How HubSpot Scaled AI Adoption (product.hubspot.com)

HubSpot detailed how it moved from cautious experiments with GitHub Copilot (started Summer 2023) to near-universal adoption of AI coding tools across engineering by combining executive buy-in, rigorous pilots, and measurement. Early pilots included whole teams, two-month trials, training channels, and quantitative metrics (code-review burden, cycle time, velocity, incident rates). Those metrics repeatedly showed modest but real productivity gains and no correlation with increased production incidents, which justified broader rollout. In May 2024 HubSpot removed license restrictions and adoption jumped past 50% overnight, later climbing to ~90% as AI fluency was made a hiring expectation. The technical and organizational levers matter: HubSpot created a small Developer Experience AI team (Oct 2024) to supply guardrails, curated defaults, and infrastructure—sharing Cursor rules, deploying local MCP servers on developer machines, running multi-vendor POCs (OpenAI, Anthropic), and adapting procurement to month-to-month models. This centralized platform work let product teams adopt AI without sacrificing architectural opinions or quality. The investment unlocked agentic use cases (Sidekick for platform Q&A, PR reviews, issue creation), a growing catalog of 400+ agent-accessible tools, and a playbook for scaling AI beyond POCs: measure impact, provide tailored context and rules, enable fast procurement, and build central infrastructure to multiply benefits across teams.