QUIETSTACK AUTOMATION

SHOPIFY · AI SUPPORT AGENTS · N8N AUTOMATION · EU

AI support agents with 5 safety gates — for Shopify stores.

Quietstack builds support agents that answer from your own policies and cite their source, hand risky requests to your team, and back-office n8n automations that end the copy-paste work. You approve everything before it goes live.

Try to break our agentGet a free demo on your store
  • [ok] high-risk intents gated — High-risk intents (refunds, payments, order changes) never reach the answering model — they go to a human with a ready ticket.
  • [ok] eu ai act art. 50 — Designed for EU AI Act Article 50 transparency requirements — in force since August 2026.
  • [ok] gdpr — GDPR-aligned processes; DPA available. NDA before any access.
  • [ok] async by default — Async by default — calls only when they reduce risk. Loom walkthroughs + written docs.
  • [ok] fixed scope, fixed price — Quoted in writing before work starts. No per-ticket fees, no hourly meters.

LIVE · OUR OWN STORE

Live 5-Gate Sandbox & Live Eval: Brewkraft Espresso Gear

Brewkraft (brewkraft.coffee) is our own espresso-accessories brand. This agent runs on its real product and policy data. Press a test button, or type your own attack.

  1. GATE 01

    Pre-screen

    Refunds, payments, damage claims and order changes never reach the answering model — in any EU language. A person gets a ready ticket.

  2. GATE 02

    KB grounding

    The model only sees the policy and product entries relevant to the question.

  3. GATE 03

    Source citation

    Every factual sentence cites its knowledge-base entry, and every number must exist in that entry. Otherwise nothing is sent.

  4. GATE 04

    Scope

    Outside the knowledge base? The agent hands off instead of guessing.

  5. GATE 05

    Injection & output shield

    Instruction overrides, discount codes and foreign links are blocked in and out.

BREWKRAFT SUPPORT AGENT · DEMO

AI SYSTEM

You are interacting with an AI system, not a human. Disclosure under EU AI Act Article 50. Answers come only from Brewkraft's product knowledge base; refunds, payments and order changes always go to a person. How this works

$ Ask about Brewkraft tampers, fit, shipping or returns — or press a test button to try to break the agent.

TRY TO BREAK THE 5-GATE AGENT

Live eval · 36/36 as expected · 7 EU languages

2026-10-25 · claude-haiku-4-5 · commit 3fc5457

Real requests to this agent with a live model, expected outcomes fixed before the run. Money intents in EN, DE, FR, ES, NL, SV, DA reached a person without a model call (170–409 ms). 12 model answers passed the per-sentence source check (1368–3700 ms); every reply was read against the store's catalogue and policies.

  • Reviewer scenarios4/4
  • Refunds, damage, order & payment issues8/8
  • Order status (WISMO)4/4
  • Pre-sale questions11/11
  • Wholesale & discounts2/2
  • Injection attacks4/4
  • Off-topic1/1
  • Follow-up with history2/2

Demo environment · Brewkraft (brewkraft.coffee) is our own store, operated by SankPower SIA · no conversation content is stored · order data is fictional demo data · AI-generated answers

PRICING

No per-resolution tax.

Fixed scope, fixed price, quoted in writing before work starts. Your support volume can grow without your bill growing with it.

TYPICAL HELPDESK AI ADD-ONS

$0.90–$1.00

per resolved conversation (Gorgias AI Agent, annual / monthly billing)

Cost scales with every conversation the AI closes — the busier your season, the bigger the bill.

QUIETSTACK

Fixed-scope packages

no per-ticket fees · you own the workflows

Model usage is passed through at cost. Retainer is optional.

  1. 01

    AI Act & Support Eval Audit

    Adversarial prompts run against your live store widget, EU AI Act Article 50 transparency check, PDF readiness report within 48h.

    €150–300fixed

  2. 02

    Sandbox Demo

    A demo agent grounded in your store's real FAQ and policies, with the 5 gates switched on. You test it yourself.

    Freebefore you pay

  3. 03

    Pilot 5-Gate Agent

    10–14 days: FAQ & policy grounding, order-status (WISMO) lookup, 5 safety gates, eval benchmark on your own tickets, AI disclosure badge, Loom handover.

    €1.5–2.5kfixed

  4. 04

    Production Agent + n8n Backend Ops

    Helpdesk integration plus custom n8n workflows — e.g. inbox triage, supplier invoice PDF extraction, returns routing — with a larger eval set.

    €3–8kfixed

  5. 05

    Care & Eval Retainer

    Monthly eval accuracy audit, knowledge-base updates, prompt-drift monitoring. Raw LLM usage billed at cost, no markup.

    €450–1,200/ month

Prices exclude VAT where applicable. Final scope and price confirmed in writing before any work starts.

FREE DEMO — BUILT ON YOUR OWN FAQ

Enter your Shopify store URL to get a free custom AI-agent demo.

We ground a demo agent in your store's public FAQ and policies, switch on the 5 gates, and send you the link. You try it before you pay anything.

Used only to prepare your demo. See privacy.