Partner guide · Policy to workflow
Turn client AI policy into a tested workflow checklist
Many clients now have, or want, an AI use policy. The policy says what should happen; this guide shows how to make the parts that concern IT support actually happen in NetZen, and how to prove it before employees rely on it.
Which client AI rules can NetZen implement?
Four kinds: what knowledge the AI may answer from, who may see which tickets and devices, which actions need a person's approval before they run, and when the AI must hand over to a person. Rules about other tools, model training or general employee conduct belong to the policy itself and the systems that enforce it.
Shadow AI is the reason this matters. Among the 600 breached organizations in IBM's Cost of a Data Breach Report 2025, one in five reported a breach due to shadow AI, meaning AI tools used without approval (IBM, July 2025).
Which knowledge can the assistant use?
Only the sources you add for that client, or deliberately share across clients. Keep confidential material out of the sources entirely: an article marked Internal is hidden from the client portal but is still available to AI answers. Give every source an owner who approves changes. Product detail is on client AI access controls.
Policy-to-control worksheet
Blank worksheet (CSV)| Policy requirement | Owner | NetZen setting | Limitation | Test | Evidence |
|---|---|---|---|---|---|
| Confidential HR documents must not reach AI answers | Client HR lead | Leave them out of knowledge sources | Internal labels hide articles from the portal, not from AI | Plant a test phrase in an excluded document, then ask for it | The phrase appears in no answer or tool output |
| Software installs need a manager’s approval | IT lead | Ticket approver: site manager | Manual and scheduled runs do not ask for approval | Request an app as an employee | Approval recorded on the ticket |
| Only staff in scope see client tickets and devices | IT lead | Roles: employee, site manager, client admin | Device operations are separate role add-ons | Sign in as each role | Allowed and denied lists match |
| Company data in external AI only through approved tools | Client security owner | ChatGPT, Claude or Copilot Studio connected through NetZen | Records a person retrieves reach that AI workspace | Connect as test users in each role | Only records allowed by each role returned |
| A person takes over when AI cannot finish | Operator | Handoff to assigned technician | Staffing and hours are the operator’s agreement | Ask an unanswerable question | Ticket assigned, AI stopped |
When do ticket actions require approval?
For actions started from a ticket, the operator chooses the approver: the requester, IT support, the site manager or nobody for low-risk actions. The AI cannot grant an approval a person is required to give. Actions run by hand or on a schedule do not ask for approval, so restrict who can run them, and note that multi-step run books cannot be scheduled. Built-in Microsoft 365 and Google Workspace checks are read-only; any change needs an automation the operator built, tested and enabled. See automations.
What data reaches external AI workspaces?
When ChatGPT, Claude or Copilot Studio is connected, each person signs in to NetZen and keeps their own role, and the records they retrieve reach that AI workspace. Decide with the client which workspaces are approved, under which subscriptions and data terms. See connect AI.
How do we test access, handoff and outcomes before launch?
Sign in as each role and compare what is allowed and denied with the policy. Ask a question from a source that should be excluded and confirm it is not used. Run the pilot action with its required approval. Ask something the AI cannot answer and confirm a person takes the ticket. Record the evidence in the worksheet, with a review date.
What remains the client’s policy, or another tool’s job?
Device-wide data loss prevention, rules for AI tools outside NetZen, staff training and the policy's own review cycle. Saying this clearly keeps the client's expectations right, and your recommendation credible.
Questions partners ask
No. NetZen enforces the parts that live in its own settings: which knowledge the AI uses, who can see what, which actions need approval, and when a person takes over. Device-wide data loss prevention, model choices elsewhere and staff training stay with other tools and the client.
Often. In Microsoft and LinkedIn’s 2024 Work Trend Index (May 2024), 78% of AI users said they bring their own AI tools to work. Giving them an approved, useful assistant is part of the answer; the policy covers the rest.
Tickets record the conversation, approvals and automation runs, with status, who started them, errors and output. Agree with the client which of these they want reviewed and how often.
Bring NetZen to your next client
You lead; we support the demo, proposal and first rollout. 15% of the NetZen subscription each client pays, for 24 months. About the program.
Become a partner