Anthropic's internal teams use Claude Code to automate tasks, bridge skill gaps, and accelerate workflows across departments from data infrastructure to legal.
- The Data Infrastructure team uses Claude Code to automate data engineering, debug Kubernetes issues via screenshots, and empower non-technical finance staff with plain-text workflows.
- The Product Development team rapidly prototypes features with Claude Code's auto-accept mode and builds core functionality through synchronous iterative prompting.
- The Security Engineering team reduces incident resolution time by using Claude Code to trace control flow from stack traces and review Terraform infrastructure changes.
- The Growth Marketing team automates ad creative generation and builds a Figma plugin, achieving a 10× increase in output with agentic workflows that run without dedicated engineering support.
- The Product Design team directly implements front-end changes and rapid prototypes via Claude Code, bridging the gap between design and engineering execution.
Adoption across departments
Claude Code is adopted by teams ranging from data infrastructure and security engineering to growth marketing and product design. Each team tailors the tool to its domain, using features like auto-accept mode, MCP servers, and Claude.md files to automate tasks and bridge skill gaps.
The tool enables both technical and non-technical staff to execute complex workflows—finance teams without coding experience can run data pipelines, and designers can directly implement front-end changes.
Key use cases and workflows
Engineers use Claude Code for fast prototyping with auto-accept mode, debugging infrastructure by feeding screenshots, and generating comprehensive test suites. Marketing builds agentic workflows for ad creation, and security engineers analyze Terraform plans and synthesize runbooks.
Dogfooding is a common practice: the Claude Code team itself uses the product to build new features, and the API Knowledge team tests model updates through the tool, providing direct feedback.
Best practices and recommendations
Teams recommend writing detailed Claude.md files, using MCP servers for sensitive data, and sharing usage sessions to spread best practices. For complex tasks, breaking them into specialized sub-agents and using iterative prompting rather than one-shot requests yields better results.
Non-developers should get help with initial setup and use custom memory files to guide Claude Code's behavior, while all users benefit from saving state before letting Claude work autonomously.
Read this at any depth.
Install Depth and pick your level — Glance for a sentence, Summary for the gist, Read for the full take. Free daily quota, no signup needed.
Add to Chrome