The Ultimate Beginner’s Tutorial for AI Personal Assistants: Everything You Need to Get Started (Part One)

Recently, seeing the names [Model A] Dot, [Company B] Muse, [Model C] Cue, and [Product D] Bot in the same sentence can be dizzying. They’re all called AI personal assistants, they all claim they can get things done for you, and yet when it comes time to actually use one, picking the right tool and figuring out where to even start can eat up a good chunk of your day.

Maybe you just want to dig up the to-do items you missed in your inbox, turn a pile of meeting notes into something readable, or have an assistant watch a single task for you and report back when something happens. But then you open the product and find some of them put you on a waitlist, others demand an invite code, and still others require you to figure out whether the subscription you already pay for will even cover this feature.

What makes AI personal assistants so exciting is that the small, repetitive chores you have to handle every day can finally be handed off to something that follows through. You state clearly what you want, it goes out and gathers information and organizes it, and whenever a decision needs your input, it comes back to you. Once you’ve found a rhythm, it can even keep picking up where it left off on the same piece of work.

So in this article I want to walk through all four of them properly. Which one should take on the work on your plate, how do you get access, how do you phrase your very first task clearly, and how do you tell whether the results came back right? Even if you’ve never used an agent before, you can follow along and build your first to-do list and comparison table, then gradually fold it into your day-to-day work.

1. Start With the Work You Already Have On Hand

All four can do research, write content, and organize information. When choosing a tool, factor in the accounts you’re already using, where your files live, and what you actually need to finish over the next week.

The first table below is for choosing a starting point. It doesn’t mean the other products can’t handle similar tasks. If you already have a suitable subscription, start with the product you’re paying for. Signing up for four new accounts just to run one small task adds a whole extra round of learning and account management.

Lining up convenience against the cost of switching makes the choice much clearer.

These trade-offs come from how each product actually works: Dot’s work-and-usage model, Muse’s goals-and-activity design, Cue’s assistant roles, and Bot’s shared computer each determine what you’ll have to manage yourself.

2. What a Personal Agent Adds: The Ability to Keep the Work Going

Think of a Personal Agent as an AI assistant you can work with over the long term. You tell it the goal, hand over the material it needs, let it use certain tools, and specify which actions must come back to you for approval. It executes, leaves behind an artifact, you check it, tweak it, and then reuse that same set of requirements later.

Take organizing a dinner party. If you just need restaurant recommendations, a normal chatbot is enough. But once you want it to check your calendar for open time, compare restaurants, prepare the booking details, and hand everything back to you for final confirmation before anything is reserved, the task now involves continuous follow-up and multiple steps.

Some of those steps run in the assistant’s cloud computer, some need to connect to your own apps, and some depend on your local machine. Sorting out which is which matters far more than memorizing model names.

Cloud-based tasks usually keep running after you close the page. Steps that read local files or operate local apps, however, may need your computer to stay on and your client to be online. Dot is a case in point: scheduling a task in the cloud and having it operate your local machine are two different conditions, and Dot’s workflow draws that line explicitly.

One more thing that trips people up. Being able to send a message to your assistant from some chat entry point does not mean it can already read your email, and it does not mean it can already control your computer. The messaging entry point, app connections, and local access all have to be checked separately.

3. From Sign-Up to Claiming Perks — Follow Your Own Page Downward

At the stage of getting access, don’t rush to buy a membership. If you see a Waitlist, you’re still in the queue. If you’ve already paid but can’t find the entry point, the feature may simply not have been opened to your account yet. First confirm whether your account can get in, then confirm whether your existing subscription covers it, and only then decide whether to spend the money. Below are the steps for each of the four. Pick the one you’re using and follow it down.

Dot: Create on the Computer, Then Keep Going on the Phone

• On your computer, open ChatGPT and log in to the account you plan to use long-term. If you don’t have one, create it via the official site, then double-check your current eligibility.

• Personal plans currently list Pro 100, Pro 200, and Pro 500, and you must also meet the official age requirements. Currently supported regions do not include the European Economic Area, the UK, or Switzerland. Business Premium and Enterprise follow their own phased rollout schedules, and Enterprise also needs an admin to enable it. When using a work account, make sure you’ve landed in the right workspace. [Eligibility details]

• Open dots from the desktop web or desktop app and follow the intro to create your agent. You can skip connecting apps for now; the desktop app will also ask whether to connect your local machine. You can get started with just two practice files.

• After creation, change the name and appearance, then log in with the same account on a Dot-supporting mobile app to keep working. Mobile web doesn’t support this path, so complete the first creation on a computer. [Creation steps]

Already on a qualifying plan but still no entry point? Check your account, workspace, desktop entry, and app version. Enterprise users should have an admin verify the switch. If the page still says it’s rolling out gradually, keep your existing account and notification channels — and don’t treat “buying it again” as a way to refresh your eligibility.

Muse: Start on the Web, Complete Account and Age Verification

• Open muse.ai/join and choose Web among the iOS, Android, and web entry points. If you later need the mobile app, come back to the same page to reach the right store.

• Once you’re on the official page, continue with the current sign-in flow using your email or phone number, receiving and entering the verification code yourself. Remember which account you used, and log back in with the same one if you step away midway.

• If you hit age verification, open the verification page offered in the current flow. It may show Confirm age, Link Instagram account, or Link Facebook account — pick the one you can actually complete. If it takes you through a credit card, read the charge and refund wording first; if the page says a temporary US$1 charge will be refunded in 5–7 business days, record it as an age-verification charge, and check the purchase page separately for subscription fees.

• Head back to the site and check your status. If you’re in the main conversation, you can start your first task. If you see Waitlist, keep that account’s application and notification channel and continue with the troubleshooting below. Current region and age conditions follow the official rollout notes and your account page.

New users: once you’re in, check the redeem entry under Settings right away. If you see Redeem invite code, follow the four-step illustrated walkthrough in Section 6 to claim the invite reward — don’t wait until you’ve been trialing for a few days to remember it.

Cue: Get a Valid Invite Code First, Then Complete Access

• Open the official Cue site, or click Agents (Cue!) in the sidebar and scan the QR code to download per the page instructions. If you received an email invite, you can also continue from the entry point given in the official email. [Official setup notes]

• Enter your code wherever the login flow asks for an invite code, confirm the page accepted the invitation, and then step into the assistant interface.

• The official announcement page still lists the limited-use universal code MEETCUE, first come, first served. You can try it at the official entry; if it says invalid or exhausted, ask a friend who’s already set up to share an invite that still has slots. [Official invite notes]

• After setup, keep one assistant and hand it three public web pages to build a comparison table. Early access via invite code is free; later pricing and quota depend on your own page, so there’s no need to buy a Manus plan just to try out the invite code.

Cue’s invite code is used to gain access. Muse’s redeem code is used to claim extra usage once you’re inside the product — different entry points, different problems being solved.

Bot: Put Your Existing Subscription to Use First

• Open the official Bot site, pick your device from the download area, and set it up per the official getting-started instructions.

• Log in with the Cursor account you want to keep long-term. If you already have Cursor Pro, Pro+, Ultra, or a self-serve Teams seat, use that same account; Enterprise users follow their account owner’s access setup.

• If you already have personal SuperGrok, SuperGrok Plus, SuperGrok Heavy, or X Premium+, open the Bot plan page and bind your Grok/X account through the relevant flow. The Link SuperGrok for Grok Bot article in the official help center has the detailed steps. SuperGrok Lite does not include Bot eligibility, and team and enterprise Grok plans can’t be pushed through the personal binding flow.

• Double-check your Cursor account before binding. Once bound, go back to the plan page and look at Weekly usage. A binding can’t be migrated to a different Cursor account on your own afterward, and Cursor and Grok/X Bot quotas don’t stack — subscribing on both sides can still generate two separate bills. [Binding and billing rules]

If you just upgraded a SuperGrok subscription that was already linked, the quota tier may take up to 24 hours to sync. That wait time doesn’t apply to other registration, waitlist, or verification issues. If you don’t have an existing subscription, check your account page for a usage trial first, then decide whether to buy.

Stuck on a Waitlist? Check Which Step You’re Actually Blocked On

The waiting you can skip usually comes down to choosing the right entry point, finishing your verification, and reusing eligibility you already have. Invite codes, subscriptions, and age verification each do their own job — read what the page is actually asking for and you’ll waste far fewer rounds. The early-access mention in Muse’s invite announcement is an eligibility for some new features; registration waitlists and regional rollout are still handled through their own processes.

Muse’s remote-browser route can be used to check how different access exits change what a page looks like. Some community members run a browser overseas and then operate it from their own computer. If you’re willing to maintain one more remote session, you can try it with the steps below.

• Open Browser Use Cloud and log in with your own account. First check your current trial quota and pricing — it’s a standalone service that provides remote browsers.

• Go to Remote Browsers, start a session, and click Launch & open to bring up the remote window. Confirm the exit region the session is showing; the provider’s proxy notes say it uses US residential proxies by default.

• Inside the remote window, open the Muse join page and choose the web entry point. Keep using your own Muse account, and complete the verification codes and current page requirements yourself.

• Follow the page’s outcome. If you get into the conversation, do a small task. If you’re still waitlisted, keep your original application. This route changes the access exit only; account region, age, and rollout eligibility are still judged under Muse’s current rules. Seeing “US” in the exit display does not mean all verification has passed.

• After registration, try continuing on your everyday device with the same account. When you’re done with the remote browser, click Stop browser (or the stop-session control on the current page) and confirm the instance has ended. Simply closing the local tab may leave the remote session running.

Browser Use’s current pricing page offers a one-time $1 credit for qualifying new sign-ups, with browser duration and traffic billed separately. Do small, scoped runs to watch how fast it burns, and don’t keep a remote instance open just waiting for Muse to unlock. Only use sessions you control, and complete verification according to the real situation. If you saved the browser configuration, login state may be retained along with it. [Configuration notes]

Claim What’s Available Now, Then Check the Usage Page

The Bot usage trial has a 7-day window, and it’s also consumed by actual task usage — one big task could burn through it early. Per the official notes, these usage trials don’t auto-convert into paid plans; if what you chose is an App Store subscription trial instead, check the renewal rules on the purchase page separately. Start with one small task to check the results and the consumption. If the entitlements you already have cover your needs, there’s no rush to buy a second subscription.

4. Your First Delegation: Spell Out Six Things

Asking an assistant to “help me be more productive” gives it almost nothing to work with. Swap that for “merge the to-dos from these two minutes into one table, and leave anything without a date as TBD,” and you’ll immediately be able to tell whether the result is correct.

Your first task can use the generic format below. Replace what’s in the brackets and keep the scope small enough that you can check it line by line.

Once you get the result, do one small check first: pick one date, one source, and one uncertain item, and go back to the source material to see whether it was handled correctly. A tidy table can still write “discuss next week” as “complete next week.” For a beginner, simply being able to spot that difference is already the start of learning how to use the assistant.

If the result has problems, stay in the original task and point to the exact spot to fix. For example: “Change the date in row 3 to TBD — the minutes only said to discuss this week; leave the other rows alone.” Redoing the whole document from scratch may end up breaking parts that were already correct, and it burns more usage.

Compare All Four on the Same Piece of Work

To experience the differences between the four, use the same source material for a side-by-side exercise. This exercise includes two fictional minutes of a meeting, a copy-paste task, and a human answer key — no need to hand over your own company files first. Give the two minutes and the same task to whichever assistant you can already use, and keep the human answer key for the acceptance check. If there’s no upload entry point, paste the full text of the minutes and keep the file title and numbering.

This exercise lets you choose based on actual results. Get a correct finished product first, then compare which product is more convenient to connect to your own data and which makes it easier to view progress and stop the work. A single run only speaks to that one task’s performance — it can’t rank every product for every use.

5. Dot: Turning Scattered Material into Work You Can Keep Following Up On

Dot suits people who already have some material, conversations, and projects and want an assistant that keeps following up. Because it can hand work off to ChatGPT Work or Codex, ordinary office work and content prep are in scope too. Coding capabilities only need to be pulled in when writing code is required — you don’t have to learn to code just to use Dot.

Talk on a Call While Dot Opens Pages and Does the Work

When doing content research, you can talk to Dot while telling it which accounts to look at and what to compare, all while watching its cloud computer work through the steps. In the run shown below, Dot had already opened a Xiaohongshu account page, and the conversation kept laying out the requirement to organize topics into Feishu. When it got to the write-in stage, it handed the required authorization back to you; you confirm the target table before it continues.

This call and cloud-computer run show a task still in progress — the Feishu write-in is still waiting on authorization. You can follow the same pattern: have it check several specified pages, organize titles, prices, or plan differences, and add requirements on the spot wherever it gets something wrong.

After content goes out, you can hand Dot a separate task to schedule the review. In one such review setup, a Feishu table already lists five checkpoints — D2, D3, D5, D7, and M1 — each mapped to a due date, with the status all “To be tracked.” Whenever you want to know which content deserves more effort going forward, you’ll have a table to keep checking and filling in.

You can have Dot review web pages while it listens to your requirements, and you can also have it schedule the follow-ups that need checking into a table. For your first pass, pick a few pieces of material — the meeting-minutes exercise below makes it easy to see whether it missed anything.

First, prepare two non-sensitive meeting minutes and keep the meeting date in the file name. You can also just use the fictional practice material above, and you don’t need to connect your whole set of work accounts yet.

The [creation notes] and [control notes] lay out these paths. Pausing work does not undo external actions that have already happened.

Build Your First To-Do Table from Two Sets of Minutes

If the minutes only say “look at the design draft again next week,” the table should not contain a Friday deadline it invented. You can deliberately pick two minutes with overlapping items to check whether it can merge them, and then find one item with no owner to see whether it keeps it as “to be confirmed.”

Once that step is done, tell Dot to fix the output format and sorting preference. If you need daily follow-up, add the execution time, the data scope, and the end date, then go to Scheduled to check the saved arrangement. Mentioning “remind me about this going forward” in the conversation still requires a corresponding scheduled task.

Dot’s long-term value has to be earned through correct context, accumulated over time. When handing it an existing project, send the project brief first — list what’s already been decided, the current status, and the next step — and then have it recite its understanding back to you. Just because it knows your preferences doesn’t mean it has read every historical task.

When things get stuck, troubleshoot by where the work is actually running. If a file can’t be found, first confirm whether the file is on your local machine or already provided to the current task. If a local action doesn’t proceed, check whether your computer and client are online. Once work has been delegated as an independent task, go to Activity, open that task, and see whether it’s waiting on a connection, a login, or an approval.

6. Muse: Start by Digging One To-Do Item Out of Your Inbox and Calendar

Muse is [Company B]’s personal AI assistant, and Muse Spark is the model behind it. It has its own cloud computer, can connect to your everyday apps, keep processing tasks, and come back to you when your approval is needed. It also extends into small-business uses — organizing customer information, preparing content, and analyzing operational data can all start from a concrete task.

Turn Scattered Articles into a PDF You Can Actually Open

You’ve bookmarked plenty of articles, and when you actually need them, you end up digging through them one by one. That kind of organization work can go to Muse: combine the articles into a PDF with text and images laid out properly, so you can just open it and keep reading.

Whether you’re studying a topic, researching several products, or compiling your own writing into a handbook, you can hand it the accessible links and files and explain how to categorize them, what order to read in, and which images to keep.

What you end up with is a file you can keep flipping through. Have it organize three articles first, open the result, and check whether the headings, images, and pagination look right before feeding it the rest. For any links it can’t open, have it list them separately so you can supply the files.

Muse Invite Rewards: Four Steps to 1B Extra Tokens for Both Sides

If you’re a new user who sees the redeem entry, claim it this way — current terms are governed by the redeem page. Get a friend’s currently valid 6-digit invite code, and first confirm your own page is still within the redeemable window. After a successful redemption, both you and the inviter receive 1B Muse tokens each. The entry point is in Settings; pasting the code into chat will not complete the redemption. The four steps below use the desktop app as the example — the web and mobile menus may differ, so look for the same-named entry on your own page.

Step 1: Open the bottom-left menu, then click Settings.

Step 2: Go to General and find Redeem invite code.

The Usage on the same page shows your usage — take a quick look at the existing balance under Additional tokens first so you can verify how much you gained after redeeming. Click Redeem invite code to open the input box.

Step 3: Enter your friend’s 6-digit code, check the terms, then click Confirm.

Both the page and the official instructions require redemption within 48 hours of joining. The six boxes correspond to the six-digit code — enter it exactly as given, and double-check you haven’t mistyped a letter or digit. Click Confirm only when the redeem page states that both sides get 1B. If you see an invalid-code message, verify the code and the remaining redemption count with your friend first; you don’t need to reinstall the app.

Step 4: After you see Congratulations, go back to Usage to confirm the credit.

The success popup states that each side received 1B. Close it, then go to Settings → General → Usage → Additional tokens and compare the balance before and after. The balance includes your original quota, so judge the claim by how much this redemption added.

When you want to share your own invite, check the sharing instructions and remaining uses inside your account first, then give your currently valid code to a friend and explain that “once redeemed, both sides get a reward.” How many tasks this quota can complete depends on task size, how you use it, and how many rounds of rework are needed — it can’t be converted into a fixed number of days.

Connect One Service First, Then Check the First Result

Once you have access, connect only the service you need this time around. Tidy up email → handle your inbox first. Check for schedule conflicts → handle your calendar first. Reading an email, drafting a reply, and sending an email are three different actions, and Muse’s permission design allows read/write scopes to be separated when the service supports it.

Open the currently supported email connection under Settings → Connectors, log in to your own account on the authorization page, and check the read and send scopes before confirming. Go back to the conversation and have it first list three emails it can read, verify that they come from the mailbox you just connected, and only then delegate the weekly to-do task below. If the page doesn’t offer your email provider, you can paste a few email copies to it manually first.

The avatar and Goals design give the background work somewhere to be checked. Browser takeover is for temporarily taking over a web page’s operation; turning off notifications only reduces interruptions — it doesn’t count as stopping the work.

Find the Emails That Need Handling in the Past Week

Spot-check three emails first. The subject lines and links should match, deadlines should come from the email body, and anything already handled shouldn’t be re-listed as a new to-do. If the priority looks wrong, point it out specifically — “this customer confirmation needs a reply today, that event promo can be ignored” — and have it re-order according to your actual plan.

Once the first table passes, pick one email and have it draft a reply. Have it check the recipient, the promises, the attachments, and the send time before executing the send. That way you’ll understand much more clearly what each layer of authorization actually adds.

For example, when building next week’s content plan, you can send: “Use only the features and offers already on the product page; don’t invent customer feedback; give me a theme, audience, image requirement, and draft for each item.” Muse’s business features have already expanded around common business tools, but you still have to tell it the specific product, the timing, and the goal.

If it isn’t progressing, click the avatar to check the activity and approval status. If it’s waiting on an approval, read the action carefully before handling it. If it’s only missing a specific app, check that one connection — there’s no need to reconnect every account. If the notifications are too many, narrow the target and require notification only when something is due, in conflict, or needs a decision.

Privacy, too, should be understood case by case. [Company B] states that Muse conversations and cloud-computer data are not shared directly with ad systems, and that model training can be turned off in Settings; the assistant’s access to third-party websites may still indirectly affect advertising. The current standard service may also be accessed by [Company B] for operational, security, or similar needs, so only hand over the material the task truly requires. If the email authorization scope is too broad, you can practice with a separate mailbox or a file you’ve selected.

At this point, hand one concrete small task to Dot or Muse and see whether the result is usable. Part two continues with the real-world use of Cue and Bot, and then covers permissions, costs, and long-term follow-up in more detail.