Your customers never wait again.
Your customer support, resolved by an AI agent trained on your tickets.
J. Carter · web chat · 9:41 pm
“My order never arrived, I want a refund”
- Customer verifiedemail + order #48213
- Order and tracking readShopify · UPS, lost in transit
- Policy checkedrefund allowed, under $150
- Refund issuedStripe · $84.00
- Customer answered, ticket closedZendesk · solved
- −70 %cost per ticket
- 3×customer satisfaction
- < 1 minfirst response, 24/7
- 1,000of your tickets replayed before you sign
Figures published by the market leaders in 2026 (Parloa −70 % cost per contact, Decagon 3× CSAT, Lorikeet ~1 min first response). Yours are measured on your tickets before you sign.
How it works
Support that resolves. Not a chatbot.
We train the model on your tickets, plug it into your helpdesk and run it for you.
01 Model
Trained on your tickets
It has read your resolved tickets. It recognises the patterns and applies the resolution that worked.
02 Resolution
It resolves, it doesn’t deflect
Refunds, changes, access: it acts in your tools, within the limits you set.
03 Operation
Run by M18, paid per resolution
Monitoring, QA, improvements. A ticket handed to your team is not billed.
No more L1. No more L2.
What a support team gains. Everything you already measure, moving the right way.
J. Carter · web chat · 9:41 pm
“My order never arrived, I want a refund”
- Customer verifiedemail + order #48213
- Order and tracking readShopify · UPS, lost in transit
- Policy checkedrefund allowed, under $150
- Refund issuedStripe · $84.00
- Customer answered, ticket closedZendesk · solved
- Order never arrived21:41 → 21:42
- Wrong size received22:07 → 22:08
- Refund not showing23:30 → 23:30
- Reset my password02:15 → 02:15
- Your refund of $84.00 is on its way.
- Votre remboursement de 79 € est en route.
- Ihre Rückerstattung von 79 € ist unterwegs.
- Tu reembolso de 79 € está en camino.
- Tone98
- Accuracy99
- Policy100
- Resolution96
- Parcel lost1,240
- Double charge860
- Lost access after update410
- Change address390
→ sent to product
The model
Trained on your tickets. Not on the internet.
Before it talks to a customer, the model reads your last twelve months of resolved tickets.
- 01
It learns the patterns
Parcel lost, double charge, lost access: it groups the recurring reasons and how often they come back.
- 02
It learns your best resolutions
For each pattern, the answer your best agents gave, the steps they took, the tone they used.
- 03
It knows what it has never seen
A new case goes to a person, with everything already gathered. Then the model learns from the answer.
| pattern | tickets | resolution |
|---|---|---|
| Parcel lost | 1,240 | learned ✓ |
| Double charge | 860 | learned ✓ |
| Lost access after update | 410 | learned ✓ |
| Change delivery address | 390 | learned ✓ |
| Supplier dispute | 36 | new → human |
Proof
Proof, not promises.
- 01
Replayed on 1,000 of your tickets
Before launch, the model answers your past tickets. We compare with what your team did.
- 02
The resolution rate, in writing
You get the report. It becomes the acceptance criterion of the build.
- 03
Paid per resolved ticket
A ticket resolved end to end is billed. A ticket handed to your team is not.
Published resolution rates
Better every week.
During the replay, the model learns from every gap with your team. Example of a replay report.
The market leaders publish 70 to 85 % of tickets resolved by AI (Fin: 76 % on average; Decagon: 70 % at Chime, 2026). Yours is measured on your tickets before you sign.
Compliance
- GDPR
- DPA & NDA first
- EU or US hosting
- Full audit log
- Human in the loop
- No training on your data
- SOC 2 ready
Helpdesks & tools
Works inside the helpdesk you already run.
No migration. The agent picks up tickets where they arrive and acts in your shop, billing and CRM.
+ any tool with a login
- Zendesk
- InIntercom
- FrFreshdesk
- GoGorgias
- HubSpot
- Salesforce
- ShShopify
- StStripe
- Gmail
- Slack
Diagnosis · 60 seconds
What does L1 / L2 cost you? What would AI give back?
Three answers. You get a yearly figure and what it replaces. Conservative by design.