Logs and traces in one lake, with CtrlB MCP in the incident path

CtrlB holds logs, metrics and traces on one lake. Blitz runs CtrlB's MCP server alongside its AWS and GitHub MCP servers. When any alerting system fires, an agent runs the RCA, scans the code for context, raises a pull request with a fix and informs the team.

- **Traces and logs, correlated**: the failing service and the log line that explains it come out of the same investigation.
- **Plain language instead of query syntax**: engineers get back the instance, the API and the function that is breaking, and why.
- **Anomaly detection on agent behaviour**: alerts fire on conversation patterns outside what a use case permits, which surfaces misuse and jailbreak attempts early.

## The results: Root cause in 5 to 7 minutes, and a 99.99% uptime target

### Time to root cause: from days to 5 to 7 minutes

The whole alert-to-fix cycle takes about 5 to 7 minutes, against days to establish the final RCA on the previous stack.

### 25+ agents in production

Blitz runs more than 25 agents, including in-house voice agents on over 110,000 minutes a month and a support agent on more than 25,000 conversations a day.

### From firefighting to a 99.99% uptime target

Blitz works to 99.99% availability and treats latency and availability as goals rather than things it reacts to. Next is onboarding onto CtrlB's self-improving agents.

- **Markdown mirror**: Fetch `/customers/blitz.md` for this page as plain Markdown