Getting Started with Claude

You are already halfway there

If you have ever pasted a messy email into an AI assistant and asked it to sound more polite, you have used most of what people think AI is for. This course is about the other half: setting something up once so you stop doing the same fiddly work every single week.

We are going to do that with Claude. Lesson one is the necessary groundwork: getting you comfortable with what Claude is, where it sits next to the other assistants, and what a genuinely useful first request looks like.

What Claude is

Claude is an AI assistant built by Anthropic. You talk to it in a chat window, you can hand it documents to read, and it writes back. It is good at the things every assistant of this kind is good at: drafting, rewriting, summarising, turning a mess into something with headings.

Where it tends to earn its keep for office work is longer material. Claude handles a big pile of text reasonably well, so pasting in a whole meeting transcript, a long policy document or six months of client notes is routine rather than heroic. It also has a fairly plain, unfussy writing voice out of the box, which means less time deleting exclamation marks.

Compared to the others, honestly

This is not a sales pitch, so here is the fair version. ChatGPT, Gemini, Copilot and Claude are more alike than any of their marketing suggests. All of them will draft your email competently. If your organisation has already standardised on one, use that one, because a tool your IT department blesses beats a slightly better tool you have to smuggle in.

The differences show up at the edges. Gemini sits close to Google's documents and mail. Copilot sits close to Microsoft's. Claude and ChatGPT are more standalone, which makes them easier to use across whatever mixture of tools you actually have. Claude's particular strengths for this course are that it keeps reusable context tidily and that it can run work on a schedule, which is exactly what lessons two and three are about.

Your first genuinely useful request

Most first attempts fail the same way: too vague, so something bland comes back, and you conclude the tool is overrated. The fix is to say who it is for, what you want to happen, and how long it should be.

Here is a first request worth running today. Pick a document you have to read anyway and make Claude do the reading.

Read the document below and give me three things.

THE POINT: what this is actually asking of me, in two sentences.
WHAT I MUST DO: any action or deadline that involves me, one line each.
If there are none, say so plainly.
WHAT I SHOULD QUESTION: anything vague, missing or worth pushing back on.
Do not summarise section by section. If something is unclear, say it is unclear
rather than guessing.
DOCUMENT:
[paste it]

The second and third things to try

Below are my raw notes. Turn them into a clear structure, not prose.
Group related points under headings you choose.
Put anything that was agreed under DECISIONS.
Put anything needing a person to act under ACTIONS, with the owner named
or written as UNASSIGNED.
Do not add any information that is not in my notes.
MY NOTES:
[paste them]

Below are three things I have written. Read them and describe my writing voice
back to me in five bullet points: sentence length, formality, how I open,
how I close, and words I clearly avoid.
Then rewrite the fourth piece of text in that voice.
MY WRITING:
[paste three samples]
REWRITE THIS:
[paste the new text]

That last one matters more than it looks. Getting your voice described once and keeping the description is the first small step towards never re-explaining yourself, which is the whole subject of the next lesson.