Claude 101: teaching notes

"There are twelve ideas in this course, one per screen. For each one I'll show you the thing working, then stop and ask where it fits in your work. You don't need to remember it all today. What matters is that you see each one actually work."

"This is the plan for the session, in four parts. First, what Claude is and how to work with it. Then, how to write an instruction that gets a good result. Then, how to organise the work so you stop repeating yourself. And last, how to connect it to the tools you already use. If you get lost at any point, we come back to this screen."

"You may have used a chatbot before: you ask a question, you get an answer, and that's the end of it. Claude works best when you treat it differently. It knows a great deal in general, but it knows nothing about your work: your clients, your standards, your way of doing things. You supply that part. And when the first answer isn't right, you don't start over; you tell it what's wrong and it corrects, the same way you'd work something out with a colleague." (Pairs with framings.md: Claude as translator/teacher, and the "brief it like a junior" line.)

"A good instruction has three parts. The context: who you are and what this work is for. The task: what exactly you want done. The rules: the tone, the format, and if you can, an example of what good looks like. Most people write one short line and expect it to guess the rest. It can't read your mind. The more you tell it, the better the result. A useful comparison: brief it the way you'd brief a capable new colleague on their first day." (framings.md: brief like a junior; the six-year-old version for informal rooms. Never use both in one session.)

"When you get a bad answer, it's usually not because the tool can't do the job. It's because something you know never made it into the instruction. Too generic? It didn't know your role or your audience. Wrong format? Show it an example instead of describing what you want. Confident but wrong? That happens. Ask it to cite sources, and check the facts that matter. Spotting what's missing and adding it is the main skill in working with AI. It has a name, AI Fluency, and there's a whole Praxis session on it."

"The app gives you the same Claude in three modes, and you pick by the size of the job. Chat is for quick questions and short drafts. Cowork is for the bigger jobs: it works across several files or sources for a stretch and comes back with a finished piece of work. Code is for building software. Nothing to memorise here: when a job feels too big for a chat window, there's a mode built for it."

"Right now, every new chat starts from zero, so you end up re-uploading the same files and re-explaining the same rules. A project fixes that. You put the files and the rules in once, and every conversation you start inside the project already knows them. The practical effect: you stop repeating yourself, and the work stops getting lost when a chat ends." (Previews framings.md: the folder is the memory; the context window. In Claude Code that 'place' becomes a folder on your machine.)

"When it produces something substantial (a document, a table, a working page), it opens it in its own window next to the conversation, properly formatted, where you can read it, edit it, download it and share it. If it ever answers inside the chat when you wanted something usable, say 'make that an artifact' and it will."

"A skill is a set of instructions Claude keeps and applies whenever that task comes up. The built-in ones you've already used. That's how it makes spreadsheets and documents. The valuable ones are the ones you make yourself: you describe how you do a repeatable task (the steps, the rules, the format), and from then on it does that task your way, every time, without being re-told. You build it by talking to it; there's no code involved. For a firm, this is the part that matters: the way your most experienced person does the job can be written down once and used by everyone." (framings.md: skill file as institutional memory; adapt "senior person" to their role.)

"Connectors let Claude work with the tools you already use (Drive, Gmail, Slack and the rest) through one standard. Instead of copy-pasting things in, it reads the actual source. On the question everyone asks: it can only see what your own account can see. Connecting your email gives it your inbox, nobody else's. You choose what to connect, and you can disconnect any of it at any time."

"Enterprise Search lets you ask one question across everything your organisation has (the documents, the chat, the email) and get one answer, with citations back to the sources. It only returns what you already have permission to see; an administrator sets it up, and nothing changes about who can access what."

"The Research function is for questions that need real digging. Instead of one search, it runs many: it makes a plan, searches, reads what it finds, decides what to look up next, and writes the result up with a citation for every claim. That takes five to fifteen minutes, not seconds. Use it when a good answer needs twenty sources rather than one."

"We've seen a lot of features today. You don't need most of them yet. What you need is one task, a real one from your actual week, to start with. Let's pick yours now." (Then go to their task from intake, not the grid on the slide.)

"Everything we did today was in the Claude app, but the same Claude is available inside other tools: in Slack, in Excel and PowerPoint, in the browser, and in a version called Claude Code that works directly with the files and folders on your computer. So the question to leave with isn't whether to switch tools. It's which of the tools you already spend your day in would benefit most from having Claude inside it." (Hold the one term: it's Claude Code, every time.)

"To sum up the session: give it proper context and a clear instruction, the way you would with a capable colleague, and when the answer isn't right, add what's missing and go again. That's the core of it. Projects, skills and connectors exist so you can do that with less repetition and in more places."