Handing an AI agent the keys to your business is a reasonable thing to be careful about. This article says plainly what your agent can see, what it can change, and how to check any of it yourself.
The short version: it sees only what you connect, it acts only where you have allowed it, and everything it does is on the record.
What your agent can see
Three things, and nothing else.
What is already public about you
From the moment you run your free audit, before you connect anything: your Google listing, your public reviews, your photos, your website, and the businesses Google puts next to you. This is what any customer could look up.
What you connect
Each connector is a door you open deliberately, through the provider's own sign-in screen. Your Google Business Profile, your calendar, your inbox, your customer system, your Google Ads, your Beside phone line.
Beside never sees your passwords. You approve the access on the provider's side, and you can withdraw it at any time from Connectors. 📖 Connect your tools to Beside
What you tell it
Your business details under Settings → Business info, and anything you put in your Library - notes, price lists, PDFs, voice memos, photos. Your agent reads these before it writes anything.
Where your contacts come from
Your agent builds its own contact list out of the conversations it already has: the calls and texts on your Beside line, your Gmail, and your calendar. You can also connect Google Contacts.
It never reads the address book on your phone. So a customer whose number lives only on your iPhone is someone your agent has never met. It asks you for the number rather than guessing, which is why it sometimes asks about a customer you are sure you have saved.
✅ Disconnect anything, at any time: Withdraw a connector from Connectors and your agent stops reading that tool immediately, along with any work that depended on it.
What your agent can do
Your agent works in two steps that are deliberately kept apart: deciding and doing.
Watching your business, working out what is worth doing, and putting a plan in front of you changes nothing and costs you nothing. Something only happens to your business at the moment work is carried out - and that step is gated.
The two gates
Gate | What it controls |
Per tool | Whether your agent may act with a given tool at all. The first time it asks, you answer Allow or Reject. You can change it later in Autonomy to Always allow, Always ask or Deny, separately for each one, and you can keep some on Always ask indefinitely. |
Per plan | Whether a specific piece of work goes ahead. Your agent shows you what it noticed, what it would change, and how it will know the job is done. You choose Approve once or Reject, and where it has proposed a particular value you can revise that first. |
It spends credits, never money
Your agent cannot bill you. When your credits run out it pauses new work rather than charging you for more - buying credits or changing your plan are decisions only you make.
What your agent cannot do
Some limits are not settings you can turn on. They are things your agent genuinely cannot reach, and it is quicker to know them up front than to discover them mid-job.
What people ask for | What actually happens |
Your agent cannot read or reply to WhatsApp. Its messaging goes out as a text from your Beside line. | |
Posting to Facebook or Instagram | Your agent writes the post, you publish it. Google Ads is the one advertising channel it runs itself. |
Images and artwork | Your agent does not create or edit images. It writes the words, and tells you which of your own photos will work best with them. |
Sending email | Your agent drafts the email either way, but it can only send once Gmail or Microsoft 365 is connected. |
Anything on your phone | Your address book, your camera roll and your personal messages are out of reach. Beside works from what you connect, not from the device in your hand. |
💡 When your agent says it cannot confirm something, believe it:
It is built to say so rather than hand you an answer it cannot stand behind.
That is usually a sign it is missing a connection or a detail, so telling it the missing piece is the fastest way forward.
How to check what it did
Everything your agent finishes is listed under Activity in Beside's profile (select Beside's name at the top of Chat), grouped by day.
Open an item to see what was done, where the result was Saved in, when it was Completed, and a link straight to the result, for example the review reply it published or the page it changed.
How to stop it
One recurring job: open it under Routines in Beside's profile (select Beside's name at the top of Chat) and select Pause routine, or ask Beside to pause it in the chat or by text. Everything else keeps running.
One tool: set it back to Always ask in Autonomy, or disconnect it entirely from Connectors.
A specific plan: choose Reject in Review rather than approving it.
Your data
How Beside stores, protects and hosts your data is the same for your agent as for the rest of Beside:
FAQ
Will my agent contact my customers without me knowing?
Will my agent contact my customers without me knowing?
Only where you have chosen Always allow for that kind of work. On Always ask, it checks with you every time. Either way, anything that was sent appears under Activity in Beside's profile with what went out and when, so nothing reaches a customer without leaving a trace you can read.
Can my agent handle my WhatsApp messages?
Can my agent handle my WhatsApp messages?
No. Your agent works on calls and texts through your Beside line, on email once Gmail or Microsoft 365 is connected, and on the tools you connect under Connectors. WhatsApp is not one of them.
Why does my agent ask for a number I already have saved?
Why does my agent ask for a number I already have saved?
Because it cannot see your phone's address book. It knows the people it has met through your Beside line, your Gmail and your calendar. Send it the number in the chat and it will use it, or connect Google Contacts so it has the whole list. 📖 Connect your tools to Beside
Is anything read-only?
Is anything read-only?
Yes. Your free audit reads only what is already public and changes nothing. And a newly connected tool starts read-only: your agent studies it and asks separately before it is allowed to act with it.
Can I undo something my agent did?
Can I undo something my agent did?
It depends where the change landed - a review reply lives on Google, a listing change on your Google Business Profile. The item under Activity links straight to what changed, so you can go and correct it at the source. If you are not sure, the Support chat in the app can help.
Who on my team can see all this?
Who on my team can see all this?
Your agent's work, its connectors and its activity belong to the business, not to one person - anyone with access to that business in Beside sees the same thing.