anatoliadata
Menu

Illustrative pilot / 03

From a vague answer to a reviewable decision.

One fictional Turkish customer-support item shows how a guideline, a human review and a handoff record fit together.

Demonstration only — not client workInput → review → handoff

01 / The input

A small rule.
A consequential nuance.

The policy and both candidate answers below are invented for this demonstration. They are not a real customer policy or a production dataset.

Fictional policy excerpt

“A delivery address can be changed in the account before a shipping label is created. After that, the customer should contact support.”

User question · Turkish

“Siparişimin teslimat adresini değiştirebilir miyim?”

02 / Human review

Which answer earns the label?

The reviewer checks policy fidelity, a useful next step and clarity. Fluency alone is not enough.

Candidate A · not preferred

“Evet, sipariş verdikten sonra adresi her zaman hesabınızdan değiştirebilirsiniz.”

Fails the policy check: “always” removes the shipping-label condition.
Candidate B · preferred

“Kargo etiketi henüz oluşturulmadıysa adresi hesabınızdan değiştirebilirsiniz. Oluşturulduysa destek ekibiyle iletişime geçin.”

Preserves the condition and gives the next step.
01 / Policy fidelityKeep the condition intact
02 / Next stepTell the customer what to do
03 / ClarityUse plain Turkish

03 / Example handoff

A decision your team can inspect.

The selected answer, reason and review status travel together. In an actual pilot, the schema and acceptance criteria would be agreed with your team first.

See the full pilot process ↗
Illustrative record · JSON
{
  "item_id": "demo-001",
  "preferred_response": "B",
  "reason": "Policy condition preserved; clear next step",
  "review_status": "illustrative"
}

No client data or measured outcome is represented here.

Your data, your rules

Let's define a pilot
worth reviewing.

Talk about a pilot