Hire AI
employees.
Let them earn the keys.
One AI employee per job, in your Slack and email. Each starts by asking before it acts, and takes on more only as its track record earns it.
Onboard them like a new hire. Ramp them like one too.
Day one, every employee asks before it acts. Each action it gets right without an edit builds its record. When the record is good, it offers to handle that one action alone: a reply, an invoice, a ticket. A manager signs off, can start with a two-week trial, and takes it back in a click. Your support employee can reply on its own while your finance employee still asks first.
Locked down from day one. Trusted in stages.
Access control comes first: an employee sees only the knowledge and uses only the tools its manager grants. Freedom comes after, one action at a time, each step tied to clear goals: numbers like edit rate, and judgment calls like tone, rated by the people it works with. The manager decides every step and can take it back in a click.
- Every action logged, immutably, exportable.
- Your Slack history is never mirrored.
- Employees always disclose they're AI.
One assistant with every key is one prompt away from all of them. Hire a team of specialists instead.
Each employee earns its own level and holds only its own tools, knowledge and channels. The HR employee can see payroll; the support employee never can. A bad week pulls back one of them, not everything you have automated.
Eight roles, ready on day one. Meet the team you can hire today.
Four of eight. Meet the whole team, or build your own from scratch: an employee is a role, knowledge, tools, and permissions, not a workflow diagram.
Cited answer in #sales, corrected once, remembered forever.
Meet PageIt doesn't wait to be asked. It brings you the work.
Each morning an employee looks at what moved in its corner of the company: what is waiting, what slipped, what someone promised. Then it proposes the next piece of work, already drafted. You say yes or not now, and anything it does still runs at its level.
- A check-in only when there is something worth saying.
- It comes back when it said it would, in the same place.
- Every suggestion has a one-tap useful or not needed, so it learns what is worth your time.
Remy · Support
Morning check-in · direct message
Morning. Three things I can take on today:
Acme opened 4 tickets about SSO this week.
I drafted one help article that answers all four.
12 saved replies still quote a 14 day refund window.
It changed to 30 days on Tuesday, per the decision log.
You told Dana you'd confirm the SLA by Friday.
It's Thursday. I can draft the note from the contract.
Correct them once. It sticks, and you can see who taught them what.
Anyone on your team can correct an employee, right in Slack. Corrections become memories with a name and a date. Feedback rolls up into proposals you approve like a code review, and anything wrong is one click from reverted. No black-box learning: a ledger.
New hires spend weeks learning the company. Yours already know it.
One company memory with a source on every fact: your docs, decisions, policies and every correction. Each employee reads it within its role, so your fifth hire onboards in minutes.
Work crosses jobs. They hand it to each other, and you decide who may ask whom.
When a request belongs to another employee's job, it is passed on instead of guessed at. Remy hands a billing question to Theo; Theo answers from the books.
- No grant, no handoff. You choose which employees may delegate to which.
- Each handoff has a level, like any tool: ask first, or hand off and log it.
- They pass a summary and links, never your raw message history.
Job descriptions take a week. Hiring here takes a sentence.
Say what the job is, in your own words. Pagerox drafts the employee: a name, what it does, the tools it will want and what it should know. You read one review, change anything, and hire. Every tool starts at ask first.
Plenty of work has no API. Each one gets its own computer.
It drives its own browser while you watch live. Logins sit in a vault it never reads, and using one always asks a person first. When the docs run out, it reads your codebase in a sandbox your keys never enter.
Invoices
October 2026| Invoice | Amount | Status | |||
|---|---|---|---|---|---|
| INV-2291 | Sep 20 | Oct 20 | $4,180.00 | Open | |
| INV-2304 | Sep 28 | Oct 28 | $912.40 | Open | |
| INV-2310 | Oct 2 | Nov 2 | $2,650.00 | Open | |
| INV-2275 | Sep 4 | Oct 4 | $1,204.90 | Paid | |
| INV-2262 | Aug 29 | Sep 29 | $3,318.00 | Paid | |
| INV-2248 | Aug 21 | Sep 21 | $640.00 | Paid | |
| INV-2231 | Aug 12 | Sep 12 | $2,075.50 | Paid |
Ask anything. It only shares what you're allowed to see.
Managers decide what each employee can see and which tools it can use. An engineer can ask the People agent about holidays all day; salaries aren't in what it can read, so no phrasing gets them out of it. The limit is set in code, not asked of the model.
No per-seat math, no credits. One plan, whole team.
$20 a month for up to 10 agents, with $15 of model usage included across all of them. Beyond the allowance, overage bills at metered cost plus 40%, and you can cap it yourself, per agent and workspace-wide.
Workspace
the only plan$20/ month, per workspace, not per seat
- Employeesup to 10
- Every role, tool, and integrationincluded
- Frontier models, every turnno surcharge
- Model usage, each month$15 included
- Beyond the allowancecost + 40%
- The ceilingyou set it
Hire your first employee today
One plan, $20 a month for the whole team. It asks before it acts.