Core Product

The AI engine behind every automated resolution.

Six AI modules, one unified platform. Here's exactly how each works under the hood — the models, the data flows, and the outcomes they produce for your store.

01 Computer Vision

AI Return Approval Engine

When a customer submits a return, they attach a photo of the item. Our computer vision model analyzes the image against your configured return policy — checking for damage, verifying the product matches the order, and cross-referencing your SKU catalog. Approved returns trigger an automatic refund to Stripe or PayPal. A human is only involved when a return falls outside your defined rules.

  • Photo analysis detects product condition, damage, and item identity
  • Policy engine supports value thresholds, SKU exclusions, and time-window rules
  • Auto-refund dispatched to Stripe or PayPal within 4 seconds of approval
  • Rejected requests routed to an agent with reason code and image annotation
  • Full audit trail for every decision — overridable at any time
94%
Auto-approval rate
3.8s
Avg decision time
0
Human touchpoints
return_analysis · #RTN-8821
Vision Analysis Output
✓ Product identified: Blue Slim-Fit Jacket (SKU-4421)
✓ Condition: Minor seam defect — rear left panel
✓ Policy match: Manufacturing defect → Full refund
✓ Order age: 12 days (within 30-day window)
Approved
Decision in 3.8s
$64.00
Refunded to Stripe
96%
Model confidence
0
Human review
02 Live Order Data

Smart Order Tracking Bot

The most common support ticket — "where is my order?" — is handled entirely by AI. It connects to your Shopify or WooCommerce order API in real-time, retrieves live tracking data, and responds conversationally. No scripted flows. The bot understands natural questions like "I ordered 3 days ago and haven't heard anything" and gives an accurate, useful answer every time.

  • Live API pull from Shopify, WooCommerce, Amazon, and BigCommerce
  • Natural language understanding — not keyword matching
  • Handles edge cases: delays, split shipments, cancelled items
  • Proactively notifies customers of delays before they ask
  • Escalates to an agent if a shipment is genuinely lost or overdue
1.1s
Avg reply time
24/7
Availability
99%
Resolution rate
live_chat · session #LC-2284
Hi, I ordered 3 days ago (order #4882) and haven't gotten any shipping update. Getting a little worried — is it on the way?
Customer · 1.1s ago
Hi Jessica! Good news — order #4882 shipped this morning from our California warehouse. It's with FedEx now, estimated delivery Thursday by 8 PM. Tracking: FX-559204117. Want me to text you when it's out for delivery?
AI replied in 1.1s · pulled live FedEx + Shopify data
03 Machine Learning

Payment Fraud Detection

Payment fraud and chargebacks quietly destroy ecommerce margins — stolen cards, repeat fraudsters, billing/shipping mismatches, and velocity abuse. Our ML scoring model, trained on over 2 million historical orders, analyzes every incoming order in real-time and assigns a risk score before fulfillment. Orders above your threshold are automatically held for review; everything below ships without friction.

  • Signals include card history, order velocity, device fingerprint, IP geolocation, and address mismatch
  • Configurable threshold — hold anything above score 70, 80, or custom
  • Flagged orders sent to Slack or the dashboard queue instantly
  • Model retrains continuously as you confirm or clear flagged orders
  • Reduces chargebacks and protects your payment processor standing
91%
Fraud catch rate
<2%
False positives
140ms
Scoring latency
fraud_scoring · order #ORD-9923
Risk Signals Detected
Billing/shipping ZIP mismatch+34 pts
Card linked to 2 prior chargebacks+30 pts
5 orders from same device in 1 hour+16 pts
IP geolocation 1,200mi from address+11 pts
Risk Score91 / 100
LowThreshold: 75High Risk
⛔ Order held. Slack alert sent to store owner.
04 Natural Language

AI Chat Assistant

A fully automated support agent across live chat, SMS, and social DMs. Unlike scripted chatbots, it uses a fine-tuned NLP model that understands natural conversation — customers don't need specific keywords or menus. It handles returns, tracking, FAQs, and escalations, and logs every conversation to your CRM automatically.

  • Deploys on your website, SMS, Instagram, and Facebook Messenger
  • Context-aware across multi-turn conversations, not one-off queries
  • Logs to HubSpot, Zendesk, or Gorgias automatically
  • Rich replies: order cards, return labels, tracking links
  • Seamless handoff to a human agent with full transcript attached
98.9%
Intent accuracy
98%
CSAT score
nlp_pipeline.process()
1
Inbound message received
Webhook fires within 400ms. Message queued for NLP processing.
2
Intent classification
Fine-tuned model classifies: return / tracking / complaint / FAQ / escalation at 98.9% accuracy.
3
Entity extraction
Pulls order numbers, product names, dates, and sentiment from the message text.
4
Action execution
API call to Shopify, Stripe, or CRM. Response generated with live order data.
5
Reply sent + CRM synced
Customer gets a reply; a CRM ticket is created with full context.
05 Generative AI

Ticket Summarization Engine

When a complex, multi-thread ticket lands in your helpdesk, agents waste minutes just catching up. The summarization engine reads the full thread, extracts the core issue, scores sentiment, assigns urgency, and surfaces a suggested action — all before the agent opens the ticket. Average handle time drops 60% from day one.

  • Summarizes tickets of any length into 3 structured lines
  • Outputs core issue, sentiment (1–10), urgency, and suggested action
  • Native integration with Zendesk, Gorgias, Freshdesk, and email
  • Detects duplicate tickets and groups related issue clusters
  • Agents rate summaries — the model improves from feedback
60%
Handle time saved
98ms
Summary latency
ticket_summary · #TK-3388
Original Thread — 11 messages
"Hi I placed an order 8 days ago... still not here... tracking says pending... I emailed twice already... need it for an event this weekend... can someone actually help..."
AI Summary
Issue: Order #5521 stuck at dispatch 8 days — tracking not updating.
Sentiment: 8/10 frustration — time-sensitive event deadline.
Suggested action: Carrier trace + offer express reship or full refund.
🔴 Urgent Shipping Delay 2nd Contact
06 Analytics

Multi-Store Analytics Dashboard

A unified operations view across all your connected storefronts. See ticket volume, resolution rates, refund totals, fraud flags, and AI performance — updated continuously. Built for operators managing multiple brands or regional stores who need a single source of truth without logging into five dashboards.

  • Supports unlimited storefronts — Shopify, WooCommerce, Amazon, BigCommerce
  • Real-time data refresh — no manual report generation
  • Exportable reports: CSV, PDF, or scheduled Slack digest
  • Per-module AI performance scores — see where automation is working
  • Anomaly alerts: ticket spikes, fraud surges, refund clusters
Stores supported
Live
Data freshness
dashboard · all stores · this week
4
Stores connected
91%
Avg AI resolution
$31.4K
Refunds processed
18
Fraud blocked
Store Breakdown
Apex Gear (TX)
93%
Luxe Beauty (NY)
88%
UrbanKicks (OR)
95%
Security & Compliance

Enterprise-grade from the ground up.

We handle order data, payment signals, and customer PII at scale. Security isn't an enterprise upsell — it's foundational.

🔒

SOC 2 Type II

Independently audited annually against the SOC 2 Type II framework for security, availability, and confidentiality controls.

🛡️

Encryption Everywhere

TLS 1.3 in transit, AES-256 at rest. Enterprise plans support Customer-Managed Encryption Keys (CMEK).

🌍

GDPR & CCPA Ready

Full data subject rights support, configurable data residency, and a Data Processing Agreement available for all customers.

🔍

Full Audit Logging

Every AI decision and user action is logged immutably with full reasoning, exportable for compliance review.

99.9% Uptime SLA

Redundant infrastructure with automated failover. Enterprise plans get a 99.99% SLA with service credits.

🧪

Annual Pen Testing

Third-party penetration tests every year, plus a responsible disclosure program for security researchers.

Developer API

Headless integration for any custom stack.

Every module is accessible via REST API. Build custom workflows, trigger AI actions from your own backend, or pipe results into your data infrastructure.

Available Endpoints
POST/v1/returns/analyzeSubmit return + photo
GET/v1/orders/{id}/statusLive tracking data
POST/v1/fraud/scoreOrder risk scoring
POST/v1/tickets/summarizeSummarize thread
GET/v1/dashboard/metricsStore analytics
DEL/v1/conversations/{id}Delete session
View full API docs →
// Node.js — Submit a return for AI analysis
const WMTech = require('wmtech-sdk');
const client = new WMTech({ apiKey: process.env.WMTECH_KEY });

app.post('/shopify/returns', async (req, res) => {
  const result = await client.returns.analyze({
    orderId: req.body.order_id,
    photoUrl: req.body.photo_url,
    storeId: 'store_apex_001'
  });

  // result.decision → 'approved' | 'rejected'
  // result.refundAmount → 64.00

  if (result.decision === 'approved') {
    await stripe.refunds.create({
      charge: req.body.charge_id,
      amount: result.refundAmount * 100
    });
  }
  res.json(result);
});