What’s inside: routing · assistants
The question this file answersWhich of today's tickets is the complaint that will cost us a customer — and how deep in the queue is it?
Fits: e-commerce teams handling 200–2,000+ tickets a day where “where is my order?” and return questions share a queue with complaints — and where turnover means the team is partly new every season
Not for: shops with a few dozen tickets a day, or teams already running a helpdesk bot that answers status from the OMS — the sorting is not your cost
Typical, from the retail and e-commerce playbook. A mid-size shop sees 200–2,000 tickets a day. An agent reads each one, looks the order up in the OMS, writes a reply, escalates the hard ones and logs it. “Where is my order?” outnumbers everything else — and buries the complaint that needed a person an hour ago. It hurts most in peak season, when the same team gets twice the queue and turnover means half of them are new.
Nine o'clock in peak week. The agent's queue is not two thousand messages; it is the complaints and the odd cases, with the “where is my order?” and “can I return this?” tickets already answered from the OMS and closed — or waiting as drafts until you approve auto-send. A new agent can work the queue in their first week, with a summary and a suggested reply on each ticket. You can see, per category, how many tickets closed without an agent and how fast complaints were first answered. The support team is still there; the lookups are not. Zendesk's (2024) first-response figure is in the metric block — theirs, not ours.
Typical, not a measured client result. Every figure here comes from the playbook source named below.
faster first response on triaged tickets — Zendesk (2024)
Before: status questions drown complaints in one queue. After: Zendesk (2024) reports 60–75% faster first response with triage — judgement calls stay with agents.
Zendesk CX Trends (2024) reports that AI-powered triage reduces first-response time by 60–75%; Gartner (2023) puts 30–40% of interactions handled without escalation in deployed organisations. Playbook range 50–70%, retail and e-commerce playbook. Vendor and analyst figures, not our measurement.
Your helpdesk stays — Zendesk, Freshdesk, Gorgias, Intercom or the one you run — and the OMS keeps the order truth: NetSuite, Brightpearl, Shopify, Magento. In front of the agent's view we add a sorting and drafting step:
First scope: the three categories you already know outnumber everything.
Judgement on complaints. High-value customers. Policy exceptions. Anything that is not a lookup.
If the order record in the OMS is late — a shipment marked despatched that is still on the dock — the automatic status reply is confidently wrong, so auto-send starts where the OMS is trusted. Classification is not perfect: a complaint phrased politely can be sorted as a status question, so an agent samples the automated categories daily and a customer reply reopens the ticket. A platform without a usable API leaves a manual step. The playbook's 50–70% is on triage and routine answers; complaints needing judgement, high-value customers and policy exceptions are outside it.
Audit, about two weeks (€1.5–3K): we read a month of tickets, count them by category and by whether the answer was a lookup, and check what the helpdesk and OMS APIs expose. Pilot, 2–4 weeks — low-to-medium complexity in the retail and e-commerce playbook (€10–20K): the top three categories, replies as drafts until you approve auto-send per category, the support team reviews a sample daily. Production: further categories, then returns end to end. From you: helpdesk admin access, OMS read access, your returns policy and tone-of-voice notes.
The path: free 60-second estimate → free 20-minute review → paid audit of this one process (€1.5–3K, typically two weeks) → pilot with your people in the loop (€10–20K, weeks, not quarters). No transformation programme. Prices are public, on the services page →
That depends on your volumes and wage costs — this page will not invent the number. The free 60-second estimate runs that calculation from your answers, with every multiplier sourced.
Not a named Aperanda client. Process file · Ops.
Deep-dive process file. Volumes, weeks and sources come from the industry playbook; nothing here is a named client.