Verify your installation
Documentation for Tutela v3.9.0.
Check that you can sign in, inspect a harmless request, and find its activity record before inviting your team. These checks complement deployment health checks: a running server alone does not demonstrate that your chosen AI connection is inspected.
Who runs this checklist
This is a customer workflow check after you configure your own provider and policies. It is not permission for a deployment operator to seed a delivered tenant. A fresh handoff contains no customer policies, provider connections, conversations or demonstration activity. Operators verify bootstrap access, service/model readiness and empty application state separately. Run acceptance fixtures in an isolated environment; do not leave test records in the delivered tenant. Real-time search is built in, but a live search check requires your supported search-capable provider connection.
Prerequisites
- A deployment that has completed activation and an administrator account.
- One configured AI provider or connected client that you are authorized to test.
- A test account and a narrow policy scope that will not interrupt other users.
- A policy owner who has confirmed that the harmless test sentence below is permitted for the test user and client. Use an isolated test scope if an organization-wide policy intentionally blocks it.
Check your installation
-
Sign in at your deployment's HTTPS address. Open Settings > Activation and check subscription and activation separately. If either status cannot be confirmed, retry that status check before continuing.
-
Open Settings > Support & diagnostics > Service health. Record any unavailable service. If an inspection dependency is unhealthy, follow Inspection unavailable or blocked.
-
Open Policies > Test, select Text, and enter the following harmless sentence:
Documentation check: explain why leaves change color in autumn. -
Set Expected outcome to Allow, then select Run test. The test must complete and report an allowed simulated decision for the permitted test scope. Needs information, Test unavailable, and an outdated result are not passing results. Resolve the displayed cause and run again.
-
Send the same sentence through your configured AI Chat or connected client. Record the time, client, and request ID if one is shown. A simulated policy test does not send this live request for you. The expected result is a provider response about leaves and an allowed policy decision. If it is blocked, inspect the matching control with its owner; do not disable a control just to receive an answer.
-
Open Activity in the same workspace. Select a time range that includes the request and locate its session. Check the application, time, policy decision, and any evidence gaps. An allowed decision and a completed execution are separate results.
-
If this connection reports token usage, open AI Spend > Usage telemetry. Compare the latest event time and source with your request. Missing usage, unpriced usage, and zero cost mean different things; use Check AI usage reporting if the request is absent.
For a protection check, repeat steps 3–6 with a synthetic example supplied by the policy owner and its expected action. Use a dedicated test scope. Do not copy real customer information into a prompt to demonstrate detection.
If a check fails
| Observation | Next action |
|---|---|
| A page is not visible | Confirm the test account's role with an administrator. Do not grant full administration just to finish this checklist. |
| The policy test works but the client request fails | Inspect the client's connection, credential and provider settings; the test does not exercise that connection. |
| The client responds but no session appears | Check the workspace and time filters, then follow Missing activity events. |
| The decision differs from your expectation | Review the published policy's scope, priority and supplied facts before broad rollout. |
| Health or evidence is unavailable | Record the error and request ID. Retry the failed source; do not record it as a pass. |
Validate
- Sign-in and activation succeeded for the intended account and deployment.
- The policy test completed with the expected result.
- The connected client produced the expected response or policy action.
- The matching Activity session identifies the tested application and decision.
- Applicable usage reporting was checked, with missing data recorded explicitly.
Keep a short record of each result, the deployment and client versions, and the test time. An unresolved check means the installation is not yet verified for that workflow.