How Intercom 2x’d their engineering velocity in 9 months with Claude Code | Brian Scanlan

The episode explores how Brian Scanlan, a senior principal engineer at Intercom, led the company’s rapid AI-driven transformation, which doubled R&D throughput in just nine months. The discussion highlights how integrating Claude Code allowed every engineer, product manager, and designer to contribute to code shipping more efficiently, fostering a culture focused on innovation, speed, and quality. The episode underscores the importance of measuring AI adoption through telemetry, building standards into skills repositories, and reimagining technical workflows to be agent-first with CLIs, MCPs, and ephemeral APIs.

Intercom extensively instrumented their internal tools with telemetry to track usage, effectiveness, and quality of their AI Skills. They built a skills repository with hooks enforcing standards automatically, enabling consistent and high-quality output across teams. This infrastructure grants visibility into how often skills are used, their impact on code quality, and allows continuous improvement driven by data. The discussion emphasizes the importance of systemic transparency and measurement to scale AI efforts confidently across large organizations.

The episode details how AI has reshaped development practices at Intercom—remarkably increasing PR throughput, reducing cycle times from code inception to deployment, and enabling backlog elimination. They’ve created a ‘factory-like’ environment where standards are embedded into the code pipeline, leading to more reliable and predictable outputs. The discussion also touches on how AI fosters a culture of fun, speed, and ownership, transforming internal workflows and lowering barriers to innovation.

Intercom’s approach to AI skills involves creating a varied repository of plug-ins and skills, with quality maintained through rigorous evaluation and monitoring. They leverage tools like Honeycomb for real-time usage analytics and session data collection, providing insights for continuous refinement. Skills are designed to be modular, reusable, and integrated deeply into engineering processes, which allows for rapid problem-solving, tech debt reduction, and quality improvements. The team emphasizes the importance of infrastructure, guardrails, and ownership in scaling AI capabilities.

The conversation shifts to how AI is influencing the customer experience—specifically, how agent-driven workflows, API integrations, CLI interfaces, and intelligent automation can empower customer support. They discuss making SaaS products increasingly agent-friendly by building APIs and mini-interfaces that enable users (both internal and external) to perform complex tasks quickly and efficiently, reducing frustrations, and boosting growth. The insights suggest that embracing AI-driven agent workflows will significantly impact product design, customer satisfaction, and business agility.

Brian emphasizes the importance of leadership, permission, and a high-trust culture to foster AI experimentation and scaling. He advocates for giving engineers and teams the confidence and responsibility to push boundaries, even in the face of initial skepticism. Leaders model accountability and emphasize the strategic importance of AI as an existential lever for competitive advantage, encouraging a culture where rapid iteration and learning are normalized. The overall message is that with proper mindset and frameworks, skepticism can be transformed into innovation.

Listeners are left with vital lessons: the significance of systemic measurement, embedding standards into workflows, and fostering a culture of experimentation and risk-taking. The episode illustrates that AI can dramatically boost velocity, quality, and morale if approached thoughtfully. Building trust, metrics, and infrastructure upfront creates a sustainable path to continuous improvement, and organizations should prioritize getting their developer workflows agent-friendly to remain competitive in an AI-driven future.