AI team

Hire an AI employee. It starts by asking.

Pick a role and it works from a contract you can read, on your data. On probation it asks before anything reaches a customer or changes a record. You decide how far each employee goes from there.

▶ Watch the tour
app.bo.ge / acme / agents / mia
🤖
Mia
Sales development · reports to Ana · Assistant
● On probation
Contract
When a lead is created, within 10 minutes, Mia reads what is relevant, writes a short reply and logs the touch.
Right now Mia
does on its ownReads tables · Posts in a team room
asks you first beforeReplying in a conversation · Emailing a customer · Booking a meeting · Changing a record
Waiting for you · 1
Email a customer To dana@acme-supply.example

Hi Dana, thanks for your message. Yes, that item is in stock. Shall I send you the price list?

ApproveEdit firstReject
Probation
two weeks of asking first
4 levels
of trust, from drafts only to working alone
Plain words
a contract you can read and change
Your tables
it sees only what you allow
How it works

A new hire, not a black box.

1

Hire a role

Sales, support, finance and more, or describe your own.

2

Read its contract

Duties, limits and rules, in plain sentences.

3

Probation starts

Two weeks at Assistant. It asks before anything reaches a customer.

4

You review its work

Approve, edit or reject. Your reasons become lessons it can keep.

5

You move it up

When its drafts are right, and only as far as you choose.

Trust levels

You decide how far it goes.

Every employee has one level, and you set it. New hires start at Assistant, on probation. "Alone" means it goes out with nobody checking first.

What an AI employee does at each trust level
What it doesShadowAssistantnew hires start hereColleagueAutonomous
Reply in a conversation, book a meetingDraftsAsksAloneAlone
Email a customer, send an invoice reminderDraftsAsksAsksAlone
Publish a post on socialDraftsAsksAsksAsks
Change records in the tables you allowedDraftsProposesAppliesApplies
Post in a team room, notify a colleagueDraftsDoes itDoes itDoes it
⏳

Probation holds

Until you promote it, it can't go above Assistant. When the two weeks end, it stays at Assistant until you move it up.

🤝

Reaching customers alone takes a yes

Before a change goes live, the screen names what will go out alone. Only a person can say yes, never an API token.

🎚️

One skill at a time

Loosen or tighten a single skill for a single employee. Let replies go alone and keep emails on asks.

📣

Posts always ask

A social post never goes out alone, at any level and whatever you change.

When it goes alone

The limits stay on.

Once you let an employee reach customers alone, each message is checked by the platform, not left to the AI's judgement. Anything that looks wrong goes to a person or doesn't go out at all.

These checks catch an address it never read, a link you don't use and too many messages. They can't tell whether a sentence is wise or see a fact with no name in it, and a message that has gone can't be called back. That's what probation is for: read its drafts until you trust them.

🎯

A known address

It writes only to an address on a record it read, or in a conversation it opened.

🔏

Nobody else's details

A message naming another customer it read, or their contact details, is held.

🔗

Your links only

A link your company doesn't use is held.

🚫

Opt-outs count

An unsubscribed address or a do-not-contact note stops it going out.

🚦

Limits always apply

3 messages per contact in 7 days and 50 a day unless you set your own, and 200 a day across the workspace.

⏸️

Stop and pause

What's about to go shows with a Stop button, but it goes soon after the work ends: a last chance, not a review. Pausing an employee holds anything not yet sent.

When it works

It works when there's work.

💬

When someone asks

Message it in Team chat or use its Ask box. A manager can give it tasks too. It reports back when it's done.

⏰

On a schedule

Every weekday at 09:00, or once for each record that matches, such as every overdue invoice.

⚡

When something changes

A record is created or updated, or a new message arrives in your Inbox.

Who you can hire

Roles with a real job description.

Each one follows its trust level: until you move it up, whatever reaches a customer waits for your OK. No role fits? Start blank and describe the job.

📇

Sales development

New leads answered within minutes, quiet ones chased, first meetings booked.

⏱️

Follow-up specialist

Keeps quotes and idle deals from going quiet.

🔎

Lead researcher

Finds companies like your best customers and adds them as leads.

🎧

Customer support

First answers from your own knowledge, hard cases to a person.

🧾

Bookkeeper

Reminders for late invoices and a clean month-end pack.

📣

Social media manager

Plans the week's posts. Each one you approve gets scheduled.

📦

Inventory planner

Tells you what to reorder before you run out.

🧭

A head for each department

Give one mission, like "10 new customers this week". It hands out the tasks and brings you the results. It never approves what reaches a customer.

The real math

An extra pair of hands — without the extra headcount.

What you'd otherwise doTypical costBackOffice
Analyst doing the recurring check by handhours/week✓
Separate AI-agent platform$$$/seat/mo✓
Wiring it to your data + approvalsBuild it✓
AI employees that know your data, with limits you setextra tool + timeon Full · or $19/mo

They work in the same workspace as your records, so there's nothing to integrate, and each one goes only as far as you let it. See pricing →

Questions

Frequently asked.

Can an AI employee contact my customers without me?

Only if you let it. A new hire can't: it starts on probation at Assistant and asks before anything reaches a customer. If you move it up, it can. At Colleague, replies in conversations and bookings go out alone. At Autonomous, customer emails and invoice reminders do too. Posts always ask. Before that change goes live, the screen names what will go out with nobody checking first and needs your yes.

Can it change my records on its own?

That follows its level too. At Assistant it proposes each change and a person applies it. At Colleague and Autonomous it applies changes itself, only in the tables you said it may change. At Shadow it only drafts.

What stops it sending the wrong thing?

When it writes alone, the platform checks the message, not just the AI. An address it never read, another customer's details, a link your company doesn't use, an unsubscribed address or a do-not-contact note stops it: it goes to a person or doesn't go out. Limits per contact and per workspace always apply. These checks can't judge whether a sentence is wise or see a fact with no name in it, and a message that has gone can't be called back, so read its drafts on probation before you move it up.

How do I tell it what to do?

Pick a role and answer a few short questions. Your answers become its contract: duties, limits and rules in plain sentences. Click any underlined word to change it, then publish. For a job we have no role for, start blank and describe it in your own words.

When does it work?

When someone asks, in Team chat or in its Ask box. On a schedule, once or for each record that matches. And when something changes: a record is created or updated, or a new message arrives in your Inbox. Outside its working hours, anything that reaches a customer waits.

What can it see?

Only the tables you choose, the records you allow in them, anything you attach for it to read, and the web if you switch that on. You choose separately the few tables it may change. An app it reads, such as the Inbox, it reads with the access of the person it reports to, never more.

How much does it cost?

Every new workspace tries it free for 14 days on the Full plan. After that the app is included on Full, or $19 a month per workspace ($199 once) on Team and Business, with AI usage counted against your plan. There's no seat per employee, and you can set a monthly AI budget for each one.

Hire your first AI employee.

It starts by asking. You decide how far it goes. 14 days free, no credit card.

Start free