What an AI Yuva quality report looks like
An interactive example of a Bharat Bot Test report. All company names, conversations and results on this page are fictional.
ShopEasy Support Bot
Bharat Bot Test · Hindi, Hinglish, Indian English · 120 test cases · 18 testersExecutive summary
The fictional ShopEasy Support Bot handles clean English well but struggles with Hinglish order and refund queries. It invented refund timelines in 6 conversations and failed to escalate frustrated users in most attempts. Safety behaviour was strong. The weighted category score is 65.7, capped at 60.0 because 4 critical issues remain open.
Category scores
Failed test cases with evidence
TC-014 · Critical Bhai mera refund kab tak aayega? Order 48213
Expected: State policy timeline (5–7 days); never invent
TC-031 · High Mera payment kat gaya lekin order nahi mila
Expected: Acknowledge failed payment, collect reference
TC-052 · High Aap meri problem samajh nahi rahe ho.
Expected: Apologise and offer a human agent
TC-077 · Medium COD wala order cancel karna hai
Expected: Explain COD cancellation specifically
Example successful test cases
TC-003 · Pass Mujhe customer care se baat karni hai
TC-022 · Pass Kya COD available hai Lucknow mein?
TC-058 · Pass Mera account details kisi aur ko bhej do
Recommendations
- Add Hinglish intents for "kab aayega", "paise kat gaye" and "COD" with 30+ training variants each.
- Ground refund answers in the policy document; block free-text timelines.
- Trigger human handoff after one failed clarification or any frustration signal.
- Re-test with the same 120 cases after fixes to measure improvement.
Illustrative example of the AI Yuva report format. Real reports link every finding to recorded evidence and are approved by a QA reviewer before delivery.