How answers stay accurate
It only answers from your data: your catalogue, your orders, your policies. When it doesn't know, it hands over rather than guess.
At the same brand, customers often rate it higher than the support team.
E-commerce customer service
Fewer tickets. Better answers. More sales.
Your own AI system instead of one more account on a shared platform. It adapts to how you work. Gem CX builds it on your helpdesk, your catalogue and your orders, and runs it for you. It answers your customers 24/7, guides them all the way to the purchase and hands your team anything that needs a person, with the full context. Nothing to build on your side.
Your team
66%*
of conversations resolved without it
Your customers
85%*
average CSAT
Sales with AI30 days
€300k*
in orders where the AI helped the buyer
*Measured in production, Sep to Oct 2026. Sales: two brands, counted only when the AI helped the buyer.
Chosen over a helpdesk vendor's AI within a week.
Your store
AI agent, replying now
Message
An example exchange. Lookups shown as they happen.
Connected to your tools
Zendesk
Freshdesk
Gorgias
LiveChat
Shopify
Magento
WhatsApp- and many more
Trusted by
Behind Gem CX
Built by someone who ran a support team
Fabien Monnier founded Gem CX after ten years in e-commerce customer service: first leading a customer service department, then working alongside 20+ companies on their CX and automation projects. AI vendors know software. CX consultants rarely write code. Gem CX does both. Every system is designed and followed by Fabien, from the first call to go-live.
- 10 years
- in customer service
- 20+
- companies supported on their CX and automation projects
Case studySofa ClubFurniture retailer, United Kingdom
Put head to head with its helpdesk's AI,Gem CX was kept within a week.
| Helpdesk vendor's AI | Gem CX | |
|---|---|---|
| Setup | Setup3 weeks, with the vendor's technical team | Setup1 week, nothing to configure on the brand's side |
| How it works | How it worksDecision trees to configure | How it worksConnected to the catalogue and the orders |
| Outcome | OutcomeReplaced | OutcomeKept, live since August 2026 |
Before
Thousands of questions a month on web chat and WhatsApp. The team first rolled out its helpdesk vendor's AI, over three weeks, with the vendor's technical team.
The switch
The result fell short: decision trees to configure, and a ceiling set by the platform. Gem CX built a system connected to the catalogue and the orders. Put head to head after one week, it was the one the team kept.
Today
The system answers on web chat and WhatsApp, advises customers up to the order and hands the team what needs a person. Gem CX improves it every week.
3 in 4
conversations resolved without the team
8 in 10
positive customer ratings (CSAT)
1,750
conversations a week on web chat and WhatsApp
24/7
instant answers, no queue
Gem CX production data for Sofa Club, September 2026. CSAT: share of positive ratings left after the conversation. Resolved: closed without a handover to an agent.
How it works in practice
When a customer wants a person
The moment a customer asks for someone, or a decision belongs to your team, it opens a ticket in your helpdesk with the conversation, the order and a summary your team reads in 10 seconds. Nobody asks the customer to repeat anything.
7,000 conversations a month don't become 7,000 tickets. Only the ones that need a person reach your helpdesk.
What it asks of your team
Very little. I learn your processes from a sample of past tickets and macros, not from hours of meetings. Your team signs off the rules and reviews the first conversations. Live in about two weeks, billed monthly, no commitment.
“In one week, it was ten times better, and really low effort from me.” CX Director of a client brand.
Sales, not just fewer tickets
It advises like a good salesperson: it compares products, checks stock and delivery, and takes the customer to the basket. The orders it helped are tracked, so you see what it brings in.
An order counts only when the AI really helped the buyer, not because the chat was open.
Tested on your traffic before it goes live
Before it answers a single customer, it runs in the background on your real conversations. Its replies are compared with the ones your team actually sent. On email, it starts as drafts your team approves. Each type of request switches to sending on its own only once its replies match your team's.
It only goes live where it has matched your team, or done better.
We measure our mistakes
Every reply sent on its own is checked against what happened next in the ticket. You see how many needed a correction, not just a resolution rate.
Each correction is traced back to what the AI was given, then fixed in its rules.
What your helpdesk's AI can't do
Need something your helpdesk's AI can't do? Your request is built into your system instead of waiting on a vendor's roadmap. It's tested on your past conversations and goes live once your team approves it.
Already built for clients: an outlet item's dimensions found from its original product page, an order number typed with a space or a stray letter.
Why not build it in-house
You could build a first version in-house. Keeping it right every week is the real job: new products, new policies, new edge cases. Gem CX runs it, watches it daily and improves it every week.
Every conversation is logged and rated by the customer.
Complex catalogues, several languages
It reads your catalogue as it is, with your own attributes, and your spreadsheets if part of the data lives there. One system can cover several stores and answer each customer in their own language.
In production in English and French, on catalogues where the answer depends on compatibility, dimensions or a photo.
Security
- Identity verified
- Approved actions only
- Rules in code
- No payment data
- No model training
- GDPR and DPA
- Conversation logs
- Gradual rollout
- Off switch
What the system does
One system, connected to your helpdesk, your store and your carriers. It covers the requests that fill your queue, from the first product question to after-sales.
What it does, not just says
- Verifies the customer with a one-time code before showing any order
- Tracks the parcel live with the carrier and flags a lost or stuck parcel to your team
- Puts an order on hold in your store when the customer asks for a change
- Opens warranty claims and return requests with photos, checked against your rules
- Builds a priced quote and sends it as a PDF
- Checks that a product fits the customer's space, and the delivery date
- Suggests in-stock alternatives
- Books a showroom visit or a callback
- Hands over with a ticket and a 10-second summary in your helpdesk
Every action is agreed with your team first, and only the actions you approve are switched on.
What their customers say
Every conversation ends with a rating and an optional comment. Three of them, left by customers the system served.
85%
satisfaction
average, across all brands
The best AI Assistant that I've encountered. Questions are understood and the answer is directly related to the question.
Excellent service, I am stunned by the quality of the exchange, bravo!
One of the best customer service chats I have dealt with. Thank you so much for all your valued help.
Pricing
From
€900per month
AI usage included. One-time setup from €1,500 to €6,000, depending on your integrations.
| Conversations a month | Monthly | Email replies |
|---|---|---|
| Up to 5,000 | €900 | + €300 |
| Up to 12,000 | €1,500 | + €500 |
| Above 12,000 | Quoted | Quoted |
- Web chat, WhatsApp, every language and every store included
- Designed, deployed, operated and improved every week by Gem CX
- Above your tier, the next one applies from the following month, after a talk with you
- Most AI agents bill per "resolution", and a customer who simply stops replying counts as one. We bill a flat monthly fee instead, with no annual prepayment.