Copilot agents for robotics teams
As AI Enablement Engineer at ABB Robotics, I brought AI into the way engineering teams already worked: agents, prompts and skills fitted to each team’s process, gathered into a library they could build on.
AI that fits how a team works.
Each engineering team at ABB Robotics has its own way of building software, so I tailored GitHub Copilot agents, prompts and skills to each team’s SDLC workflow.
The pieces that worked became a reusable library of AI developer tools for code generation, review and documentation. It grew into an internal platform other engineers built their daily workflows on. Later I extended the same approach with the Claude API and MCP-based agent tooling.
- Personalised Copilot agents for each team
- A shared library of prompts and skills
- Code generation, review and documentation
- Self-serve setup for every team
- Agents built on the Claude API and MCP
- Role
- AI Enablement Engineer
- When
- Jan – Jun 2026
- Teams
- Four engineering teams
- Stack
- Copilot, Claude API, MCP
From one team to a platform
Fit each team
Studied each team’s SDLC workflow and built agents, prompts and skills that matched it.
Build the library
Collected what worked into a reusable library of AI developer tools.
Hand it over
Documented the workflows and set every team up to run its own Copilot agents after deployment.
Add a second ecosystem
Extended the tooling with the Claude API and MCP-based agents.
Built to be owned by others
Shaped around real teams
Agents, prompts and skills followed each team’s existing SDLC workflow.
Documented for self-serve
Every workflow was written down so teams could set up and adjust their own agents.
One owner for adoption
I was the internal owner for AI tooling adoption, from documenting workflows to setting each team up to self-serve.