Agentic AI & team productivity
I apply agentic AI to team software development: agents framed by shared rules, speeding up the cycle from ticket to review without taking decisions away from developers.
What I set up at TF1:
- Introduced the first CLAUDE.md files in the team’s project repositories, with rules, standards and an agent that plans, then asks questions before sensitive decisions
- Tooling for the pull request lifecycle: automatic descriptions, one-command creation, agent-assisted review focused on bugs, security and architecture
- MCP chain: Azure DevOps ticket → Figma mockup → design system → implementation
- Internal talk on harness engineering (October 2026)
- Ticket
- AI agent
- Review
- Shipped
- The agent reads the ticket and plans
- It asks questions before sensitive decisions
- Review and sign-off by a developer



