The IT team's people changes and security housekeeping, prepared. A joiner gets an access plan from your role matrix (which systems, which groups) and the accounts are created behind approval. A leaver gets a removal plan across every connected system, executed behind approval with the evidence recorded. Devices out of compliance get a note to the person and their manager. New security findings become tickets with an owner. Unused seats and upcoming renewals get a reclaim or renew note every Monday. Quarterly access reviews are packaged per system for the reviewer. Works with your HR system, Okta or JumpCloud, 1Password or Bitwarden, Intune or Jamf, Drata or Lacework, and Jira Service Management or Freshservice. No agent ever reads or resets a password or grants an admin role. Comes with Kit, an IT Operations Coordinator, and a staff portal.
Installs in about 35 min, verified with a smoke test.
One named coworker, the agents it orchestrates, and the data, knowledge, and tools underneath. This is the real manifest, not a mockup.
The employee routes each request to the right worker. Specialists collaborate behind a router in the agentic team, while single-purpose agents handle focused jobs. Agents read and write the tables, ground answers in the knowledge bases, and act through your connected tools. Every write is gated and traced.
Every agent's tasks, every table's schema, every trigger. Read it all before you sign in.
Methodical, security-minded, kind to people who broke their laptop. Kit never touches an account without an approval and never asks anyone for a password.
The chat on the staff portal. Answers how-to questions from the IT articles (VPN, MFA, encryption, software requests) and says where the viewer's own ticket or access request stands. Never resets anything, never grants access.
When a people event arrives it reads the person from the HR system, applies the role matrix, and writes the plan: one Access Plan row per system with the group or role and the action. For a joiner it drafts the welcome note for the manager. It never creates, changes or removes an account; the plan waits for approval.
When a leaver event is set to In progress it takes each Approved Access Plan row and performs it in the connected system: deactivate the identity account, remove vault access, reassign seats, retire the device. Every step is held for approval by policy, and the system's response is written as evidence. Nothing is deleted.
Every morning it reads every managed device from the device tool, updates encryption, OS and last check-in, and for anything out of compliance drafts a short note to the person with their manager copied. After 7 days it raises a ticket. Never locks or wipes a device.
Every hour it reads new findings from the compliance or cloud security tool, keeps medium and above, closes duplicates against open tickets, picks the owner from the asset, and drafts the ticket with what to do and the evidence needed. Creating the ticket waits for a person; Critical findings alert the IT lead immediately.
Every Monday it reads seat usage from the identity provider's application assignments and the last-login data, counts seats unused for 60 days per product, lists renewals within 45 days, and drafts the reclaim or renew note for the licence owner. Never removes a seat.
For each system due a review it lists every user with their last login and manager from the identity provider, recommends removal for leavers and 90-day-idle accounts, writes the package, and emails it to the reviewer. Decisions are recorded by the reviewer; removals go through the executor behind approval.
When a ticket arrives it categorises it, drafts the reply from the IT how-to articles, and marks anything that needs a person: access requests (checked against the role matrix), password or MFA resets, security matters. Never resets anything, never grants access.
The role matrix (which systems and groups each role gets), the leaver rules, device standards, and the IT how-to articles. Replace the starter text with your own.
Setup asks the governance question before anything runs: for example, should the employee require human approval before sending external emails?
We use analytics cookies to see which pages help and which don’t. Nothing loads until you choose. Cookie Policy
Hello there.
AI agent. It can make mistakes, and a human reviews anything that matters.