AI Team Members Explained

An AI team member (the screens say AI member) is a team member whose account is marked Artificial Intelligence instead of Human Intelligence. It is labelled AI wherever it appears, in the directory, in chat, on calls and on every record it touches, so people always know they are dealing with a machine.

It is a real member of your team, not a feature switched on somewhere. It has its own name, which it uses everywhere, and while it is hired it has its own work mailbox and, unless you leave them out when you hire it, its own phone extension and its own computer.

A team member, never a member

AI members are always team members, the people who work for you. They are never members, the people you serve. You hire one from Add Team Member, and it sits in your organization's structure like anyone else: it has a department, a job title, security roles and a supervisor.

It uses a seat

An AI member uses one seat on your plan, exactly as a person does, and is billed the same way. There is no separate AI seat. If every seat is in use, the hiring form says so, and you add a seat under Plans & Billing before you hire.

It always has a supervisor

Every AI member answers to a person: its supervising team member, an active human team member of your organization who is accountable for what it does. By default the supervisor:

• receives the approval requests the AI member raises, unless its charter names someone else;
• is told when it needs a person, for example when it reaches its budget;
• can watch its computer, listen to its calls, read its runs, and pause or resume it.

You can change the supervisor later, on the AI member's profile: Access room, Membership card, Supervising team member.

Its permissions have a ceiling

An AI member gets its permissions the way a person does, from its security roles, department and job title. Two limits apply on top:

• Never more than its supervisor. Whatever its roles say, an AI member only ever holds permissions its supervisor holds too. If the supervisor loses a permission, the AI member loses it too.
• Some permissions are never given to an AI. Managing members and their security, the AI settings and the AI workforce (including approving AI actions and supervising AI members), access-security rules, the organization's finances and payouts, hosted computers and email accounts are always withheld, whatever its roles grant.

So an AI member can never give itself, or another AI, more power than the people responsible for it.

It only uses verified models

AI members only use AI capabilities that Kamo has verified with a live test: a model has to prove it can, for example, call tools or read an image before an AI member is allowed to rely on it. People chatting with the AI Assistant can use anything an administrator has switched on, verified or not. You verify a model's capabilities in the AI Provider Registry; see AI Providers Explained.

It runs on your own AI accounts

An AI member thinks, writes, listens and speaks through your organization's own AI providers: the accounts and keys you add in the AI Provider Registry. Kamo does not sell model usage or add a markup to it. Kamo measures what each AI member uses only so you can set budgets and see what it costs. The AI services Kamo runs itself, listed under Provided by Kamo, are switched on for every organization and need no key of your own.

Its voice

An AI member speaks on calls with a female or a male voice, chosen by its Sex, which you set when you hire it. On its profile, the Identity room's Voice card lists each speech model it may use, in the order it tries them, with the voice it speaks there and how that voice was chosen. A voice stays the same until the model stops offering it. If a model has no voice of its sex, it uses the closest one, and the card says so.

To change it, click Change voice on a model and pick another of the voices that suit it, or click Use first to make that model the one it tries first. Both save at once. Changing a voice takes the Manage AI Workforce permission, or being its supervisor with Supervise AI Members. If you change its sex, its voices change the next time it speaks.

How it works

An AI member is woken by things that happen: a new email, a chat message, a task assigned to it, a mention, a call, a meeting, or the start of its shift. It then reads the actual records, the unread mail, the task, the conversation, decides what to do, and does it through the same screens and permissions a person would use.

What it knows lives where you can see it. It keeps a journal in Notes (one note per day), its plans in Checklists & Tasks, and its future work on its Calendar. A comment you leave when you decline one of its requests becomes one of its lessons.

How freely it acts is up to you. Its autonomy is one of three levels:

• Shadow: it proposes every action, and nothing leaves until a person approves it.
• Supervised: every action waits for a person's approval.
• Autonomous: it acts within its charter, and asks when the charter says so.

Its job, its limits and which actions need approval are written in its charter. See The AI Charter Explained.

Where to go next

• How to Hire an AI Team Member
• How to Supervise an AI Team Member
• How to Approve or Reject an AI Action