AI rules employees can find, read and follow.
A policy in a shared drive changes nothing. The employee AI hub gives every person one page: which AI tools are approved, how to ask for another, the policy to read and acknowledge, a 13-minute course, and a plain account of what monitoring records about them.
Interactive example with fictional records. Try it. Open the policy and acknowledge it, request a tool, look at the privacy tab.
Which AI tools we have approved, how to ask for a new one, our AI policy and a 13-minute course.
- For employees
- One page, no console account
- Course
- About 13 minutes, in 5 modules
- Evidence produced
- Acknowledgements and completions
- Plans
- Hub on every plan. Course on Govern
What an employee sees.
The hub of Demo Corporation is live, with fictional data, seen as one of its employees. It has six parts:
- Home. What needs doing: a policy to acknowledge, a course to finish, a request with a question on it.
- Tools. The approved AI tools, each with what it is approved for.
- Requests. Ask for a tool that is not on the list, and follow the request to a decision.
- Policy. The current AI policy in full, with one button to acknowledge it.
- Course. Five short modules on using AI safely at work, each ending in a few questions.
- Privacy. What your employer monitors, what it never collects, and what has been recorded about you.
An employee does not get a console account and cannot see the inventory, risk tiers, other people's activity or internal comments on a request.
- The browser extension opens the hub from its popup, and from its warning and block pages
- Works on a phone
Give people a faster route than going around you.
Shadow AI is mostly a queueing problem: someone needs a tool today and the approval route is unknown or slow. The hub puts a request form where the list of approved tools is.
A request names the tool, what it is for and what kind of data would go into it. In the console, requests arrive in one queue with the demand for each tool counted, so five people asking for the same thing is visible as one decision. A reviewer can ask the requester a question, add an internal note the requester never sees, and approve or decline. Declining needs a reason, which the requester sees, and can point to an approved alternative.
Approving is a real decision, not a reply: the tool enters your inventory as sanctioned, with any conditions you attach, and appears on everyone's approved list.
- Open and overdue requests are counted on the console overview
- Check the vendor fact card for the tool before you approve it
Acknowledgement and training you can show.
Policy acknowledgement. Publish a version of your AI policy to the hub. Each employee reads it and acknowledges it. Each acknowledgement is recorded against the version that was read, with the date. You see coverage for the current version and can export the list.
AI course. A vendor-neutral course of about 13 minutes: what generative AI is and where it fails, what must never go into an AI tool, checking outputs and who is accountable, spotting risky tools and asking for approval, and a module generated from your own rules. You can add a module of your own. Each module ends with questions, and completion is recorded per person.
The EU AI Act asks providers and deployers to take measures to support the AI literacy of their staff. A course with completion records is one such measure. It is not, by itself, compliance with the Act, and the course gives no legal advice.
Acknowledgement and completion figures can appear on your trust center and are cited as evidence by the questionnaire engine.
Signing in, and what the hub itself stores.
There are four ways in, and an administrator chooses which are allowed:
- Single sign-on through your OpenID Connect provider, the same connection the console uses.
- A personal link an administrator creates for one person, or a shared link for a group.
- The browser extension, which hands the device's identity to the hub.
- A one-time link by email to an address at one of your own domains. This needs an email provider. The hosted service does not send email yet, so this option is available only on a self-hosted server that has one configured.
How someone signed in decides what they may do. Someone who opens a shared link without identifying themselves can read the tools, the policy and the course. Someone who types their own work address there can acknowledge the policy and take the course, recorded as self-declared, and is never shown activity data.
The hub follows your identity mode. In pseudonymous or anonymous mode it stores a salted hash in place of the work email, and the admin views show counts and departments, not names. A hub session is a separate credential from a console session: it cannot call a console route.
- When an employee is erased on request, their acknowledgements and course records go, and the totals are kept as counts
- Hub sessions last 7 days by default
Hub details
- Address
- /hub/your-company on the service's domain. An unknown or switched-off hub answers exactly like one that never existed.
- Sign-in
- OpenID Connect single sign-on, admin-created personal or shared links, the browser extension, and an emailed one-time link (self-hosted with an email provider only, at present).
- Employees can see
- Approved tools, their own requests, the policy, the course, what is monitored, and their own recorded activity.
- Employees cannot see
- The inventory and risk tiers, other people, internal reviewer notes, or any console page.
- Evidence
- Policy acknowledgements by version, course completions by module, tool requests and decisions. Kept for the audit-log retention period.
- Identity modes
- Named, pseudonymous or anonymous, the same setting the sensors use.
- Plans
- The hub, tool requests and policy acknowledgement are in every plan. The AI course and completion tracking are in Govern and Enterprise. Pricing
Questions
- Do employees need an account?
No console account. They reach the hub through single sign-on, a link an administrator gives them, or the browser extension. None of these lets them into the console.
- Can the hub email people a sign-in link?
The code supports it, and it works on a self-hosted server with an email provider configured. The hosted service does not send email yet, so there you would use single sign-on, admin links or the extension.
- Does the course make us compliant with the EU AI Act?
No single measure does. The Act, as amended, asks for measures that support AI literacy and does not fix a level or a format. The course and its completion records are a reasonable measure to point to. Whether it is enough for your organization is a question for counsel.
- What can an employee see about themselves?
The same things you can see about them, and no more: which AI tools were recorded and for how long. No content is recorded for anyone, so there is none to show.
- Can we change the course?
You can add one module of your own text and questions, and the "our rules" module is generated from your own policy settings. The four general modules are fixed.
Put the rules where people work.
Start a 14-day trial with every Govern feature. Mark your approved tools, publish your policy to the hub and send the link to a pilot group. No card is needed to start.