Skip to content
NTPU AI Learn

AI 301 Apply

Design how a real team of agents divides work, communicates and coordinates, with a Canvas

What the Canvas is for

The Multi-Agent Architecture Canvas is the capstone of AI 301. For one real use case you state the shared goal and why one agent is not enough, define the roles and their responsibility boundaries, decide how tasks are allocated and delegated, define the messages agents exchange, separate shared from private context, choose the orchestration pattern, mark the handoff points, plan conflict resolution and human oversight, and decide how you will evaluate whether collaboration helped. Human, AI, Human: a person defines the task, the agents assist or execute, and a person reviews and stays responsible.

What you fill in

  • Shared Goal and Why Multi-Agent?
  • Agent Roles and Responsibility Boundaries
  • Task Allocation and Delegation Rules
  • Communication Protocol
  • Shared Context & Memory
  • Orchestration Pattern and Handoff Points
  • Conflict Resolution and Human Oversight
  • Evaluation and Risks & Boundaries
  • Final System Output, and the reflection question

Warning

A high-impact or irreversible decision should never be settled by agents alone. State which outputs and decisions a person must confirm before anyone relies on them.

Examples

The canvas page shows a worked example (preparing a grant proposal) and ideas for other occupations: customer support, ESG report review, research literature scan, financial analysis and administrative operations. All use simulated data.

Before you submit

Warning

Do not enter secrets, credentials, personal data, client data or sensitive information you are not authorized to share. Use simulated or public examples only.

What comes next: AI 302

AI 301 teaches how multiple agents divide work, communicate, delegate, coordinate and resolve conflicts. AI 302 (Agent Harness Engineering & Reliable Automation) builds on it by looking at how such a system runs reliably over time, with state, checkpoints, verification, recovery and observability.

Academic mapping: M5276 Week 13

AI 301 maps primarily to Week 13 of M5276 Agentic AI Innovative Applications: Multi-Agent Collaboration & Interoperability: Role Division, Communication Protocols, Agent-to-Agent Delegation and Orchestration. The topics covered are role division, communication protocols, agent-to-agent delegation, shared context, orchestration, and conflict and oversight.

Durable execution, checkpoints and production reliability belong to the following course, so they are not part of this one.

Go to the canvas

What you need to complete

  • Complete all eight AI 301 modules
  • Score 80% or higher on the post-test
  • To earn the Multi-Agent Architecture Capability Certificate, also submit a Multi-Agent Architecture Canvas and have it reviewed and passed against the rubric by a Credential Admin

How you are assessed

A Credential Admin scores the Multi-Agent Architecture Canvas across seven weighted domains: Goal & Multi-Agent Justification, Role & Responsibility Design, Task Allocation & Delegation, Communication & Shared Context, Orchestration & Handoff, Conflict Resolution/Human Oversight/Evaluation, and Risk Management/Final Output/Reflection. A PASS requires at least 17 of 21 points with no critical-fail rule triggered; 14 points or fewer, or any triggered critical-fail rule, is a FAIL.

Info

This human-review rubric is separate from the automatic advisory score shown right after you submit (which scores each of the fifteen Canvas sections and the reflection question on its own, for reference only). This scoring logic is published so you know what is being assessed -- it is not the same as quiz answer keys.

Evidence of practice

  • A recorded rubric result (PASS, REVISION_REQUIRED, or FAIL) with no critical-fail rule triggered
  • A passing review leaves capability evidence for each of the eight fine-grained AI 301 competencies individually (Single-Agent vs. Multi-Agent Decision, Role & Responsibility Design, Task Allocation & Delegation, Agent Communication Contracts, Shared Context & Coordination, Orchestration Pattern Selection, Conflict Resolution & Evaluation, Multi-Agent Architecture Design)

What you can earn

AI 301 Course Completion Certificate: earned by completing all eight modules and meeting the post-test threshold -- no Canvas submission required. Multi-Agent Architecture Capability Certificate: requires a separately submitted and reviewed Canvas. The two are independent of each other.

What this does not mean

This credential confirms the holder completed AI 301's required course content and post-test, or (for the capability certificate) demonstrated goal framing, role design, task delegation, communication and coordination, and conflict handling on the Multi-Agent Architecture Canvas.

  • It is not a formal professional certification or hiring qualification in multi-agent system design
  • It does not guarantee that any specific multi-agent system the holder designs later is correct or safe
  • It does not replace human review before high-risk decisions (Human, AI, Human)

What next?