Skip to content
Convor
Convor Docs

Test automation before publishing

Save realistic visitor messages and check whether Rules and Knowledge still behave as expected.

Changed a shipping Rule? Try the visitor message "When will my parcel arrive?" and compare the result against the response or handoff you expect. Automation tests save these cases for the next round of changes. The test checks decision logic without contacting a real customer.

Describe the visitor's situation

Open Automation → Automation tests and select Add test. Give the test a clear name, such as "Shipping deadline question". Enter a realistic sample under Visitor message. Set the relevant channel and content language so Convor evaluates the same context a visitor would have.

A useful test starts with a plausible visitor message and the conditions under which it arrives.

Decide what should happen

Under Convor should, select the expected outcome. You can choose a Rule or Knowledge article and, for more detailed checks, inspect rule matching and workflow actions. For a request to speak to a person, the expected result can be a handoff to an operator.

Expected results can include replying, adding an action or handing the conversation to an operator.

Recreate the conditions behind a decision

Use the context controls to specify whether an operator was available, whether the bot was enabled and whether someone was already assigned. For instance, test a TopSoft4U refund question first during a staffed shift and then without an available operator. You can also choose a specific evaluation time for time-limited Rules or enter the last reply time to exercise cooldown behavior.

The channel and account fields affect matching. Allegro-specific tests can include the thread type and connected account. Running a test evaluates this context without sending a provider message.

Choose Conversation event to test AI startup, a need for more information or handoff. Keep Visitor message received for an ordinary customer question. For the Convor site, compare AI startup in the widget with Email: ready-made rules should match the widget and select no message for Email.

The event, channel and operator availability form the test context.

Check actions as well as reply text

For TopSoft4U, a shipping-status question should use a published Knowledge article, while a refund approval request might require handoff. Save both cases. For Rule checks, turn on Check all matching rules and select Expected matching rules. Then enable Check selected actions to verify that the billing tag, priority change or operator assignment really matches the outcome you expect. Don't expect an AI answer when the right result is a human decision.

Developers can manage Rule definitions through POST Create an automation rule and multilingual Knowledge content through the Knowledge API. These methods configure the system, while the dashboard test checks its decisions.

Save the test and run it. After changing automation, use Run all cases to check your saved cases again. Review any unexpected results before enabling new Rules. Read the Rules and Knowledge guides for the underlying decision flow.

Was this page helpful?

On this page