Claude AI [NXC]

Course subjects

The New Claude

• Claude as a prediction engine, why it can be confidently wrong, and the fix: a source plus verification

• The five model tiers: Haiku 4.5, Sonnet 5, Opus 5, Fable 5 and Mythos 5, which Anthropic restricts to invitation-only access

• The one heuristic for choosing a model, and the vendor-risk lesson of Fable 5

• The product map: Claude as advisor, as executor and as fabric

• Hands-on: bring one real task from your week and run it through Claude

Getting Great Output

• RCTF prompting: Role, Context, Task and Format, and why context carries most of the quality

• Projects, Memory and Styles: your knowledge, your continuity and your voice

• Artifacts and file creation: build inside Claude, or download finished .xlsx, .pptx, .docx and .pdf files

• Hands-on: rebuild your own task as an RCTF prompt and create a custom Style

Answer to Deliverable

• Inline visuals: charts, diagrams and timelines produced inside the answer

• Quick data analysis with the working shown, so you can check it

• Output compared with action: the trust model of plan, permission and a human in the loop

• Hands-on: nominate a weekly metric for Claude to watch

Claude That Does the Work (Cowork)

• Cowork, Claude’s digital coworker: give it a goal and a folder, and it returns finished work

• The approval gate, human oversight by design

• Running Cowork on your computer compared with running it in the cloud

• Hands-on: delegate one real job, then map five weekly tasks you could hand over

Claude Everywhere

• Claude in Chrome: the browser agent and its safety gate

• Claude for Microsoft 365 in Excel, PowerPoint and Word, and Claude for Outlook

• Browser rules: safe, drafts only, or never

• Hands-on: put Claude inside a real spreadsheet and review every edit it makes

Wiring Claude In

• Connectors and MCP: one-click links to Gmail, Drive, Slack and many more

• Scheduled tasks: moving from reactive answers to proactive reports

• Skills and plugins: packaging your best prompt into a repeatable capability

• Hands-on: spec your first Skill and turn it into a scheduled task

AI Agents and the Build Lab

• Agent design: a goal, plus tools, plus guardrails, written up as your own agent spec

• Vibe coding: building working tools by conversation rather than by writing code

• The frontier: Claude Science and where the direction leads

• The Build Lab: 30 supported minutes to ship an automation, a tool or a live agent

Trust, Governance and Your Playbook

• The three questions every business asks: training data, certifications and safety

• The honest limits of the technology, and how to manage each one

• Your one-page governance brief for IT, legal and the board

• The 30-day playbook: get fluent, make it yours, delegate, make it permanent

The Gauntlet and Certification

• A 16-question finale drawn from all eight modules, with four minutes on the clock

• A certificate of completion, with distinction at 13 out of 16 and all eight badges earned