Associate Foundations: working with Claude
The practical course for non-engineers: prompting that reproduces, documents at scale, accuracy review and team conventions.
If you use Claude in your job but have never had structured training, this is the gap. It concentrates on repeatable workflow rather than prompt tricks, and on the review discipline that keeps you accountable for the output.
Course outline
12 lessons · 3h 40m · v1
01Getting productive with ClaudeThe working habits that separate people who get value from people who get frustrated.3 lessons
- What Claude is good and bad atFree preview18 min
- Your first reusable promptFree preview16 min
- Iterating instead of restarting18 min
02Working with documents and dataLarge inputs, long outputs and staying in control of both.3 lessons
- Summarising long documents without losing the point22 min
- Extracting structured data reliably20 min
- What not to paste in16 min
03Accuracy and reviewTreating AI output as a draft to be checked, with a system for checking it.3 lessons
- Spotting a confident wrong answer20 min
- A review checklist you will actually use18 min
- Being accountable for AI-assisted work16 min
04Team workflows and governanceShared conventions so quality does not depend on one enthusiastic person.3 lessons
- Building a shared prompt library20 min
- Simple guardrails for a whole team18 min
- Measuring whether it is helping18 min
What this covers
You will be able to
- Write prompts that produce consistent, reusable results
- Summarise and extract from long documents safely
- Spot and verify confident but wrong answers
- Set sensible team conventions
Included
- 12 lessons across 4 modules
- Prompt library templates
- Review checklists
- Lifetime access and updates
Preview: What Claude is good and bad at
The first two lessons of the course are free to read, so you can judge the material before buying.
A useful mental model of Claude is narrower than the marketing and more useful than 'it's an AI'. It is a text predictor that follows instructions well, reasons over what you give it, and has no access to facts it was not given or trained on.
What it does well
- Transforming text you supply: summarising, rewriting, restructuring, translating.
- Extracting structured data from unstructured documents.
- Drafting first versions where a human will edit.
- Explaining and tutoring over material in the prompt.
- Classifying and routing short inputs against criteria you define.
Where it is weak
- Facts it was not given: specific numbers, dates, citations and identifiers.
- Anything after its training cut-off, unless you provide current data.
- Arithmetic and precise counting over long inputs.
- Knowing what it does not know, unless you give it a way to say so.
- Consistency across repeated runs without structure and validation.
The practical consequence
Treat the model as a capable colleague with no memory of your context and a tendency to fill gaps confidently. Your job is to supply the context and verify the output. Everything in this course builds on that division of labour.
Checkpoint
Name a task in your own work where the model's weakness list makes it unsuitable without a human check.