role-tiered-training

Source Metadata

Original cangjie-skill frontmatter from the distillation run:

id: f25
title: "分层AI培训 — 高管/经理/工程师的差异化学习"
book: "AI for Everyone"
author: "Andrew Ng（吴恩达）"
source: p22
tags: [training, change-management, role-based, education, organization]
related_skills:
  - slug: ai-transformation-playbook, relation: depends-on
  - slug: ai-team-building, relation: composes-with

R — Reading (original quote)

"For executives and senior business leaders... most of the content can be covered in about 4 hours of training... department heads... need at least 12 hours... software engineers... plan for at least one hundred hours of training." — p22

I — Interpretation (restate in your own words)

The most common mistake in AI training is "giving everyone the same course." Based on his training experience at Google and Baidu, Andrew Ng proposed a layered training framework:

Level	Duration	Learning objective	Core content
Executives/senior leaders	~4 hours	Make resource allocation decisions	What AI can and cannot do, how to evaluate AI projects, how to formulate an AI strategy
Department/project managers	~12 hours	Manage AI projects	How to set project direction, how to conduct business due diligence, how to manage an AI team, how to set acceptance criteria
Engineers/technical leads	100+ hours	Build AI systems	Machine learning fundamentals, model training and evaluation, data engineering, model deployment and operations

Why layer?

• Different roles have completely different AI knowledge needs
• Teaching executives 100 hours of technical details is a waste, and teaching engineers 4 hours of strategy is not enough
• Layered training maximizes learning efficiency and avoids "not understanding" or "too shallow"

A1 — Past Application (past application cases)

• Google Brain training: In the early days, the Google Brain team designed differentiated courses for employees at different levels. Executives spent half a day understanding AI strategy, while engineers went deep into technical implementation over several weeks. This layered model allowed Google to rapidly expand the reach of its AI capabilities despite a shortage of AI talent.

A2 — Future Trigger (future trigger conditions)

Trigger scenarios:

• The company wants to roll out AI training but does not know what different levels should learn
• Training results are poor — executives think it is too technical, engineers think it is too shallow
• Designing an AI training system
• Evaluating the return on investment of AI training

Language signals:

• "AI training"
• "AI learning for all employees"
• "What different levels should learn"
• "AI training system"
• "AI literacy"
• "What AI to teach executives"

Distinguishing points:

• Does the user needtraining designorspecific course content? This skill provides a layered framework and does not cover specific course content.
• Is itorganizational trainingorindividual learning? This skill is oriented toward organizational scenarios.

E — Execution (execution steps)

1. Identify audience levels: Divide the target audience into three categories: executives, managers, and engineers.

2. Design layered courses：

   • Executive level (~4h): AI capability boundaries, project evaluation methods, AI strategy formulation, common misconceptions
   • Manager level (~12h): project scope definition, business due diligence methods, AI team management, acceptance criteria setting
   • Engineer level (100h+): ML fundamentals, data processing, model training, deployment and operations

3. Choose training methods：

   • Executives: workshops, case discussions, external expert sharing
   • Managers: hands-on projects, cross-department rotations, collaboration with AI teams
   • Engineers: systematic courses, hands-on labs, hands-on projects

4. Set learning objectives: Set clear learning objectives and assessment methods for each level.

5. Continuous iteration: Adjust training content and duration based on feedback.

Completion criteria: Complete the design of a tiered training plan, specifying the training objectives, content outline, duration, and assessment methods for each level.

B — Boundary (usage boundaries)

Not applicable scenarios:

• Individual self-study of AI (no need for tiered design)
• Pure technical teams (no executive/manager levels)
• One-off lectures (not systematic training)

Failure modes:

• Teaching the same course to everyone — executives cannot understand technical details, engineers find the strategy too shallow
• Insufficient training duration — a 4-hour executive training compressed into 1 hour cannot establish basic understanding
• Training only technical staff and neglecting management — management does not understand AI and cannot effectively support AI projects
• No practice opportunities after training — learn it and forget it

Author blind spots:

• Andrew Ng's duration recommendations (4/12/100 hours) are based on experience at large companies.Small and medium-sized enterprises may need to compress— for example, 2 hours for executives, 6 hours for managers, and 50 hours for engineers can also achieve basic results.
• Tiered training assumes you can clearly distinguish "executives/managers/engineers" — but in flat organizations or startups, one person may simultaneously take on multiple roles.

Related skills

• ai-transformation-playbook — depends-on: Tiered training is the concrete implementation plan for the third step of the five-step AI transformation guide, "provide broad AI training," and needs to be understood within the overall transformation roadmap.
• ai-team-building — composes-with: Tiered training and AI team building complement each other — training provides a talent pipeline for the team, and the team provides practical scenarios for training.