A Word Before We Begin
This guide assumes nothing. Not that you've used AI before, not that you're comfortable with computers, not that you know what half these words mean. That's the point of it.
Claude is the AI assistant we use across Instant Insight AI BUILD. Think of it less like a search engine and more like a genuinely capable colleague who happens to live inside your computer — one you can hand real work to, in plain English, and trust to do it properly.
You don't need to be "good with computers" to use it well. You need about twenty minutes, a Mac, and a willingness to click things and see what happens. Nothing here is fragile. You cannot break your computer by exploring this guide.
A Mac. An internet connection. An email address (or a Google account) to sign in with. And a paid Claude plan — Pro is the one we recommend. If you already have Pro or Max, you're set.
Setting up your account and choosing your plan is genuinely the first piece of AI BUILD homework — done before we go anywhere near the programme itself. Claude's free plan is for conversation: it doesn't let Claude carry out tasks on your files and apps, and it doesn't run scheduled tasks — and that is what almost all of AI BUILD runs on. There's no version of this working for you on the free plan.
Setting up your account, step by step
-
1Create your Claude account first (see "Get the App" and "Signing In" below) Use whichever email you'd like this tied to long-term — this becomes your AI BUILD working account, so pick accordingly rather than a throwaway address.
-
2Click your name (bottom-left) → Upgrade plan It's in the small menu that pops up from your name. (You'll also find your plan under Settings → Billing.)
-
3Choose Pro and complete payment This is a self-funded subscription you take out in your own name, currently $20 (USD) a month billed monthly, or a little less a month if you pay for the year. Prices vary by region, so check the page for the exact figure. You're welcome to cancel any time; there's no lock-in.
This is genuinely a "get it done in your own time, before we begin" step, not something we walk through together live. If this guide has landed in your inbox ahead of a session, treat it as pre-work — account created, paid plan active, ready to go — the same way you'd turn up to a workshop with the right materials already in your bag.
A few words we'll use a lot
Getting the App
Before anything else, let's get Claude properly installed on your Mac. It only takes a couple of minutes, and once it's done, you'll never have to think about it again.
-
1Open your browser and go to claude.com/download Click into the address bar at the top of the browser window, type
claude.com/download, and press Return. This takes you to Claude's official download page. -
2Click "Download" for macOS The page should recognise you're on a Mac and offer the right version automatically. Click the download button — the file will land quietly in your Downloads folder.
-
3Open the downloaded file Click the little download arrow near the top of your browser window and select the file, or open your Downloads folder and double-click it.
-
4Drag Claude into Applications A small window will open showing the Claude icon next to a picture of your Applications folder. Click and drag the Claude icon onto the Applications folder to install it — this is the standard way every Mac app gets installed.
-
5Open Claude from Applications Find Claude in your Applications folder (or search for it with Launchpad) and double-click it to open.
-
6If your Mac asks "Are you sure?" — that's normal The first time you open any app downloaded from the internet, macOS double-checks with you. Click Open to confirm. You'll only ever see this once, for this app.
You can sign in through claude.ai in a browser too, and most of this guide looks the same there. But only the app can work with the files and folders on your Mac — and it needs to stay open while Claude is working on them. So the app is where AI BUILD happens. It also behaves like any other program on your Mac: it lives in your Dock, opens with one click, and doesn't disappear when you close a browser tab. The rest of this guide assumes you're using it.
Occasionally, instead of the simple "Are you sure?" prompt, macOS refuses to open it altogether — "Claude can't be opened because Apple cannot check it for malicious software." That's not a real problem, just Mac being extra cautious about anything not downloaded from the App Store. Go to the Apple menu → System Settings → Privacy & Security, scroll down, and you'll see a line mentioning Claude was blocked, with an Open Anyway button beside it. Click that, confirm once more, and it'll open normally from then on.
Signing In
This is the only part of the whole process you'll only ever do once. After today, Claude will simply remember you.
-
1Open your browser Click on Safari or Chrome in the row of icons at the bottom of your screen (this row is called the Dock). If you're not sure which icon it is, hover your cursor over each one without clicking — the name will pop up.
-
2Go to claude.ai Click into the address bar at the very top of the browser window (it's the long empty strip, usually showing a little padlock), type
claude.ai, and press Return on your keyboard. -
3Choose how you'd like to sign in You'll be offered a couple of options — usually Continue with Google (fastest, if you already have a Google or Gmail account) or signing in with your email address. Either is fine; pick whichever account you'd like this to be tied to.
-
4If you use email, check your inbox Claude will send you a short verification code, or a "click to confirm" link. Open that email and follow it — this is simply confirming the address is really yours. You'll never need to set or remember a password for this.
-
5You're in You'll land on Claude's home screen — a simple page with a message box in the middle, waiting for you to type something. That's it. That's the whole sign-in process.
There's no way to do this "wrong." If you land somewhere unexpected, close the browser tab and start again from Step 2. Nothing you do here is permanent or risky.
Getting your bearings
Before we go any further, a quick orientation to the strip running down the left-hand side — you'll use this constantly, so it's worth thirty seconds now.
The left-hand sidebar — this is your home base. Your recent conversations are listed underneath it.
Beside the Claude name, you'll see a speech bubble and a </>. The speech bubble is where you'll spend all your time. The </> is Claude Code, a tool for software developers, so you can leave it alone.
Meet the Models
Once you're in, you'll notice a small button just under the message box, showing a model name and an effort level — something like "Opus 5.5 Medium." Click it, and you'll see this.
Click the model name next to the message box to see this menu. New models ship from time to time, so the names on your screen may be newer than this picture — the shape of the menu stays the same.
A "model" is simply which version of Claude you're talking to. They're not different products — they're the same assistant, tuned for different kinds of effort, a bit like choosing between a quick chat over coffee and sitting down properly with someone for an hour.
You do not need to think hard about this. Here's the plain-English version of each:
Leave it on whichever model Claude opens on and get on with your day. If you find yourself running into your limits, switch to Sonnet 5.5 for the simpler work, because it makes your allowance go further. You can switch models mid-conversation at any time, and you lose nothing by starting simple.
What about "Effort"?
In that same menu, underneath the model list, you'll spot a setting called Effort. This is worth understanding properly — it matters more than the model choice for most people.
Hover over "Effort" in the model menu to see the five levels.
Where the model is who you're talking to, Effort is how hard they think before answering. Claude explains it plainly right there in the menu: "Higher effort means more thorough responses, but takes longer and uses your limits faster."
There are five levels — Low, Medium, High, Extra, and Max — running from quickest-and-lightest through to slowest-and-most-thorough. Medium is marked "Recommended", and it's a sensible one to leave alone. Notice the warning beside Max: it uses four times or more of your allowance.
Turn it down (Low) for quick, simple asks, when you want the answer fast and the task doesn't need deep thought. Turn it up (High or Extra) for something genuinely hard or high-stakes, where you'd rather wait a little longer for Claude to think it through properly. Keep Max for the rare occasion that truly warrants it. Most days, you'll never need to touch this.
Free vs Pro — and how "usage" actually works
Behind every message you send and every reply you get back, your account draws on a usage allowance. You'll never see a raw number counting down — Claude tracks it for you — but it's worth understanding what moves it, because it explains why two messages that look similar can use very different amounts.
Four things make usage move faster: a heavier model (Opus 5.5 or Fable 5.1 use more than Sonnet 5.5 or Haiku 5.5 for the same question), a higher Effort level (Extra and Max think longer, which costs more than Low or Medium), longer conversations and bigger tasks (Claude re-reads the whole thread each time, and every step of a multi-step task uses some allowance), and large file attachments. None of this is something you need to calculate. Just know that "Sonnet 5.5, Medium effort, a clear question" is the light, efficient combination, and "Opus 5.5, Max effort, a huge attachment" is the heavy one.
Your session allowance tops back up on a rolling five-hour clock — if you hit a limit, it's temporary; check back in a few hours and you'll have room again. On Pro, there's also a weekly allowance that resets on a fixed day each week, tied to when your plan started. You can check exactly where you stand any time under Settings → Usage.
Ask for what you need in one well-formed message rather than five small back-and-forth ones — Claude re-reads the whole conversation each time, so fewer, fuller messages are more efficient. And if you're working from documents you use often, keep them inside a Project, so they're always to hand without uploading them again. For anything multi-step, group related work into one task rather than several.
One Message Box, Two Ways to Use It
If you've seen older guides, or older screenshots of Claude, you may remember two buttons above the message box: Chat and Cowork. They've gone. There's now one message box, and Claude works out from what you ask whether you want an answer or a piece of work done.
One message box. Underneath it: Project on the left; the model, effort level and permission mode on the right.
Here's the simplest way to tell them apart. Ask "what should I say to a client who's gone quiet?" and Claude will write you a suggested reply, which you then copy somewhere and send. Hand it the work instead ("look through my inbox, find the clients who've gone quiet, and draft replies") and, once your email is connected, the drafts sit in your email, waiting for your nod.
You don't need to choose a mode first. Same box, same habit: say plainly what you want.
Manual or Auto — how often Claude checks with you
Next to the model name sits a small word: Manual or Auto. This is your permission mode, and it's worth setting deliberately on day one.
Click the word beside the model name to choose.
Switch to Manual. You'll see exactly what Claude is about to do, every time, and that is how trust gets built. Move to Auto once you know how Claude works with your files and tools. Whichever you choose, Claude asks for your explicit permission before it permanently deletes a file.
Try these first
The blank message box is, honestly, the hardest part for most beginners — not because it's technical, but because it's a blank page. You don't need a special formula. Here are three you can copy exactly as written, to get a feel for it:
-
💬To warm up:
Explain what you can help me with, in plain English, using examples from a small business. -
💬Something genuinely useful:
I need to write a short, friendly email to a client who hasn't replied in two weeks. Draft three versions with different tones — warm, brief, and slightly more formal. -
🗂️Your first piece of real work — we'll do this together next:
Look through my Downloads folder and tell me what's in there — just a summary, please leave everything exactly as it is.
Try It, Live
Before we tour Settings, let's hand Claude one real piece of work, on something so low-stakes it doesn't matter if anything goes sideways. This is where you'll see an approval prompt for the first time, and understand exactly what it protects you from.
-
1Open the Claude app on your Mac The app, not a browser tab: Claude reaches the files on your Mac only through the app, and the app needs to stay open while it works on them.
-
2Check you're on Manual Look at the word beside the model name, under the message box. If it says Auto, click it and choose Manual.
-
3Type the Downloads prompt from above, and send it
Look through my Downloads folder and tell me what's in there — just a summary, please leave everything exactly as it is. -
4Claude will ask permission to see that folder This is Claude checking with you before it looks at anything on your actual computer. It names the exact folder it wants to see, and waits for your answer. Nothing is opened until you click Allow.
-
5Click Allow, and watch what happens Claude will look through the folder and come back with a plain-English summary of what's sitting in there. That's the whole loop — ask, approve, done.
Click the + in the bottom-left of the message box, then Devices, and choose your Mac under "Run tasks on." It appears as Claude Desktop (macOS). Then send the prompt again.
+ → Devices: the computer Claude works on when a task needs your files.
Claude can only read and write in the folders you've allowed. When it wants to reach something new, it stops and asks first, by name. It never quietly does something you haven't seen coming. Keep that in mind, and handing Claude real work stops feeling like a leap of faith.
Finding Your Way Around Settings
You won't need to touch most of Settings for a while — but it's worth knowing where it lives, and what each part is for, so nothing feels hidden.
Getting there
-
1Find your name in the bottom-left corner On the Claude home screen, look at the very bottom-left of the sidebar. You'll see your name, with your initial in a small circle and your plan beside it.
-
2Click it, then click "Settings" A small menu will pop up. Settings is the first option, at the top. Usage, just below it, shows where you stand against your allowance.
The menu that pops up from your name.
This is what opens — everything you can adjust, listed down the left-hand side.
You'll see a fair number of options here. Most of them you'll never need to touch. The four worth knowing about — because they're the ones that make Claude genuinely useful for your work — are Account, Connectors, Plugins, and Skills. The last three sit together under Customize, the same place the Customize link in your sidebar takes you.
Account — the basics about you
This is where your name lives, along with what Claude should call you, what best describes your work, and a box called Instructions for Claude. Use that box to tell Claude a little about yourself and how you like to work, so it doesn't need reminding every time. Nothing here is complicated; think of it as filling in a short form once. (Light or dark screen, fonts and text size are under General.)
Connectors — linking your other tools
The Discover tab shows the most popular connectors. A tick means connected; the + adds one. "Yours" lists the ones you've connected.
A Connector is simply a bridge between Claude and one of the tools you already use — your email, your calendar, Google Drive, Notion. Connecting one means Claude can genuinely read from and write to that tool when you ask it to, rather than you having to copy and paste things back and forth. You don't need to connect anything on day one. Add a connector only when there's a real task waiting for it.
Plugins — ready-made bundles
A plugin is a bundle — a set of tools and know-how grouped around one kind of work. "Try" tests one in a conversation; + adds it.
Think of a Plugin as a themed toolkit — everything Claude needs to be good at one particular area of work, packaged together. You won't need to build these yourself or go hunting for them; we set them up with you in the programme, matched to what Instant Insight AI BUILD is helping you build. When it's time, Step 11 of your AI Team Setup Guide walks you through it.
Skills — the individual instructions inside
If a Plugin is the toolkit, a Skill is one tool inside it — a specific, well-written set of instructions for doing one particular job well. Skills sits directly above Connectors in the same list, laid out the same way, with Yours and Discover tabs. Once a skill is added, you can call on it from the message box: type / and pick it from the list. Again, this is something we'll usually set up with you rather than something you need to hunt for.
Get comfortable typing to Claude first — that's the only prerequisite. Connectors, Plugins and Skills earn their keep once you know what you actually want help with.
Your First Scheduled Task
This is where Claude stops being something you have to remember to use, and starts quietly working for you in the background. Let's set up one small, genuinely useful example: a Monday-morning round-up of what changed in AI that week, written for a business like yours.
Scheduled tasks run on Claude's side, in the cloud, so they keep running when your Mac is asleep or the app is closed. That's the whole point of them. It's also why your first one should be something that doesn't need the files on your computer. This one only reads the public web and reports back to you.
Where to find Scheduled Tasks
-
1Look at the left-hand sidebar On the Claude home screen, you'll see
Scheduledlisted underneathArtifacts. Click it. -
2Have a look at the Templates tab You'll find a page of ready-made ideas: Daily briefing, Inbox triage, Meeting prep, Weekly review, Content ideas and Monitor a topic. They're there to give you a sense of what's possible; you don't need to use any of them.
The ready-made ideas on the Templates tab. Your own tasks appear under "Yours."
Creating your own
-
3Click "New task," top right A small menu appears with two choices.
"Create with Claude" talks you through it in a conversation. "Set up manually" shows you the plain form — we'll use that here, so you can see exactly what each part does.
-
4Click "Set up manually" A form opens with a Name, a large box for your instructions (with the permission mode and model underneath it), a Trigger for when it runs, and Advanced settings, which you can leave closed.
-
5Give it a name Under Name, type something plain, like
My Monday AI round-up. -
6Write your instructions in plain English In the big box, tell Claude exactly what you'd tell a helpful assistant. For example:
Look for the most useful developments in AI from the past week for a small service business. Give me three, in plain English, with one line on what each could mean for my business. Keep it under 300 words.
Plain, direct instructions — exactly as you'd explain the task to a person. Notice the form starts on Auto.
-
7Switch Auto to Manual The form opens on Auto, and the note underneath says what that means: Claude will work and use your connectors without pausing for approval. Click Auto and choose Manual, so Claude checks with you before it uses a new tool or opens a new site — the right setting while you're still getting used to this. Leave Skip all approvals well alone.
Choose Manual, the top option.
-
8Choose how often it should run Beside Trigger, click the dropdown that says Daily and choose Weekly. Then pick Monday and a time before you start work. One small tip from the form itself: tasks set for exactly on the hour can start several minutes late, so choose a time a few minutes earlier, such as 07:52.
Hourly, Daily, Weekdays, Weekly or Monthly — pick whatever rhythm actually suits the task.
-
9Check it over, then click "Create" Have a last read of your instructions, then click Create in the bottom-right corner. That's it — it's set, and it'll run on the rhythm you chose from here on.
A finished, ready-to-create task: weekly, Monday morning, Manual.
It needs nothing connected and touches none of your files. It reads the public web and writes you a short note. That makes it the perfect place to watch Claude work on a schedule for the first time, and to build a little trust before you hand over anything that matters more. When you're ready, the same form will run a task across your connected email, calendar or documents.
On Manual, Claude checks with you before it uses a new tool or opens a new site. If something doesn't look right, saying no is all it takes. You can pause, edit or delete a task any time from the Scheduled page.
How you'll actually know it's run
Claude doesn't need you to sit and watch. Each run appears as its own conversation in your sidebar list, marked with a small clock, and a blue dot shows when there's something new for you to read. Click in, and you'll find exactly what Claude found, laid out the same way as any other conversation.
There's no need to check constantly. Once a week, glance at Scheduled the way you'd glance at a physical in-tray — nothing gets lost in the meantime.
What's Next
You now know more than most people ever bother to learn about Claude before they start using it properly. That's genuinely enough to begin.
| What you've learned | Where it lives |
|---|---|
| Choosing your plan | Your name (bottom left) → Upgrade plan (or Settings → Billing) |
| Signing in | claude.ai, via Google or email — no password to remember |
| Choosing a model and effort | The model name under the message box — Medium effort is recommended; Sonnet 5.5 makes your allowance go further |
| Asking or handing over work | One message box — just say plainly what you want |
| Manual or Auto | The word beside the model name — Manual while you're learning |
| Approving Claude's first request | Claude names the exact folder or account it wants to reach, and waits for your Allow |
| Letting Claude reach your Mac | The Claude app, kept open — or + → Devices → Claude Desktop (macOS) |
| Settings, Connectors, Plugins, Skills | Your name (bottom left) → Settings, or Customize in the sidebar |
| Scheduled tasks, and checking their results | The sidebar, underneath Artifacts — watch for the blue dot |
Reach out the moment a screen doesn't match what's described here, or something isn't behaving the way you expect — the same way you would about anything else in AI BUILD. That's exactly what we're here for, and it's very rarely you: screens change, and a quick message sorts it out fast.
Twenty minutes today. A little time back every week from here on.