Evidence · first-party tested

The router followed the billing logic but substituted its own assignee label, 'Billing Support / Human Agent', instead of the instructed 'Billing Team'.

The observation

Criterion: Reliability · Scenario: Angry duplicate-charge customer routing case (group: customer-routing)

Query this

get_evidence({
  tool: "pickaxe",
  scenario: "customer-routing"
})

MCP · mcp.aidemos.com/api/mcp
Free with attribution.

Same input, other tools

Dust ✓ Worked Gumloop ✗ Failed Relevance AI ✓ Worked Zapier AI Agents ✓ Worked

Real outputs, no retouching · every cell queryable via API & MCP · aidemos.com