One evidence-backed reply — from five backend systems

The retailer runs an online shop with a five-figure catalog and a few thousand orders per week. Requests come in across four or five channels in parallel — Zendesk, support@ email, Amazon messages, chat widget, Instagram — each channel delivering the same thing in a different form.

Every request cost 5–10 minutes, three of them pure data-gathering: the order number in the free text of the email, the answer in Postgres, the tracking at the carrier — three to five system switches per case.

Reporting across all channels was impossible: “How many shipping complaints last month?” meant three channels, three categorization worlds, and a manual Excel exercise.

Marketplace SLAs are strict — Amazon demands 24 hours, or account-health penalties follow. As volume grew, individual cases slipped through.

Piko presents every request in structured form — customer, order reference, request category, urgency — no matter which channel it came from. Classification and references are fields, not free text.

Backend lookups run automatically based on those fields — order status, availability, return eligibility — and appear in the reply draft with a source note: the order data with its primary key, the tracking with its last event and timestamp.

Zendesk remains the ticketing system, the shop backend the order system, the PIM the product system. Piko reads, classifies, and writes reply drafts — and changes no records. Staff confirm or correct.

A request that used to take six minutes is confirmed and answered in about one — the draft sits ready beside it, complete with evidence.

“Shipping complaints last month” is now a filter instead of an Excel job — reporting across all channels has become a standard query.

The 24-hour marketplace SLA is reliably met again, because no cases are left sitting unclassified.