Your work lives in a dozen tools.
Claude ties them together.

Most AI tools are fancy search bars. Claude connects to Slack, Google Drive, GitHub, and more — then builds on what it finds. Write documents. Analyze data. Draft entire presentations. All from one conversation.

The landscape

You've probably used ChatGPT. Here's what's different.

ChatGPT put AI into everyone's hands. That mattered. But it's still mostly a chat window — you paste things in, you get answers back.

Claude, built by Anthropic, takes a different bet. Instead of being the smartest thing in a box, it tries to be the most useful thing connected to your work.

ChatGPT / Gemini

Good at answering questions

Strong general-purpose chat. You bring the context — copy-paste from documents, describe your situation, attach screenshots.

Claude

Good at doing your work

Connects directly to your tools. Reads your Slack threads, browses your Google Docs, then creates new documents, spreadsheets, and presentations inside the conversation.

Key Concepts

Three ideas to know.

Everything Claude does falls into three categories. Here's the mental map before we go deeper.

🔗

Connectors

Live links to apps like Slack, Google Drive, and Notion so Claude can read and act on your real data.

🧠

Memory

Claude remembers your preferences, past conversations, and context so it gets better the more you use it.

Skills

Built-in abilities to create documents, spreadsheets, and presentations — plus partner integrations.

Connectors

Claude talks to your tools.

It's 2pm. A teammate drops a question in Slack about a project plan that lives in Google Drive. Before Claude, you'd open both apps, read the thread, find the doc, then type out an answer.

Now you ask Claude.

It reads the Slack thread, pulls up the Google Doc, cross-references the two, and drafts a response. One conversation, two tools, ten seconds.

Claude · with connectors
Y
Sarah asked in #product-launches about the Q3 timeline. Can you check her Slack message and compare it to our launch plan in Google Drive?
C
I found Sarah's message in #product-launches and your Q3 Launch Plan doc. Sarah's asking if the beta can move up two weeks. Looking at the plan, the eng milestone is set for June 15 — moving beta up would compress QA by 8 days. I'd suggest either scoping down the beta feature set or keeping the original date. Want me to draft a reply to the thread?
G
Google Drive
S
Slack
N
Notion
J
Jira
Z
Zapier
GH
GitHub
A
Asana
+
More
  1. Open Claude settings — click your profile picture in the bottom-left corner, then Settings.
  2. Go to "Connectors" — you'll see a list of available integrations.
  3. Click "Connect" next to the tool you want. You'll be redirected to authorize access.
  4. Grant permissions — Claude only requests read access by default. You control what it can see.
  5. Start a conversation — Claude will automatically search connected tools when relevant. You can also ask directly: "Check my Slack for..."
Skills

Claude doesn't just chat. It creates.

Skills are where Claude goes from assistant to workhorse. Think about the tasks you repeat every week — status updates, pipeline reports, Monday-morning summaries. Skills turn those into one-prompt automations.

Ask once, get a polished document, spreadsheet, or slide deck. Ask the same way next Monday and Claude remembers the format, the audience, and the level of detail you like.

Partner skills extend this further. Figma, Notion, and others plug in so Claude can work directly in the tools you already use.

📄

Documents

Write memos, reports, blog posts. Formatted and ready to share.

📊

Spreadsheets

Excel files with formulas, charts, and structured data.

📥

Presentations

PowerPoint decks with layouts, bullet points, and speaker notes.

🔌

Partner Skills

Figma, Notion, and a growing ecosystem of third-party integrations.

Claude · recurring workflow
Y
It's Monday — time for the weekly team update. Pull this week's closed tickets from Jira, highlights from #wins in Slack, and any blockers flagged in standup notes. Same format as last week.
C
Done — here's your weekly update using last week's format. 14 tickets closed, 3 highlights from #wins (including the payment-flow fix), and 2 open blockers. I kept the same section order and tone.

📄 Weekly Team Update — Feb 3.docx
  1. Built-in skills are always on — document, spreadsheet, and presentation creation work out of the box.
  2. Partner skills appear in the skills menu (sparkle icon in the chat input). Enable them individually.
  3. Just ask naturally — say "make me a spreadsheet of..." or "create a presentation about..." and Claude picks the right format.
  4. Reuse and repeat — Claude remembers the format from last time. Ask for "the same weekly update" and it follows your template.
Memory

It remembers you.

Monday you tell Claude you prefer bullet points over paragraphs. Wednesday you start a new conversation. Claude already knows.

This sounds small. It isn't. The difference between a tool that forgets you every session and one that accumulates context is the difference between a temp worker and a colleague.

Two toggles control it.

Memory Claude remembers preferences and facts you've shared across conversations.
Personalization Claude adapts its tone, format, and level of detail to match how you work.
Claude · using memory
Y
Summarize this article for me.
C
Here are the key points (using bullet format since that's your preference):

• The Fed held rates steady at 5.25-5.50%
• Core inflation dropped to 3.2%, below expectations
• Markets reacted positively — S&P up 1.2%
• Next decision expected in September
  1. You control what's stored — visit Settings > Memory to see everything Claude remembers. Delete individual memories or clear all.
  2. It's not training data — memories are stored for your experience only. They're not used to train Claude's models.
  3. Works per-account — memories are tied to your account. Different accounts have separate memory.
  4. You can correct it — say "Remember that I prefer tables, not bullet points" and Claude updates its memory.
Get started

Six steps. Five minutes.

Everything above sounds useful in theory. Here's how to make it real.

1

Open Claude

Go to claude.ai and sign in.

2

Set up your profile

Click your avatar > Settings > Profile. Tell Claude your name, role, and what kind of work you do. This seeds memory from day one.

3

Turn on memory

Settings > Memory. Flip both toggles on. Claude will start learning your preferences.

4

Connect your tools

Settings > Connectors. Start with Google Drive and Slack — they give Claude the most context about your work.

5

Pick your default model

The model selector is at the top of each conversation. Sonnet is the default — best balance of speed and quality. More on models below.

6

Start with a real task

Don't test with "tell me a joke." Bring a real piece of work: summarize a doc, draft an email, analyze a spreadsheet. That's where you'll feel the difference.

Models

Three gears.

Claude isn't one model. It's a family. Think of them like gears on a bike — you shift based on the terrain.

Fast

Haiku — the quick one

Near-instant responses. Use it for simple questions, rewording text, quick lookups, brainstorming lists. It won't write your thesis, but it'll answer "what's the word for..." before you finish the thought.

Best for: rapid-fire tasks, simple Q&A, when speed matters more than depth.

Balanced

Sonnet — the daily driver

The default for a reason. Sonnet handles most work tasks well: writing, analysis, coding, summarization. Fast enough that it doesn't feel slow, smart enough that it rarely misses the mark.

Best for: everyday work, writing drafts, code generation, document analysis.

Deep

Opus — the deep thinker

For hard problems. Opus takes longer but catches nuances the other models miss — subtle logical errors, complex multi-step reasoning, tasks where getting it right matters more than getting it fast.

Best for: complex analysis, critical writing, nuanced reasoning, when accuracy is non-negotiable.

Projects

Organize repeat work.

Some conversations are one-offs. But when you keep coming back to the same type of task — writing blog posts, reviewing contracts, planning sprints — Projects give Claude persistent context.

A Project is a folder with instructions. Every conversation inside it inherits those instructions automatically. Set it up once, benefit forever.

Blog Writing

Project · 3 custom instructions · 2 reference files

Instructions: "Write in a conversational, direct tone. Keep paragraphs to 2-3 sentences. Use subheadings every 200 words. Target audience: product managers who are technical but not engineers."

Reference files: style-guide.pdf, past-posts-examples.md

Every conversation in this project automatically follows the style guide and matches the voice of previous posts — without re-explaining each time.

  1. Open the sidebar — click the Projects icon (folder) on the left.
  2. Click "New Project" and give it a name.
  3. Add custom instructions — tell Claude how to behave in this project. Be specific: tone, format, audience, constraints.
  4. Upload reference files — style guides, examples, data. Claude uses these as context in every conversation.
  5. Start a conversation inside the project — hit "New chat" from within the project folder. Instructions apply automatically.
Prompting

Five ways to get better answers.

The gap between a mediocre prompt and a great one is enormous. These aren't tricks — they're habits that compound.

01

Let Claude prompt itself

Start with your goal, then ask: "What information do you need from me to do this well?" Claude will ask clarifying questions that produce better results than your original prompt would have.

02

Show, don't describe

Instead of saying "write it casually," paste an example of casual writing and say "match this tone." Examples beat adjectives every time.

03

Iterate, don't restart

If the first response is 70% right, say what's wrong and ask for a revision. Claude improves within a conversation faster than starting fresh.

04

Give it a role

"You're an experienced copywriter reviewing a junior's work." Roles activate different patterns of knowledge. Be specific about the expertise level.

05

Drag in files

Claude handles PDFs, images, spreadsheets, and code files. Dragging in the actual document beats describing it every time. Upload first, then ask.

Going further

This is just the desktop app.

Once you've got a rhythm with Claude, there's more to explore. Cowork lets Claude work alongside you in the background — researching, drafting, preparing while you focus on something else. Claude Code is a terminal-based tool for developers that can write, test, and ship code.