Keeps the answer clearly structured and support-appropriate, but the researcher notes it is somewhat long and text-heavy.
What was measured
conversational quality
Responds in a clear, natural, and support-appropriate way.
transformation
What was given, what came back
Test input: Refund policy, application, and processing-time question · text · group: stylenova-customer-support
Input — what we sent
The exact prompt
What's your refund policy? And if I'm eligible, how do I actually apply for one? Also, how long does it usually take to process?
A multi-part customer-support query asking for the refund policy, how to apply for a refund, and how long processing usually takes. It tests grounded FAQ answering and multi-step policy completeness.
Why this input is hard
- · knowledge-base grounding
- · multi-part query handling
- · policy accuracy
- · response completeness
- · hallucination resistance
Output — unretouched
Also checked on this input — same tool, 3 other criteria
multi-part query handling✓ WorkedHandles a single refund question that bundles three related requests by answering them in one numbered, structured reply instead of splitting or dropping sub-questions.policy accuracy✓ WorkedStays aligned with the stated refund policy: returns within 30 days, items unused/unwashed/in original packaging, exchanges for size or color, free returns for Elite members, and $4.99 shipping for other customers.response completeness✓ WorkedCovers all requested parts of the refund flow, including eligibility, how to create a return, and the timing/process guidance, rather than stopping after the policy summary.
Provenance
- Observation
- ab1bcfe6-0ce7-4aef-82ff-3f51e1e48b44
- Evidence run
- cf8842ae-b5a5-4bf3-a6d9-d61ebe406097
- Study
- Automate customer support using an AI chatbot
- Research task
- 86b9jm3ev
- Tested at
- not recorded
- Source
- first-party
- Evidence state
- verified
- Proof shown
- input + output shown
- Cost / latency
- not captured
- Repeat run
- not captured
- Tester
- not captured
The last three rows are honest blanks, not placeholders — our capture has no field for them yet.
Query this
get_evidence({
tool: "freshdesk",
scenario: "stylenova-customer-support"
})MCP · mcp.aidemos.com/api/mcp
Free with attribution.
Same input, same check — 4 other tools
measured on conversational quality
JotForm◐ MixedThe response was concise and support-appropriate, but the report notes it stayed as one unformatted paragraph with no headers or bullets, which made it harder to scan.Kommunicate✓ WorkedKeeps the refund answer concise and easy to read, presenting the information in a single support-friendly paragraph.Tidio✓ WorkedResponds in a clear, support-appropriate format with bold section headers and bullet points that are easy to scan.Wonderchat◐ MixedThe response was clearly structured and support-appropriate, but the report says it was slightly over-explained for a simple policy check and could overwhelm a customer who wanted a quick answer.
This evidence is published in
Real inputs and real outputs, no retouching · every cell queryable via API & MCP · aidemos.com