Shopify's VP of Engineering describes AI-first strategy: standardizing infrastructure rather than tools, achieving roughly 20% productivity gains, and preparing for agentic AI in 2026.

Shopify built a centralized LLM proxy (an internal gateway routing all AI requests through one platform layer) rather than standardizing on a single AI tool. This allows engineers to experiment with Claude Code, GitHub Copilot, and other tools simultaneously while maintaining centralized cost control, usage analytics, and the ability to switch models as capabilities improve or costs change.

Farhan Thawar, VP & Head of Engineering at Shopify, estimates his team is 20% more productive, but the gains show up not in traditional metrics like lines of code or pull requests but in faster prototyping, exploring more approaches, and higher-fidelity deliverables. Weekly demos showing tangible velocity are used to measure progress.

Shopify connected AI to internal systems via MCP servers (tools allowing AI to query information from wikis, product management tools, and data warehouses) while preserving access controls—AI only retrieves information the user already has permission to see. The company also built an internal tool called Quick that lets employees drag and drop JavaScript, TypeScript, or HTML files to instantly deploy simple software.

Farhan warns that comprehension debt is the number one long-term risk: if engineers stop thinking deeply and learning, they will lose understanding of their systems. His guardrail is that engineers must understand systems 2–3 layers below where they are working, using AI to accelerate learning rather than replace it.