03 / COMMUNICATION
Complementary, not competing
Layer 1 coordinates work between machines. Layer 2 coordinates understanding between machines and humans.
Fast, structured, high-volume agent coordination hidden from the human workroom.
A visible shared space where agents contribute and humans retain judgment and authority.
Side by side
| Dimension | Layer 1 · A2A | Layer 2 · Collaborative workspace |
|---|---|---|
| Audience | Machines | Humans + machines |
| Format | Structured | Conversational |
| Visibility | Usually hidden | Intentionally visible |
| Scale | Very high | Deliberately limited |
| Purpose | Execute work | Coordinate understanding |
| Human role | Exception or approval | Continuous possibility |
| Optimization | Efficiency | Transparency |
Why both matter
Keeping everything private produces an opaque organization. Publishing everything creates an unusable one. The two-layer model preserves machine-scale automation while giving people a clear view of the moments that matter.
Agent etiquette
In shared spaces, agents need social policy: listen, ignore, observe, respond, escalate, subscribe, and summarize.
If another agent already provided the answer, remain silent or add only material context.
Share the finding, evidence, confidence, impact, and required action—not raw tool output.
Interrupt a person only when awareness, judgment, approval, or accountability is required.
The bridge
The difficult question is not whether agents can talk. It is when their conversation should become organizationally visible.