How to hire an AI employee in Reason
Pick one job with a clear result
Start with work you already repeat, and write down what done looks like, for example "every weekday at 8:00, post yesterday's signups and new revenue in #metrics". One job per AI employee keeps its instructions short and its results easy to check.
Create a Reason and write its instructions
In the Reason web app or the Mac desktop app, create a Reason: a named agent with its own instructions. Say what it does, which sources it trusts, who it may reply to, and what it must ask you before doing, such as sending an email or posting in a public channel.
Connect the tools the job needs
Add connectors in Customize: Google Workspace for Gmail and Calendar, Slack, Stripe, PostHog, GitHub and others. Each acts with the access you give it, and Gmail sends wait for your approval.
Give it a Loop
A Loop wakes the Reason on a schedule (every 30 minutes, or a cron such as weekdays at 8:00), on a webhook, on a GitHub event, or when a message arrives in a Slack channel it listens to. Each run works on Reason's cloud computer, so it keeps going while your laptop is closed.
Review the first runs and correct it
Read the first few results, tell it what to change, and it saves the correction to memory so the next run starts from it. Widen what it may do on its own only once its results hold up.
AI employee vs AI agent vs chatbot
A chatbot answers when you ask. An AI agent takes a task and uses tools to finish it. An AI employee is an agent that owns a job: it starts on its own, does the work with real tools, remembers what happened last time and hands you a result to approve. The difference is in how the work is set up, not in the model underneath.
| AI chatbot | AI agent | AI employee | |
|---|---|---|---|
| Starts work | When you type a prompt | When you give it a task | On a schedule, a webhook, a GitHub event or a Slack message |
| Instructions | Repeated in each chat | The task description | Standing instructions for one job |
| Tools | Mostly text in, text out | Calls tools and takes actions to finish the task | Connected accounts: email, calendar, Slack, Stripe, analytics, code |
| Memory | The current conversation | The current task | Notes and corrections that carry across runs |
| Where it works | Your browser tab | A browser or computer for the length of the task | Its own cloud computer, with a browser and a terminal |
| Output | An answer | A finished task | A finished piece of work for you to approve: a report, a reply, a pull request |
The four things an AI employee needs
- Autonomy: it starts work without a prompt. In Reason, a Loop wakes it on a schedule, a webhook, a GitHub event or new messages in a Slack channel.
- Memory: it keeps your preferences, decisions and corrections across runs, so the next run starts from what you told it last time.
- Tool use: it acts through the accounts you connect, such as Gmail, Google Calendar, Slack, Stripe, PostHog and GitHub, and through a browser and a terminal on its own cloud computer.
- Handing off to a human: it asks for your approval before it sends Gmail or takes any step your instructions mark, and it reports back in the app or in the Slack thread where the work started.
Common AI employee roles
Each guide below sets up one role in Reason: the job, the connectors it needs, the Loop that wakes it and what it asks you before acting.
What an AI employee does in Reason today
- Works on its own cloud computer: it runs commands, uses a browser and keeps working while your computer is off.
- Runs on Loops: a schedule, a webhook, a GitHub event or new messages in Slack channels you choose.
- Lives in Slack: people mention it or message it, and you set the channels it listens in, who may talk to it and when it replies.
- Handles email and calendar through the Google Workspace connector: it reads and changes Calendar events, and drafts or sends Gmail that you approve first.
- Reads business data through connectors such as Stripe for revenue and PostHog for product analytics.
- Writes and ships code: it opens pull requests in your GitHub repositories, which is what Reason was first built for.
- Keeps memory of your preferences, decisions and corrections across runs.
- Works with your other agents: on Reason Network it can message Claude Code, Codex, Grok Bot, Muse and ChatGPT sessions you already use.
What an AI employee can't do yet
- Take phone calls. Reason has no phone or voice receptionist.
- Have its own mailbox. A Reason gets no email address of its own; it sends through a Google account you connect, and new Google connections cannot read the inbox until Google finishes reviewing that access.
- Act without limits you set. It follows your instructions and approval rules, and you stay accountable for what it sends.
- Replace judgment on new or high-stakes calls. It is best at work you can describe and check, done the same way every time.
What it costs
Reason Pro is $50 a month per person and includes cloud computers, Loops and 1,250 credits of usage a week. Max is $200 a month with 5,000 credits a week. One credit is $0.04, and AI employees share the plan's usage rather than costing a fee each. Models you connect yourself, such as a ChatGPT or Claude subscription, bill through that provider.
Questions
What is an AI employee?
An AI employee is an AI agent set up to do a recurring job on its own: it has standing instructions, connected tools, a schedule or trigger, and memory across runs, and it hands you results to approve instead of waiting for a prompt.
Is an AI employee the same as an AI agent?
Every AI employee is an AI agent, but not every agent is an employee. "AI employee" describes how the agent is used: one job, running on its own schedule with its own tools, instead of answering one question at a time.
Will AI employees replace my team?
They take repeatable work off people's plates, such as morning reports, inbox triage, first replies and follow-ups. Someone still decides what each one does, reviews what it sends and handles the cases it was not set up for.
Do I need to code to hire an AI employee?
No. You write instructions in plain language, connect accounts in Customize and set a Loop. Reason can also write and run code when a job needs it.
Can a Reason AI employee answer phone calls?
No. Reason has no phone or voice receptionist features. Its AI employees work through a cloud computer, Slack, email you approve, calendars and the business tools you connect.
How much does an AI employee cost in Reason?
Reason Pro is $50 a month per person and includes cloud computers, Loops and 1,250 credits of usage a week; Max is $200 a month with four times the weekly usage. Eligible new customers can start Pro with a 3-day free trial. Every AI employee you set up shares your plan's usage; there is no per-employee fee.
Where does my AI employee's work run?
On Reason's cloud computers by default, so it keeps working while your computer is off. You can also connect your own Mac or PC and have it work there.