Test how the Agent handles a product
Use the product test surface to verify how the Agent presents a real indexed catalog item and its available data.
Evaluate product-specific behavior without confusing a catalog problem with a general conversation problem. The test should show whether the selected item is available to the Agent and represented accurately.
An Agent combines instructions, customer-facing interface settings, Knowledge, catalog data, and optional tools. A reliable setup is tested with realistic customer questions before deployment. Tests should cover expected answers, missing information, product scenarios, escalation, and any optional interaction that the team has enabled.
Before you begin
Access: Open Agent setup and testing with read access to the catalog and permission to run product tests.
- Select a synchronized product that should be available.
- Know its current title, variants, price, stock, and market context.
- Confirm indexing is not currently in progress.
Work in the smallest owning area described below and keep the current customer-facing state available while you prepare the change. Before clicking any final action, confirm the active company, Agent, store, language, and market shown in Humind. A missing control can indicate read-only access or a capability that is not configured for this company. In that case, record the intended task and ask an administrator to review the exact permission or dependency. Do not bypass the boundary by sharing an account, copying data into another area, or promising a capability that the workspace does not expose.
Step-by-step workflow
Choose a representative product
Start with an item that has enough data to verify. Then repeat with a variant, out-of-stock, or market-specific example if those cases matter to the storefront.
- Record the product and expected market.
- Check its source data before asking the Agent.
Run realistic customer prompts
Ask questions about fit, attributes, alternatives, variant availability, or price using language a shopper would use. Avoid revealing the expected answer inside the prompt.
- Ask one factual and one recommendation question.
- Include a boundary question for missing data.
Compare the response with the catalog
Verify that the Agent uses the correct product and does not merge details from another item. Check values that can vary by variant or market carefully.
- Compare identifiers and displayed attributes.
- Distinguish a source-data error from response behavior.
Correct the owning layer and retest
Fix source data or sync when the catalog is wrong, indexing when the product is unavailable, and guidance only when the response behavior needs adjustment.
- Make one focused correction at a time.
- Repeat the same prompt after the required sync or index completes.
Important limits and operating notes
- Product tests depend on the currently available indexed catalog.
- Prices and availability may vary by market and variant.
- Narrative Knowledge should not override volatile catalog facts.
- A product test does not validate the storefront embed or live domain.
Verify the result
- The intended product can be selected for testing.
- The Agent distinguishes product and variant facts correctly.
- Missing data is acknowledged rather than invented.
- The repeated test reflects the focused correction.
Keep a short record of what you tested, which customer scenario you used, and what changed. This makes later troubleshooting more precise and helps another teammate reproduce the result without relying on memory.
Troubleshooting
The product cannot be selected
Check synchronization, product visibility, indexing rules, and final index status. Verify that the product belongs to the current company and intended market.
The Agent uses details from another product
Compare titles, variants, metadata, and filters for ambiguous values. Improve the underlying product distinction or filter configuration before adding behavioral guidance.