How to set up your customer service stack

Customer service for a plant webshop is not "add a chat widget and call it done". It's a stack of channels — each with its own response window, audience, and failure mode. Pick the wrong combination for your volume and staffing and you create the worst outcome: a channel that promises help and doesn't deliver. This article walks you through how to design the stack honestly, sized to your reality.

Why "the stack" matters more than any single channel

Every channel you add is a promise. A chat bubble says "we are here", an open WhatsApp number says "message us like a friend", an email address says "we will read this". If you can't keep the promise, the channel actively hurts you — a customer who waits four hours on what looked like instant chat is more likely to leave a public complaint than one who never had the option.

So the design question is not "which tool is best", it's: which combination of channels can I actually staff with the response times those channels imply?

Response time benchmarks per channel

These windows reflect retail support norms in NL/BE/DE. Treat them as the contract you sign by switching the channel on:

• Live chat: <2 minutes during business hours. Outside hours the widget should show "offline — leave a message".

• WhatsApp Business: <30 minutes business hours, <4 hours always. Silence reads as ignoring.

• Email: <4 hours business, <24 hours always. Auto-reply is the bare minimum.

• Contact form: same as email.

• FAQ + AI agent: instant.

Miss these and the channel becomes a liability.

Three honest scenarios — pick yours

Customer service design is a function of messages per day and hours you can actually be reachable. Three realistic plant-shop profiles:

Solo reseller, ~5 messages per day

You answer when you can, often after 18:00 or between other tasks. Live chat is a trap here — you'll miss the 2-minute window 80% of the time, and every miss is visible. Recommended stack:

• Contact form on a /contact page, with one auto-reply that sets expectations ("we reply within one working day").

• WhatsApp Business with business hours configured and a clear auto-message outside those windows.

• FAQ page built from your last 30 days of support questions (see How to use FAQ as self-service).

No chat widget. Adding chat at this volume looks ambitious and ends up looking neglected.

Solo reseller, ~50 messages per day

You can't keep up by hand anymore — but hiring isn't economical yet. This is the AI-agent sweet spot:

• AI agent on top of your FAQ + product data + order data, answering pre-purchase plant questions ("Is Calathea pet-safe?", "What's the pot size of this Monstera 12cm?"), delivery-status checks, and policy questions.

• Contact form / WhatsApp as the human escalation when the AI can't resolve. Set the rule: 3 turns without resolution = handover to human.

• Email for everything async (returns, complaints, B2B requests).

The AI handles the volume; you handle what AI can't. Done well, this configuration absorbs 60–70% of incoming questions automatically.

Small team (3 people), ~200 messages per day

Now live chat earns its keep, because you can actually staff it:

• Live chat with a clearly published SLA (e.g. "live 09:00–18:00, <2 min response") and visible offline state outside hours.

• AI agent as front line on chat — handles 50%+ of questions instantly, hands over to a human when needed.

• WhatsApp Business connected to the same inbox (Trengo, Tidio, or similar) so the team isn't switching apps.

• Email + contact form for async, ticketed in the same inbox.

The unifying inbox matters here — at 200/day you cannot have three apps and three notification streams.

Tool comparison for plant resellers

Our recommended default: Trengo. Dutch-built, strong multi-channel inbox (live chat + WhatsApp + email + Instagram + Facebook DM in one view), AI agent built in, scales from solo to team without switching tools. For NL/BE plant resellers, Trengo is the cleanest choice — set it up once and it grows with you.

Other options worth knowing

If Trengo doesn't fit your scenario, the alternatives:

• Tawk.to — free, basic live chat only. Fine for validating whether chat fits before paying. No AI, weak mobile app, limited integrations.

• Tidio — strong mid-market option, good Shopify integration, has an AI agent (Lyro), reasonable price. A close second if you don't need the multi-channel inbox yet.

• Crisp — clean UI, good shared inbox, fair AI features. Comparable to Tidio; pick on UX preference.

• Intercom — enterprise-grade, expensive (€100+/month and up quickly). Skip unless you're past 500 messages/day and need workflow automation. Overkill for most plant resellers.

• WhatsApp Business app (free, mobile-only) — fine for the 5/day scenario as a standalone. For higher volume, use the WhatsApp Business API via Trengo or Tidio so multiple agents can answer.

What an AI agent can and can't do

AI is useful for plant shops because most questions are repetitive and factual. AI handles well: "Is this plant pet-safe?", "How big does Monstera Deliciosa get?", "What's my delivery status?", "Do you ship to Belgium?". Those are FAQ + structured-data lookups, and AI nails them at 90%+ accuracy when fed your real catalog and policy.

AI fails on: account issues, specific complaints ("the plant arrived broken"), refund disputes, B2B requests, anything emotional. Hard rule: 3 turns without a clear resolution = human handover. Don't let AI loop the customer through "could you rephrase?" — that's worse than no AI.

Plant example: "Is my Calathea Orbifolia safe for cats?" — AI answers from your spec data instantly. "My Calathea arrived with three broken leaves" — human-only, AI should escalate the moment it detects damage signals.

Make your business hours visible

Whichever stack you pick, publish the hours and response windows in three places minimum: contact page, footer, and the widget itself. "We reply Mon–Fri 09:00–18:00 within 4 hours" is a contract customers can plan around. Silence about response time creates the unmet-expectation problem you're trying to avoid.

Next action

Pick your scenario from the three above based on actual messages-per-day this past month (count them — don't guess). Tomorrow, write down on one page: which channels are on, what the response window for each is, and where those windows are published on the site. If any channel doesn't have a stated window you can keep, switch it off this week. Then build out the FAQ that sits underneath the stack — see How to use FAQ as self-service — because every channel works better when the easy questions never reach a human in the first place.