AI Team Execution Layer
The execution layer turns workforce definitions into executable work.
A Digital Employee’s job and skills can remain stable while the mechanism used to execute them changes.
Execution mechanisms
Depending on the environment, execution can occur through:
- plugins
- agents
- scheduled tasks
- coding environments
- model-native workflows
- other supported AI runtimes
The execution mechanism is therefore an implementation choice, not the definition of the Digital Employee.
Plugins as an execution layer
Plugins package selected AI Team capabilities so they can execute inside supported AI environments. A plugin can expose Digital Employees, skills, orchestration resources, templates, and approved tools required for a particular workflow.
Free plugins can provide executable implementations of selected Capture or Proposal Team capabilities. They are an entry point into the architecture, not the entire enterprise deployment model.
Keeping the plugin separate from the canonical job definition also allows the same Digital Employee or skill to execute through another environment later.
Model runtimes
The model provides reasoning and generation capability. More capable models may perform several Digital Employee jobs within one runtime while maintaining the logical job boundaries defined by the AI Team.
The architecture therefore does not assume that increasing model capability requires eliminating roles, skills, workflow state, or governance.
Tools and protected capabilities
The execution layer can invoke approved tools and enterprise integrations. MCP can be one mechanism for exposing customer-controlled resources or Compound Leverage-controlled capabilities without placing the implementation itself into model context.
See Model Context Protocol and AI Teams.
Next: Customer AI Control Plane.