Table of Contents

5. Agent-to-Agent Collaboration

At a Glance
Format Demonstration
Participants create Nothing
Key message Agents collaborate through clear responsibilities, controlled actions, and human approval
Duration to be defined

Objective

Show a realistic operational scenario in which several agents collaborate. Avoid unrealistic fully autonomous scenarios.

Concept What to show
Context passing One agent hands over enriched context to another
Separation of responsibilities Each agent has one clear role
Controlled actions Agents act only within defined boundaries
Human approval points Operators approve impactful actions

What Is Given

Item Details
Demo environment Pre-built agents on a prepared DataMiner System
Demo scenario A scripted operational situation with a clear start and end
Fallback recording A recording of the full demo

Example Scenario (Media & Broadcast)

sequenceDiagram
    participant M as Monitoring agent
    participant R as Event readiness agent
    participant O as Operator
    participant C as Communication agent
    M->>R: Encoder errors on Tier 1 event starting in 45 min
    R->>R: Apply go/no-go policy for Tier 1 events
    R->>O: Propose switch to backup encoder
    O-->>R: Approve
    R->>C: Action taken, event back on track
    C->>O: Draft update for the rights holder (for review)

Before the Session

  • [ ] Run the scenario end to end and reset it
  • [ ] Have the fallback recording ready
  • [ ] Prepare the explanation of each agent's role and boundaries

Session Flow

# Trainer Participants
1 Introduce the scenario as an operational situation, not as a technology Listen
2 Introduce each agent and its responsibility Watch
3 Trigger the situation and show the context being passed Watch
4 Pause at the human approval point and explain why it is there Discuss
5 Complete the scenario and show the outcome Watch
6 Link back to their own agents: "Which role would your agent play here?" Reflect
7 Close with the final training message Listen

Do's and Don'ts

Do Don't
Show boundaries and approvals explicitly Present a fully autonomous "lights-out" scenario
Keep the scenario realistic and close to the workshops Use a scenario unrelated to operations
Explain why each agent has a separate role Hide failed steps – switch to the recording

Success Criteria

  • Participants understand how agents with separate roles collaborate safely
  • Participants see where humans stay in control

Closing

End the training with the final training message.