Everything Lumnya can do, in one place.
Grouped by the job it's doing. Everything below is live today unless it carries a “next” tag — we'd rather show you the line than blur it.
Sell · storefront
Turns questions into orders.
The Customer AI works the front of house on every page — discovery, objections, the nudge at the right moment, and the order itself.
Aurora Supply
Customer AI · online
Do you have this in medium? I need it by Friday.
Yes — 3 left in Medium, and express delivery lands Thursday. Want me to build the order?
Nordic Hoodie — M
€64 · express by Thu
grounded in your policies · escalates to a human when it should
Support · storefront
Resolves the ticket instead of deflecting it.
Answers come from your live data and your own policies — which is why they're answers and not links.
Operate · via Lumnya
Ask in plain language. Lumnya does the work.
Everything here is confirm-gated: you see the exact change before it happens, one tap approves it, and the audit trail records what moved.
>refund order #1043 and email the customer an apology
… reading order #1043 · drafting refund + apology email
preview
Refund €42.00 (full) — order #1043
Email: “We’re sorry about the delay — your refund is on its way.”
Refunded & email sent · logged 14:32 · approved by you
Analyze · the workforce
Seven specialists live today.
Each owns one domain, runs on a cadence and on demand, and attaches its reasoning to everything it reports. All of them start in recommend mode.
Inventory Forecaster
Scans stock on a schedule, flags critical and low against your thresholds, and opens priority-ranked reorder recommendations.
Profit Lens
Computes true per-product contribution margin from your costs — so you know which SKUs actually earn.
Liquidation Planner
Finds stagnant stock and proposes bounded clearance discounts you approve before anything changes.
Bundle Discovery
Mines your real order history for products that sell together and turns the pairs into bundle suggestions.
Demand Radar
Reads storefront conversations for demand your catalog didn't capture — products asked for, blockers hit.
Revenue Detective
Decomposes a move in revenue into quantified drivers: volume, basket size, refunds, mix.
Self-Learning FAQ
Collects the questions that stumped the assistant and drafts the missing FAQ entries for one-click approval.
In development
The next hires are already in training.
Named here because they're near-term — not because they're running.
Pricing Optimizer
Watches margins and elasticity, proposes the price moves worth making.
Churn-Risk Scorer
Flags the customers most likely to slip away — early enough to act.
Sentiment Pulse
Reads chats and reviews so you feel the mood without reading all of it.
Anomaly Detector
Beyond the Store Watchdog's three alarms: the same treatment for every metric you track.
Executive Summary
Your whole store, briefed like a Monday-morning memo.
Run & control
The controls that keep it yours.
Autonomy dials
Set how independent Lumnya is per area of the store — fifteen of them — with caps on value, volume and discount, plus quiet hours. Everything ships at “confirm”.
Action journal & undo
Every write recorded with how it was authorized. Reversible ones undo in a click, replaying the same guarded path they were written by.
Simulation
“What would you do?” runs the real decision path against your real settings and names the areas it could act in alone, the ones that would still ask, and what's pending. Nothing changes.
Store Watchdog
Checks every hour and tells you when orders stop, when the widget starts failing customers, or when people ask for a human and nobody answers. Measured against your own normal, so a quiet shop never gets a false alarm — and one problem is one message, not twelve.
Goals you set in words
Say the outcome — “margin above 30%”, “recover more carts” — and Lumnya tracks it against real numbers, weekly. A goal is a standing brief, not a new permission: every step it takes still passes the same caps and gates as anything else.
Trust panel
The month's ledger of what Lumnya did, what it remembers about your store — by kind, with a delete-for-good control — and how it reaches out to you, with the switch. No values are put on display, yours or your customers'.
“What can you do?”
Ask it, and the answer is generated from the tools it actually has — area by area, what it can read, what it can change, and a sentence you can copy. It can't over-promise: a page ships with its tools or the build fails.
Give it a file, not a task
Attach the supplier price list — Excel, CSV, PDF or a photo of it — and Lumnya reads the lines, matches them to your products itself, and lays out every cost change before one of them is applied. Lines it can't match are named, not skipped, and a price it can't read unambiguously it refuses to guess. A stock count works the same way, with one extra guard: it checks each quantity against what your store holds right now, so anything that sold while you were reading the list is reported rather than overwritten. Attach your returns policy and it drafts the FAQ answers and the return rules to match. Nothing written inside a file is ever followed as an instruction.
Notifications & web push
Urgent events email immediately, the rest arrive as one daily digest. Push is opt-in per device, quiet-hours aware, and carries no customer data.
Live Chats — takeover & copilot
Watch conversations as they happen, jump in as a human mid-thread with the full history, or let the reply copilot draft your response for you.
Self-learning FAQ
Gaps in your knowledge base surface as drafted answers. Approve one, and the assistant stops fumbling that question.
Widget Studio
Appearance, tone, maximum discount, escalation threshold and the knowledge it may draw on — previewed, then published to your storefront.
Smart triggers
Rules for when the assistant should step forward, based on real on-page behavior rather than a timer.
Reports & exports
A weekly report on every closed week — what it earned, what changed it, what it cost, and how the problems named the week before turned out. It ends by naming what you could change about it, and only ever things it can actually carry out for you — each one opens the chat with the request written, and nothing is applied until you say yes. Plus analytics and conversation exports. Take any of it as PDF or CSV; your data is yours.
Quality loop
Post-chat satisfaction ratings, per-message thumbs up/down, and automatic escalation when the AI's own confidence drops.
Languages & localization
Speaks your customer's language — and stays in it.
The assistant detects the shopper's language automatically and answers in it — and so does every email that follows: a confirmation link, a price-drop alert weeks later, a return update, a newsletter opt-in. The language belongs to the conversation, not to the store's default setting, so a Greek-configured store writes to its German shopper in German. Mail Lumnya sends you follows your own Lumnya language setting instead, and with nothing on record either way, English. The widget UI currently ships in English, Greek and German, with more on the roadmap — and your brand voice is a setting, not a surprise.
Platform & engine
The right model for each moment.
Lumnya routes routine work to fast models and revenue-critical moments — checkout questions, cancellations, returns — to stronger ones. The dashboard runs embedded in your Shopify admin or standalone in its own tab; the storefront widget ships as a theme app embed.
multi-model engine
Security & trust
Built so the answer to “can it?” is your call.
The guarantees underneath every capability on this page.
Data isolation — by architecture
per storestore-aurora
1,204 facts · 8.2k orders
store-nordic
312 facts · 1.9k orders
store-atlas
88 facts · 640 orders
cross-store reads: denied — always, for everyone, including us
Private beta
All of it, pointed at your store.
Join the private beta and be first in line when Lumnya opens to your store.