A personal AI assistant is an AI you set up once for your own work: it remembers what you tell it, connects to the apps you use, and can do jobs on a schedule without being asked each time. The difference from an ordinary chatbot is that it keeps your context between conversations and can act for you, within limits you set.
What is a personal AI assistant?
A personal assistant has standing instructions ("keep answers short", "I manage the mobile team"), a memory of facts about you, and access to some of your tools. You can give it a recurring job, such as a summary of new issues every morning, and it does that job on its own and tells you how it went.
Many assistants are built on the same large language models as chat apps. The personal part is everything around the model: what it may remember, which accounts it may use, and when it is allowed to act.
How is a personal AI assistant different from a chatbot?
A chatbot answers the message in front of it and starts most conversations from nothing, apart from any memory feature you turn on. An assistant also works when you aren't typing: it can start a job on a schedule and take actions in other apps. Acting is the part to be careful with. An assistant that can change things should ask before it does anything you can't easily undo.
What can a personal AI assistant do?
- Remember context. Your role, your projects and your preferences, so you don't repeat them in every chat.
- Run routines. A morning summary, a weekly check on a project, a reminder to follow up on something.
- Work with your apps. Read and update issues, notes, tasks and bookings in the services you connect.
- Talk by voice. Take a spoken request, answer it, and leave a written record.
- Use your computer. Open an app or a web page and complete a small task while you watch.
- Work with your files. Read a document you share, draft a new one, or build a chart from a spreadsheet.
- Reach you on your phone. Reply on your phone as well as on your computer, and notify you when a job is done.
How do you choose a personal AI assistant?
Start from what you want it to do. The voice assistant built into your phone or speaker is quick for timers, short messages and smart-home controls. A chat app suits writing, research and one-off questions. An assistant that keeps your context and acts in other apps suits recurring work, such as a weekly check on a project.
Then compare the practical details:
- Your apps. It should connect to the services you use every day. Check the list before you sign up.
- Your devices. If you want it on your phone and your computer, check that it keeps the same memory and conversations on both.
- Cost. Check which features need a paid plan, and how voice calls and long jobs are counted.
Is it safe to give an AI assistant access to your accounts?
That depends on the assistant and on how much access you give it. Any assistant can misread a request, so look for one that lets you control each step. Before you connect an account, check four things:
- Can you limit which apps and tools the assistant may use?
- Does it ask before it creates, edits, deletes, sends or buys something?
- Can you see and delete what it remembers about you?
- Can you watch it work and stop it at once?
A good way to start is with one small, low-risk job, such as a daily summary that only reads. Give the assistant more access as you see how it works.
How do you set up a personal AI assistant in AI Integrator?
In AI Integrator, a personal assistant is called an Agent. You can create one on the web, in the desktop app or on iPhone, and the same Agent works on all three. Introducing Agents walks through each part.
- Your Agent. Instructions and a Home chat
- Memory. Facts you choose to keep
- Routines. Runs on a schedule
- Connected apps. GitHub, Linear, Notion and more
- Calls. Talk by voice
- Your computer. Only where you allow it
- Code chats. Starts coding work
- Create it. Choose New agent and write its name and what it does in a sentence or two. Chat drafts its instructions from that, and you can edit them (up to 8,000 characters).
- Choose what it remembers. Each Agent follows your account's memory settings, can only read saved memories, or has memory off. Every saved fact can be read, edited or deleted, as Memory in AI Integrator Chat explains.
- Choose what it can use. Skills and connectors are switches on the Agent's profile. Connectors include GitHub, Linear, Notion, Todoist, Cal.com and Granola. There are no Gmail or Outlook connectors. If a service publishes its own connection for AI assistants (an MCP server address), you can add it under Skills & Plugins. By default an Agent asks in its Home chat before a plugin creates, edits or deletes anything.
- Give it routines. Add under Routines opens its Home chat with a draft that starts "Set up a routine to". Each run posts its result in that chat.
The Agent's Home is one continuing conversation that reads like a text thread. When it needs a sign-in or an answer from you, it shows a card there and its row says Waiting on you.
Calls
You can call an Agent from its Home and talk to it by voice. You can set how it sounds and how fast it speaks. When the call ends, the transcript appears in Home. During a call, the Agent can start work on your computer or a Code chat.
Using your computer
An Agent can use the apps on one of your computers only after you turn on Let Agents use this computer for that computer. It is off by default. The computer asks before the Agent uses an app, unless you chose Always allow for that app. Purchases always ask, and a sign with Pause and Stop stays on screen while the Agent works. Watch what it does and stop it if it goes wrong. Agents can now use your computer has the full rules. Computer use is built and tested on Windows. The Mac version has not been tested on a Mac yet.
An Agent can also start and message Code chats on a computer where Let Chat and Agents use Code chats on this computer is on, which is off by default. Each chat it starts asks you before every edit and command.
What does it cost?
You can create Agents and give them routines on every plan. Their replies use your plan's usage like any Chat message. Calls, notifications and the Files library need Pro or Max. A call is charged by the minute, so it can't run past the usage you have left. Pricing compares the plans.


