Evaluate the whole conversation, including clarification and escalation, rather than a single polished answer.
Choosing a product
Start with a broad request: ‘I need a gift for someone who loves cooking.’ A helpful assistant asks about budget, preferences, and timing before showing a small set of relevant products. Test whether the suggestions match the answers, and whether each product link works.
Next, try a constraint: ‘I need a waterproof boot in my size.’ The assistant should distinguish documented product attributes from assumptions and make stock limitations clear.
Understanding the details
Try a sizing question with an ambiguous measurement. Does the assistant ask for the missing detail and refer to the right size guide? Then ask a product-care question. The response should come from the instructions for that product, without inventing a maintenance routine.
For a deadline-driven purchase, ask whether an order can arrive before an event. A useful answer explains published delivery windows and their limits. It should not turn an estimate into a guarantee.
Getting help after purchase
Ask for order status without providing enough identifying information. Check that the assistant follows your verification process before revealing order details. Then try a return that falls outside the normal policy: the assistant should explain the policy and offer a route to the team.
Knowing when to step aside
Finally, ask to speak to a person. This should be a straightforward request, with a clear handover and no loop of repeated automated replies. Include this scenario in every evaluation, even when the assistant answers the other six beautifully.
- Gift discovery
- Product requirements and availability
- Sizing clarification
- Care instructions
- Delivery deadlines
- Order status or a return
- A direct request for a person
Try the idea in your own store.
Explore the assistant and see how a useful conversation feels.
Experience Convi