Run several agents in one Ask AI request

Learn the order in which Ask AI in Test Management runs the dataset, maintenance, and generation agents, and what it asks you before each handoff.

A changed requirement rarely needs one thing. The data behind the test cases changes. The test cases that use that data are impacted. A new rule needs coverage that does not exist yet. In Ask AI you describe the change once. Ask AI works out which agents the change needs and runs them one after another. It stops for your approval at each step.

The order the agents run in

Datasets come first, updates second, new test cases last. Each agent builds on what you approved in the step before.

The order protects your repository. Datasets run first, so the updates use the new data. Updates run before generation, so new test cases fill only the remaining gaps. The generator never rewrites a test case the maintenance agent has fixed.

An agent that your request does not need is left out. The rest keep this order.

How Ask AI reads your request

Ask AI decides what to run from the words you use. The following table lists the principles it follows.

What each request triggers

The following table shows how common requests map to the three agents. Run means the agent runs. Offered means Ask AI asks you first and you can skip it. Skipped means the agent does not run.

When a request is unclear, Ask AI asks one focused question before it routes anything. The same happens when you mention a document you did not attach. Answer in the chat and it continues.

Follow a handoff in the chat

The following walkthrough uses a Jira issue that changes a promo code rule. The datasets, the test cases that use them, and the missing coverage all need work.

1. Open the folder that holds the affected test cases, click Ask AI, and send a prompt such as PAY-482 changed: update our promo code data, add test cases, and fix existing ones that use that data.
2. Ask AI summarizes what the issue changes and asks whether to update the test data first. Click Update test data, or click Skip to test cases to leave the datasets as they are.
3. The Test Data Generator Agent asks what kind of data you need, for example boundary values, random values, or both. Answer in the chat.
4. When the datasets are ready, click Save all datasets to add them to the project, or Reject. The chat confirms how many datasets were saved.
5. Ask AI names the test cases that use the saved dataset and asks whether to update them. Click Update test cases, or Iterate more to change the datasets first.
6. The Test Case Maintenance Agent returns the updated test cases in a result card. Click View to check each change in the review panel. Then click Update all test cases, or Ignore the selection to leave them unchanged.
7. The Test Case Generator Agent returns the new test cases for the remaining gaps. Their steps use the saved data as parameters, such as {{cart_value}}. Click Add all test cases, or Ignore the selection.

The chat closes with a summary line, for example: 2 datasets saved, 2 test cases updated, 5 test cases added. Each agent shows its own progress card while it runs, so you always know which agent is waiting on you.

Skip or stop a handoff

You are never locked into the chain. Click the skip option to move to the next agent without changes. A rejected result is not applied, and the next agent still runs with what you approved so far. Ask AI waits for your answer at each handoff. The chain does not move on until you click an option. Nothing that you did not approve is saved.

Limits

Keep these limits in mind when you chain agents:

• The order is fixed. You cannot ask Ask AI to generate new test cases before it updates existing ones.
• Only the Test Case Generator Agent writes new test cases. The maintenance agent updates and the dataset agent creates data.
• Agents fill system fields only. Custom field values are not generated.

Next steps

Read the page for each agent in the chain:

• Generate test datasets with Ask AI: Learn what the dataset agent asks and what it saves.
• Update test cases with Ask AI: Review each update before it is applied.
• Generate test cases with Ask AI: Steer the new test cases with follow-up prompts.