Getting started

Set up your Operator

Create your workspace, connect the systems Operator needs and set sensible approval boundaries before you expand.

6 min read

The common mistake is trying to hand over the whole company at once. Setup goes better in the opposite direction: one business motion, only the systems that motion touches, and clear approval boundaries before anything runs.

1. Before you begin

Pick the first business motion you want Operator to carry. A good candidate happens often, follows a recognisable pattern and currently costs somebody time every day: lead response, the Morning Brief, meeting follow-up or inbox triage.

Everything else in this guide follows from that choice. The motion decides which systems you connect, which signals Operator watches and which approvals matter.

2. What you need

  • one repeatable process to start with
  • access to the connected systems that process touches
  • agreement on who reviews sensitive actions

3. Create a profile for the role

A profile is one independent Operator environment: its own conversations, sessions and memory. Create a separate profile for each role or client you want kept apart, so context never bleeds between them.

NEW PROFILE

New profile asks for four things: an optional logo, a customer-facing name, which existing profile to clone the starting context from, and an optional SOUL.md.
  1. 1

    Name the profile

    Use the name you would say out loud, a client, a department or a role. Operator creates the internal ID for you.

  2. 2

    Choose what to clone from

    Cloning copies the selected profile's starting context while keeping the new sessions and memory separate. Operator (default) is a safe starting point.

  3. 3

    Write the SOUL.md

    This is the role description: who the profile is, how it works, what it may do on its own and what it must ask about first. Leave it blank only if you want the cloned default behaviour unchanged.

    • identity and who the role represents
    • working style and tone
    • core responsibilities
    • boundaries, what always needs a human decision
  4. 4

    Create the profile

    The new profile opens with its own chat, inbox, activity and memory.

WATCH: CREATE A PROFILE

The full flow in under a minute: New profile from the Today screen, name the role, choose what to clone from, paste the SOUL.md and create. The new profile opens with its own chat, inbox, activity and memory.

Boundaries written into the SOUL.md are instructions, not enforcement. Anything that must never happen without you also belongs in your approval settings.

4. Connect your systems

Connected apps is where you decide two separate things: the channel you talk to Operator on, and the systems Operator is allowed to read and act in. They are configured independently.

CONNECTED APPS

Settings, then Integrations. The top row picks where you chat with Operator. Everything below it grants Operator access to a system, each one shows its approval policy and current connection status.
  1. 1

    Pick your chat channel

    Browser chat always works. Add Telegram, WhatsApp or Slack if you would rather reach Operator where you already message.

  2. 2

    Connect email, calendar and files first

    Google Workspace covers the systems most first workflows need. Gmail supports multiple inboxes with one primary send account.

  3. 3

    Read the approval policy on each integration

    Reading and drafting usually run automatically, while sending, publishing or changing records requires your approval.

  4. 4

    Add the rest of the process only

    Connect the meeting, task and CRM tools the chosen motion actually touches, nothing more.

INTEGRATIONS

Integrations are grouped by what they do, scheduling, meetings, tasks, CRM. Each row shows whether it is connected and connects in a single step.

Access is scoped per workspace and can be removed at any time. How integrations work

5. Authorise a connection

Connecting a system opens the provider's own sign-in and permission screens. Operator never sees your password, you grant access at the provider, and the provider hands back a token that is scoped to your workspace and can be revoked at any time.

STEP 1, START THE CONNECTION

Press Connect on the integration. Operator opens the provider prompt in a new window and shows Authorization in progress while it waits. If the window was closed or blocked, Continue reopens it and Cancel resets the attempt.

STEP 2, CHOOSE THE ACCOUNT

Pick the account the role should work in. Use the business account, not a personal one, everything Operator reads and sends comes from the account you choose here.

STEP 3, REVIEW THE PERMISSIONS

Google lists every permission being requested. Read them before continuing, this is the access the role will have, and Operator's own approval policy sits on top of it rather than replacing it.
  1. 1

    Press Connect on the integration you want

    The status column changes from Not connected to Opening while the provider window loads.

  2. 2

    Sign in and choose the right account

    One workspace can hold several inboxes, so be deliberate about which account you authorise.

  3. 3

    Review the requested permissions

    Leave a permission unchecked if the workflow does not need it. Some providers will refuse the connection if a required scope is declined, reconnect with it enabled if a workflow later reports missing access.

  4. 4

    Return to Operator and confirm the status

    The row turns Connected once the provider hands the token back.

STEP 4, CONFIRM AND SET THE INBOX

After a successful connection the integration shows Connected. For Google Workspace you also control which inboxes Operator reads: In use decides whether an account is watched at all, and Primary decides which address Operator sends from. Add account links additional inboxes, and Disconnect revokes the access immediately.

If the panel stays on Authorization in progress, the provider window was closed or blocked before consent finished. Cancel the attempt and press Connect again with pop-ups allowed, nothing is connected until the status reads Connected.

The screenshots above show Google Workspace, but every integration follows the same pattern: press Connect, sign in at the provider, review and accept the permissions, then confirm the status reads Connected in Operator. Slack, Microsoft and other providers show their own sign-in and consent screens, but the flow is identical.

6. Setup steps

  1. 1

    Choose the process

    Pick the motion where a delay is most expensive for the business.

    • lead response
    • Morning Brief
    • meeting follow-up
    • inbox triage
  2. 2

    Identify the systems involved

    Connect only what the process actually needs. Lead response usually needs a form intake, CRM, email and calendar; the Morning Brief needs whichever systems hold the overnight activity you would otherwise check by hand.

  3. 3

    Define what Operator should watch

    Decide which signals start the workflow.

    • new messages
    • new leads
    • meeting events
    • CRM changes
    • scheduled checks
  4. 4

    Define what Operator may prepare

    Preparation is safe work: nothing has left the company yet.

    • response drafts
    • CRM updates
    • tasks
    • reminders
    • summaries
  5. 5

    Decide what requires approval

    New workflows should begin conservatively. External communication, financial activity, deletion and significant data changes stay human-controlled.

  6. 6

    Test with real but controlled examples

    Run the process end to end on real cases you are watching, before anyone relies on it.

  7. 7

    Expand once it is reliable

    Add the next system or the next action type only after the first loop has run without correction for a while.

The workflow itself follows Trigger, Reason, Act and Approve. Launch your first workflow

7. What to check before continuing

  • the context Operator gathered is the context you would have gathered
  • the suggested action matches what you would have done
  • approval boundaries hold for anything customer-facing or financial

If a proposed action would be awkward to undo, keep it behind an approval until the loop has proven itself.

8. What happens next

Once the workflow is reliable, expand its responsibility or connect an additional system. Widen one thing at a time so a change stays easy to attribute.

Start with one motion. Prove the loop. Then expand.

Continue learning

Launch your first workflow

Take one repetitive business motion from connected systems to a working Operator workflow using Trigger, Reason, Act and Approve.

6 min read

How integrations work

Scoped access, connected per workflow and revocable at any time.

4 min read

Changing autonomy rules

Loosen the leash one action type at a time, based on evidence.

4 min read