Evidence · first-party tested/Query Live Databases Using Plain English with AI

When a follow-up phrase could refer either to the full order-stage breakdown or to the immediately preceding pending-paid subresult, it did not ask for clarification and instead chose the narrower pending-paid interpretation.

⚠ Struggled🧾 artifact-verifiedinput + output shownTest date not recordedDraxlr →

What was measured

Ambiguity Handling

Does the tool clarify unclear business terms instead of guessing silently?

transformation

What was given, what came back

Test input: Order pipeline breakdown with paid-pending edge case and last-month comparison · text · group: ecommerce-nl2sql-benchmark

Input — what we sent

The exact prompt

How many orders do we have at each stage right now?

Follow-up 1: What percentage of our orders were successfully delivered vs cancelled?

Follow-up 2: Are there any orders that are pending but already paid?

Follow-up 3: Compare that to last month — same breakdown, I want to see if things have improved or got worse.

A deeper operational analysis of current order stages, delivery-versus-cancellation rates, pending-but-paid edge cases, and a month-over-month comparison of the same breakdown.

Why this input is hard

· Order pipeline analysis
· Percentage calculation
· Edge-case detection
· Payment/order status joins
· Multi-turn context retention
· Month-over-month comparison
· Ambiguity handling for 'same breakdown'

Output — unretouched

Also checked on this input — same tool, 2 other criteria

Business Insight✓ WorkedCan turn numeric results into plain-English takeaways; the delivered-vs-cancelled follow-up automatically summarized the split and framed delivered orders as more common than cancelled orders.Chart / Visualization Support✓ WorkedCan auto-generate a correct chart for a simple month-over-month comparison, showing Last Month at 11 pending-paid orders versus Current Month at 20.

Provenance

Observation: a37351f9-c20e-4396-aaac-98fda74b1cb5
Evidence run: db2bb5d5-0e0e-4cb3-8d76-3555c45c23cd
Study: Query Live Databases Using Plain English with AI
Research task: 86b9y6c99
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: "draxlr",
  scenario: "ecommerce-nl2sql-benchmark"
})

MCP · mcp.aidemos.com/api/mcp

JSON · ai.aidemos.com/v1/observations/a37351f9-c20e-4396-aaac-98fda74b1cb5

Free with attribution.

Same input, same check — 1 other tool

measured on Ambiguity Handling

AskYourDatabase✓ WorkedClarified 'last month' upfront as April 20, 2026 before running the comparison, rather than guessing what period the user meant.

This evidence is published in

Query Live Databases Using Plain English with AI →

From the same study (page rebuilt from a later run)

Best AI Tools to Query Live Databases Using Plain English →

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