How Intercom 2X'd engineering velocity with Claude Code | Brian Scanlan

2026-04-20 · Source: How I AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Data Science & Analytics · Depth: Intermediate, extended

Summary

Intercom's R&D department has doubled its throughput, measured by pull requests (PRs) per R&D head, in nine months by fully embracing AI-assisted development, primarily using Claude Code. This initiative, driven by a company-wide "AI-first" strategy and a CTO-set goal, has not only increased velocity but also improved code quality, as indicated by a Stanford research group's analysis. The company manages this transformation by treating its internal AI adoption like a product, implementing extensive telemetry with Honeycomb and Snowflake for skill usage and session data analysis. They've also developed an internal skills repository, distributed via IT systems, to standardize and enhance developer workflows, exemplified by a "flaky specs" skill that autonomously fixes recurring test issues.

Key takeaway

For Directors of AI/ML and VPs of Engineering aiming to significantly boost R&D velocity, you should commit to an "agent-first" development paradigm. Prioritize investing in AI tools and internal skill development, even if initial costs are high, and establish robust telemetry to monitor usage and quality. This approach can lead to substantial throughput gains and improved code quality, transforming your team's capacity to tackle technical debt and deliver features faster.

Key insights

Intercom doubled R&D throughput and improved code quality by fully integrating AI-assisted development and treating internal AI adoption as a product.

Principles

• Treat internal AI adoption as a product.
• Prioritize speed and investment over immediate cost optimization.
• Foster a high-trust culture for AI integration.

Method

Implement comprehensive telemetry for AI skill usage and session data. Develop an internal, IT-distributed skills repository. Enforce quality standards for AI-generated outputs (e.g., PR descriptions) via automated hooks and LLM judges.

In practice

• Instrument internal AI skills with telemetry (e.g., Honeycomb).
• Analyze raw AI session data for organizational insights.
• Invest in fixing tech debt using AI-driven automation.

Topics

• Engineering Velocity
• Claude Code
• AI Adoption Strategy
• Developer Experience
• Code Quality Improvement

Best for: AI Engineer, Director of AI/ML, Software Engineer

Related on AIssential

• How Intercom 2x’d their engineering velocity in 9 months with Claude Code | Brian Scanlan — Deep AI integration and telemetry can double engineering throughput and improve code quality.
• 🎙️ This week on How I AI: How Intercom 2x’d their engineering velocity with Claude Code — Intercom doubled engineering velocity in nine months by deeply integrating Claude Code, emphasizing telemetry and cultural permission.
• v2.1.80 — Recent Claude Code updates enhance agent capabilities, user experience, and system stability through new features and critical bug fixes.
• We're sharing our internal AI engineering cheatsheets — Pre-made engineering decisions and architectural guidelines accelerate AI development and improve output quality.
• AI Made Your Engineers 10x Faster and Your Product 10x Worse — AI coding assistants boost speed but degrade product quality, necessitating a rebuilt quality layer.
• v2.1.87 — Claude Code updates significantly enhance AI model capabilities, user experience, and extensibility through plugins and agent teams.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by How I AI.