Claude Cowork Course —
Stop Prompting. Start Delegating.

Build a Personal AI System That Actually Knows How You Work

I built the self-paced Claude Cowork Online Course for knowledge workers, including product managers, business analysts, consultants, coaches, agile practitioners, writers, and anyone whose work involves recurring judgment tasks, who want to move beyond one-off prompting and build a system that compounds over time.

You do not need to write code. You do not need an engineering background. You do need to understand how Claude Cowork turns a generic AI assistant into one who knows your role, your standards, your anti-patterns, and your judgment calls, and why that difference matters more than any single prompt.

The ceiling of prompting is well-known; you have done so for months. You get decent results on good days and generic filler on bad ones. Every conversation starts from zero because the model forgets who you are and what good output looks like for your specific work.

This course shows you what is above it.

The Claude Cowork Online Course is in English. 🇬🇧

14-Day Money-Back Guarantee: Cancel Anytime, Full Refund. A Risk-Free Career Investment.

14-Day Free Trial:
Cancel Anytime, Full Refund —

A Risk-Free Career Investment

What You Will Learn to Master

1. Build Skills That Encode Your Expertise and Compound Over Time: Move from one-off prompts to persistent instruction sets that capture how you work: your standards, your judgment calls, what “done” looks like. A prompt disappears after one use. A Skill persists across every future session. That is the difference between using AI and building with it.
2. Configure Your Environment So the Model Recognizes You: The instruction stack (User Preferences, Global Instructions, Project Instructions, CLAUDE.md) is the highest-leverage setup activity most users skip entirely. A vanilla Cowork installation guesses. A configured one knows your role, your constraints, and your quality bar.
3. Decide What to Delegate and What to Keep Human: The A3 Framework (Assist, Automate, Avoid) gives you a decision model for every task. "Can AI do this?" is the wrong question. "Should AI do this, and at what level of autonomy?" is the right one. The course teaches you to spot the three failure modes: rationalizing "Avoid" work into "Automate," set-and-forget on tasks that need auditing, and rubber-stamping "Assist" output without reading it. Also, you will create and document the context that makes your agents successful. 
4. Connect Skills to Your Tools and Assemble AI Agents: A Skill tells the model how to work. A connector gives it access to Google Drive, Slack, Gmail, Notion, Jira, or 300-plus other services. Combine the two, add a folder structure, and you have an Agent who handles a category of work autonomously.
5. Iterate Instead of Perfecting: Every Skill is version 1. The course teaches you how to test, break, and improve your Skills, and how to recognize when the next iteration is no longer worth the effort. The result will not match what you would produce manually. For most tasks, the time savings justify the tradeoff. When the gap matters, close it by hand.

Testimonials from the Cowork BootCamp Founding Cohort

[PLACEHOLDER — Insert BootCamp cohort testimonials with LinkedIn links here]

Claude Cowork Course Topics

The Claude Cowork Online Course is built around Vemorum Advisors, a fictional London-based strategy firm with twelve people, four partners, and a weekly deliverable that keeps everything running: the Monday Briefing.
Every Friday, an analyst produces a four-to six-page market briefing for the four partners. Every Monday at 9:00, the partners meet for ninety minutes to act on it. When that briefing is late, weak, or generic, the most senior ritual at the firm starts badly.

You are the analyst. Your job is to build an AI system that helps you produce better Monday Briefings, faster, without losing the judgment that makes them valuable. Along the way, you will configure your environment, build Skills for recurring research and writing tasks, connect Claude to the firm's tools, and assemble Agents who handle the repetitive parts so you can focus on the parts that require your name on the cover page.

The course follows the same arc that a founding cohort tested in the live Cowork BootCamp (April 2026): experience first, then configure, then build.

Course Design: The course philosophy: define outcomes, not steps. Context beats capability. The A3 Framework (Assist, Automate, Avoid) as the delegation spine that runs through every module. Pick your own recurring work to translate the exercises to.

Claude Idiosyncrasies: Three tabs, three different worlds: why Claude forgets when you switch tabs. The "Project" paradox: three definitions of the same word. Words that change meaning across Chat, Cowork, and Code. The extension stack. Cowork safety: what to check before your first real session.

Cowork Set-up: The Instruction Stack: four layers of customization (User Preferences, Global Instructions, Project Instructions, CLAUDE.md) and how they interact without overlapping. The setup checklist. CLAUDE.md principles: read every single time, less is more, WHY-WHAT-HOW, pointers beat copies, define what not to touch, the file compounds with use, strategic not clerical.

Skills: Seven principles for building effective Skills. The prompt-first method: prompt yourself to a successful output in Chat (cheaper), then package it as a Skill. Skill anatomy (YAML frontmatter, description as trigger, instruction body). Writing anti-patterns into Skills to help the model avoid your known failure modes. Four principles for maintaining Skills: test, drift, version, audit. When to run maintenance. Hands-on exercises with the Vemorum Advisors case.

Connectors: Without connectors, every Skill you build works only on what you paste into the conversation. With connectors, Skills reach into the tools where your work already lives and write results back. That is the difference between an AI assistant you feed manually and a system that operates on your actual data. The module covers setup, the risks of giving an AI tool access to live services, a trust framework for deciding how much autonomy each connection deserves, when not to connect at all, and the per-connector limits on each Claude plan.

Compounding Systems: The five-layer stack from single prompts to agentic systems, and why the first month feels slower than prompting: you are building reusable stock, not producing output yet. The payoff comes later and accelerates. Three roles you will move through as your system matures: Operator, Designer, Gardener. Why context is the only advantage that holds when everyone has access to the same models.

A3 Framework: Most teams delegate work to AI on a whim, without criteria or a plan for what to do when the output is wrong. The A3 Framework gives you a decision model: Assist (AI drafts, you decide), Automate (AI executes on rules, you audit), Avoid (stays human because failure would damage trust). The A3 Handoff Canvas documents every delegation decision: who reviews, what triggers a revert, and what good output looks like. Those documented boundaries are what your agent harnesses need to function. The A3 Decision Tree. Exercises across three parts.

Agents: What does "agent" mean inside Claude Cowork? Everyone says agent, but for this course it means something specific. The anatomy of an agent: what you have already built by this point, what not to trust blindly,and when to delegate to Cowork.

AI Agent Risks: An AI agent that can read your files, browse the web, and push data through connectors is an attack surface. The module names ten specific risks and introduces two tests for every workflow: the reversibility test (can you undo what the agent did?) and the distinction between approval and review (clicking "yes" is not the same as reading the output). You will learn where agents fail in ways that are difficult to detect and expensive to fix. When an agent acts on your behalf, the consequences are still yours.

Token Economics: The price of your Claude plan stays the same. What you get for it does not. Cowork multiplies the problem because a single session can consume what ten chat conversations would. The module covers where your tokens are leaking and how to make your consumption visible before it surprises you.

ClaudeOps: A configured system that no one maintains will rot. This module covers the daily habits that prevent that: dictating instead of typing, ending sessions so the next one starts with full context, pruning Skills before they contradict each other. You will learn when to prompt explicitly and when to let Cowork decide on its own, what to check when Anthropic ships a new model version, and how to automate the tasks you are still doing by hand.

Plugins: What a plugin is. Why plugins matter. What is inside a plugin. Where to find plugins. A practical example: the Product Management Plugin. 

The course combines video walkthroughs of key concepts and Skill structures with hands-on exercises. You build working artifacts in every module. You do not watch someone else build them.

What Makes This Course Different

1️⃣ You build, you do not watch. Every module produces a working artifact: a configured environment, a Skill, or an Agent. Theory serves construction.

2️⃣ Designed for the $20/month Pro plan. Every exercise works within the Pro plan budget. The prompt-first method front-loads work in Chat (cheaper) and uses Cowork only for packaging and testing.

3️⃣ No coding required. Skills are Markdown files. If you can write a structured document, you can build a Skill. The course is built for knowledge workers, not developers.

4️⃣ Tested with a live cohort. The course content comes from the Cowork BootCamp (April 2026), where real participants built real Skills and told me what worked and what did not.

What You Will Get

✅ 8+ hours of self-paced video modules, plus guided implementation work.

✅ The A3 Framework Decision Tree and Handoff Canvas as working handouts you can use on day one.

✅ The Vemorum Advisors case package: firm charter, evaluation criteria, team profiles, 12 sample articles, and three reference briefings (weak, medium, strong) so you can calibrate your own output.

✅ Multiple useful Skills to give you a headstart.

✅ Skill Design Principles and Skill Maintenance Principles handouts.

✅ The Claude Set-up Checklist and Instruction Stack reference.

✅ The Quick Reference Card for Claude Cowork (setup, delegation framework, model selection, limitations) and the Terminology Glossary.

✅ Exercise materials with sample inputs and explicit success criteria.

✅ Lifetime access to the version you purchase: all texts, slides, Skills, prompts, graphics; you name it.

✅ Claude Cowork Foundational Certificate upon completion.

Frequently Asked Questions

• Do I need coding experience? No. Skills are Markdown instruction files. If you can write a clear email, you can build a Skill. The course does not use Claude Code or any developer-facing tools.
• What Claude plan do I need? A paid Claude plan. Pro ($20/month) is the minimum. The course is designed for Pro. Max ($100/month or $200/month) gives more room for experimentation.
• Does this work on Windows? Claude Cowork runs on macOS and Windows (x64). On ARM64 Windows devices, Cowork requires remote sessions.
• What AI tools do I need besides Claude? None. The course uses Claude and Claude Cowork exclusively. The delegation framework (A3) is tool-agnostic, but the hands-on exercises require the Claude Desktop app to access Cowork.
• How is this different from an AI prompting course? Prompting courses teach you to write better individual inputs. This course teaches you to build persistent instruction sets (Skills) that encode your expertise across every future session within a system. Prompts disappear after one use. Skills compound.
• How much time do I need? The course has 8+ hours of video plus hands-on exercises. Most students will spend roughly twice the video time when including their own implementation work. The prompt-first method means you start building in normal Claude Chat conversations before you touch Cowork, so there is no steep on-ramp.
• I already took the AI4Agile Online Course. Is this redundant? No. AI4Agile teaches agile practitioners to use AI in their specific workflows. This course teaches any knowledge worker to restructure how they work with AI through Skills, delegation frameworks, and agent assembly. AI4Agile students will recognize some foundational concepts, but the scope and depth are different. AI4Agile students: check your email for your discount code.
• What if Anthropic changes Cowork significantly? Anthropic enhances Cowork frequently. The interface will change; the core methods this course teaches will not: identify recurring work patterns, encode your expertise as reusable instructions, decide what to delegate and what to keep, and iterate based on results. These transfer to any agentic AI tool. When changes are significant enough to warrant a new course version, I will offer existing students a discounted upgrade.
• What is the refund policy? You get a full refund within 14 days of purchase. No questions, no hoops.
• Will the content stay current as Cowork evolves? The course covers stable patterns (Skills, instruction stacks, delegation models) that persist across interface changes. You keep lifetime access to the version you purchased. When Cowork changes enough to warrant a new course version, I will release it as an upgrade with a discount for existing students.