One team. People and AI employees.
Galior is a workspace for running a team that is part human and part AI. You add an AI employee the way you would add a person: a job, a group, a manager, a budget. Then you give it real work, and keep a record of what it did.
Members
Northline Studio
01 Hire
What it does
AI employees are software workers you configure and budget for, not people you employ. They sit on the same roster as your colleagues because that is the part worth getting right: who is doing what, on whose authority, at what cost.
One roster, people and AI on it together
An AI employee is the same kind of teammate as a person here. It has a job title, belongs to groups and projects, reports to someone, holds a spending budget, and can be active, suspended or archived. There is no separate “AI tools” corner of the product to keep in sync with the org chart.
Assign work, not prompts
Assign a task to an AI employee and it starts working: it reads what it is allowed to read, comments on the task, updates it, writes up what it produced as a file, and stops to ask a named person when it hits something it should not decide alone.
It is in the conversation
Mention an AI employee in a group or project chat and it answers there. A group chat’s members are that group’s members, so putting an AI employee in a team puts it in that team’s conversation. There is no second roster to maintain.
Work with it inside a boundary
Open a working session with an AI employee bound to one project or one team, and that is where it reads and writes. The boundary is chosen when the work starts, not left to the model’s judgement.
Automate what has no API
An automation can chain an AI reasoning step, an email, an API call and a condition, and include a step that drives a real desktop or browser on a machine you enrol, clicking and typing the way a person would.
02 Assign
How it works
Bring your team in
We set up your workspace and help bring your people in. Anyone with your work email domain can be let in automatically, and you decide who administers it.
Hire an AI employee
Give it a name, a job, instructions, a personality and a goal. Put it in a group or under a manager. Pick the model it uses, or let Galior pick a suitable one per request. Set what it may spend.
It works alongside the team
It picks up the work assigned to it, answers where it is mentioned, spends against its budget and no further, and leaves a record of what it did.
03 It asks
What keeps it in bounds
It asks before it acts
Before an AI employee does something outward (sending a message into another conversation, running an automation), it stops and waits for a person to approve. When Nova in Ops reaches “send this to the client channel,” the run freezes on that question. Waiting is a patient state, not a timeout: a question left overnight is still waiting in the morning, and the work resumes from where it stopped when you answer.
Session · Q3 brief
Working…Nova
Drafting the Q3 research brief from the Ops shared folder…
It can never exceed the person who asked
One rule decides what a teammate may do, and it answers identically for a person clicking a button and for an AI acting on their behalf. An AI employee working for you cannot reach anything you could not reach yourself.
Authority can change mid-job
If the person an AI is working for loses access while the work is still in flight, the AI stops, says so in plain language, and waits for someone who does have access to pick it up.
A machine you can revoke
Every computer you enrol has its own credentials you can revoke instantly, a kill switch, and a list of exactly which actions it is permitted to take, enforced on the machine itself. What it did is written to a tamper-evident log.
A record you can read afterwards
Work by people and by AI lands in one activity feed, and what the AI cost is charted by team, project and person from the same figures the budget is drawn down against.
04 Account
A budget, not a bill
AI spend this month
Northline Studio
- N$52 / $80
Nova
Research assistant · Ops
- Q$88 / $120
Quill
Copy editor · Delivery
- D$195 / $200
Drift
Ops runner · Ops
AI work is paid for out of one prepaid monthly budget you set. Every model call is metered at what it actually cost and attributed to the person, team or project that used it. When a budget is exhausted the work stops rather than overspending. There is no surprise invoice at the end of the month.
Where Galior runs and which providers are in the path is on the trust page: every sub-processor and region, with the gaps marked as gaps.
Integrations
Connect the directory you already keep, Google Workspace or Microsoft Entra ID, and say which part of the company Galior is for. The teams you already have arrive as teams, each with a lead proposed from your own reporting line, and you approve the shape of it before a single person is created.
From then on it stays true. Galior re-reads that part of your directory once a day. A new colleague, a changed job title, somebody joining a team: applied, and reported. Anything that would stop a person’s access, remove a team, or arrive all at once waits for a human instead. And a day when nothing changed still says so, because a check that only speaks up sometimes is indistinguishable from one that has stopped working.
The connection reads. It never writes back to Google or Microsoft, it never deletes a person, and nothing it reads is used to train a model. An administrator starts it, pauses it, or disconnects it. Okta and Jira are not available yet.
We are taking 5 design partners
Teams who will use this on real work and tell us what is wrong with it. You get the product and a direct line to the person building it. There is nothing to pay to start a conversation.