Marketplace Question Draft Reply v2 | Zervanor
← All automations

Marketplace Question Draft Reply v2

Assistenza

Every hour the oldest unanswered customer question on Trendyol is picked up. A support agent drafts a short Turkish reply from the question and product context, and the seller sees the draft next to the original question and approves or rejects it. Only approved text is ever sent to the customer. Marketplace answer-time directly affects seller rating, and questions arrive faster than a small team can watch them.

1 workflows · 2 apps & agents

TR

A finished operation: the trigger, the steps and the agents it needs, already modelled and validated against the live connector catalog.

Marketplace Question Draft Reply v2 is a finished operation across Trendyol, powered by an AI agent. Built for Turkey. Its trigger, steps and connectors are validated against the live catalog.

How it works

Marketplace Question Draft Reply v2

  1. Date Time Pencere başı (2 gün önce)
  2. Date Time Pencere sonu (şimdi)
  3. Transform Pencere başı epoch ms
  4. Transform Pencere sonu epoch ms
  5. Action Cevap bekleyen sorular
  6. Transform Soruları normalize et
  7. Condition Bekleyen soru var mı?
  8. Transform En eski sorunun zamani
  9. Transform En eski soruyu al
  10. Agent Task Yanıt taslağı yaz
  11. Approval Yanıtı onayla
  12. Condition Onaylandi mi?
  13. Action Trendyol'a yanıtla
  14. End Bitti

Apps & agents 2

Trendyol

Support Agent

Frequently asked questions

How does "Marketplace Question Draft Reply v2" work?

It runs these steps automatically: Pencere başı (2 gün önce) → Pencere sonu (şimdi) → Pencere başı epoch ms → Pencere sonu epoch ms → Cevap bekleyen sorular → Soruları normalize et → Bekleyen soru var mı? → En eski sorunun zamani → En eski soruyu al → Yanıt taslağı yaz → Yanıtı onayla → Onaylandi mi? → Trendyol'a yanıtla → Bitti.

Which apps does it connect to?

Trendyol. Each one is reached through a versioned connector pack — no extra development needed.

Does it use AI?

Yes. A built-in AI agent handles the reasoning (summarize / classify / decide) and runs in the same runtime as your workflows — no separate AI subscription.

How is the Trendyol integration modelled?

Each provider is modelled as a versioned connector pack, and the steps that need reasoning fall to first-party agents hosted in the same runtime as the workflows.

Run this behind your own product