Let employees connect their own Claude, ChatGPT, or Microsoft Copilot to their SkillsDB data, govern what those assistants can do, and revoke access at any time.
Quick Summary: The AI Assistant settings let administrators allow employees to connect their own AI assistant to their SkillsDB data. You control whether assistants can only read or also make changes, whether they can run high-impact operations, and you can see and revoke every connection in the company.
The AI Assistant feature lets each employee connect their own AI assistant — Claude, ChatGPT, or Microsoft Copilot — to their SkillsDB data. Once connected, the assistant can answer questions and, if you allow it, make changes on the employee’s behalf.Every connection is scoped to that employee’s own permissions. An assistant can never see or do anything the person could not see or do themselves in SkillsDB, and every change it makes runs through the same permission checks and approval steps as the same action taken by hand. This scoping is enforced on every request, so tightening any setting binds already-connected assistants on their very next call.Governance uses three switches, layered so each one only matters once the one above it is on:
Setting
Controls
Default
AI Assistant Connections
The master switch. Lets employees connect an assistant at all, for read access.
Off
AI Assistant Writes
Whether connected assistants can make changes, not just answer questions.
Off
AI Assistant — High-Impact Operations
Whether assistants can run high-impact operations — anything that notifies whole populations, creates or deactivates sign-in accounts, or permanently destroys data (today: deleting labels, creating or deactivating people, and starting or ending assessment cycles).
Off
Everything is off by default, so nothing is exposed until an administrator turns it on. Each switch applies to the whole company. Only administrators can change these settings or manage connections.
Note: The AI Assistant settings appear only when the feature is enabled for your SkillsDB account. If you do not see the AI Assistants entry on the Settings page, contact your account manager.
Navigate to Company > Settings. In the Configuration section, find AI Assistants and select Configure.
2
Allow connections
Turn on AI Assistant Connections. Employees can now connect an assistant with read access to their own data.
3
Allow changes (optional)
Turn on AI Assistant Writes to let assistants make changes as well as answer questions. Every change is still bound by the employee’s own permissions and the usual approval steps. This switch appears only after AI Assistant Connections is on.
4
Allow high-impact operations (optional)
Turn on AI Assistant — High-Impact Operations to additionally allow the highest-impact actions. This switch appears only after AI Assistant Writes is on.
Turning AI Assistant Connections off stops every assistant in the company on its next request. The connections themselves are not revoked: if you turn the switch back on, they work again without employees reconnecting. Turning AI Assistant Writes off leaves reading working and refuses every change.A switch hidden because the one above it is off keeps its saved value. When you turn the upper switch back on, the hidden switch reappears in the state you left it.
Warning: Turning on AI Assistant Writes does not upgrade connections made while it was off. Those connections stay read-only until the employee disconnects and connects again, leaving Make changes on your behalf ticked.
Employees connect and manage their own assistants on the AI Assistants page, at app.skillsdb.com/settings/ai-assistant. Every employee can open this page, whatever their role, but nothing in the SkillsDB menus links to it yet — share the address with them, or point them to Connecting Your AI Assistant. The server address every assistant uses is https://prod.skillsdbnext.com/mcp.The page shows:
Connect your AI assistant — one tab each for Claude, ChatGPT, and Microsoft Copilot, with step-by-step setup instructions and the SkillsDB server address. Select Copy address to copy it.
Connected assistants — the employee’s own connections, each showing whether it Can read your data or Can read your data and make changes, and a Disconnect button.
When an employee connects, SkillsDB asks them to sign in and approve the access. If AI Assistant Writes is on and the assistant asks to make changes, the approval screen shows Make changes on your behalf, ticked by default. The employee can untick it to connect for reading only.
Note: Microsoft Copilot connections are set up by your IT team in Copilot Studio. Employees send IT the SkillsDB address from the Microsoft Copilot tab, then sign in to SkillsDB when Copilot prompts them.
If AI Assistant Connections is off, the page tells employees that their organisation has not turned this on yet.
The Connected assistants section of the admin settings lists who has connected an assistant and when they last used it, so you can audit access across the company.Each row shows:
Column
Meaning
Person
The employee who connected the assistant.
Assistant
The name the assistant reported about itself when it connected (for example, Claude), or — if it reported none.
Access
Read only or Read and change, reflecting the scope that connection was granted.
Connected
The date the connection was created.
Last used
The date the assistant last made a call, or Never if it has not been used.
To end a connection, select Revoke on its row. Revocation takes effect on the assistant’s next request. A revoked row is labeled Revoked.Revoking a connection does not stop the employee from connecting again — they can reconnect straight away from their own AI Assistants page. To remove someone’s access for good, deactivate the person in SkillsDB: a deactivated person’s assistants stop working on their next request.
Warning: Revoking a connection cannot be undone from here. If the revoke fails, the row stays active and shows an error — try again rather than assuming it worked.
No. Every connection is scoped to the connecting employee’s own permissions. An assistant can only see and do what that person could see and do themselves in SkillsDB, and every change runs through the same permission and approval checks.
What happens to connected assistants when I change a setting?
The change binds on the assistant’s next request. Policy is checked live on every call, so turning a switch off takes effect immediately for assistants that are already connected, without waiting for them to reconnect.
I turned on AI Assistant Writes, but an employee's assistant still cannot make changes. Why?
The connection was made while AI Assistant Writes was off, so it was granted read access only. The employee needs to select Disconnect on their AI Assistants page and connect again, leaving Make changes on your behalf ticked.
What counts as a high-impact operation?
High-impact operations are the actions that affect many people or are hard to reverse: anything that notifies whole populations, creates or deactivates sign-in accounts, or permanently destroys data. Today that means deleting labels, creating or deactivating people, and starting or ending assessment cycles. Everything else an assistant can change stays governed by AI Assistant Writes and the employee’s own permissions.
Can I allow assistants for only part of my organization?
No. Each of the three switches applies to the whole company. Within that, every assistant is still limited to what its employee can see and do in SkillsDB.
How do I tell an assistant's changes apart in the Activity Log?
You cannot. A change an assistant makes is recorded under the employee who connected it, exactly as if they had made it themselves.