We automate customer support for mid-market e-commerce. On the PandaChat framework, our own, we build a digital twin of your support operation, then run it alongside your team and improve it every week.
Predictable monthly cost · No surprise invoices · You stay in control
Customer · web chat · 21:47
Wo ist meine Bestellung #48213? Sie sollte gestern ankommen.
“Where is my order? It should have arrived yesterday.”
PandaChat · 21:47
Ihr Paket hat unser Lager am Montag verlassen. DHL hält es im Depot Graz, neue Zustellung: morgen bis 17:00. Ich melde mich, falls sich etwas ändert.
08:00 · WhatsApp · Viber · Slack · email
What happened, what we caught, what needs you. Ask it anything and get an answer from live data.
Resolved by the twin
89%
+4.1 this week
First reply
6s
day and night
Handed to your team
8%
with full context
Value this month
€2,323
0.7 FTE capacity
Conversations, last 14 days
resolvedhanded over
Needs your approval
Refund €1,200 doesn't match order #3391.
Frozen by the anomaly watcher
Runs on the stack you already have
The support team that doesn't sleep, watched by people who do.
Why teams come to us
1,284
Thousands of “where's my order?” tickets in Slovene, Croatian and German. Agents who speak all three barely exist.
Dev roadmap
Checkout redesign
ERP migration
Carrier integration
support fix · in 12 months
Every fix that would help support sits behind the roadmap, so your team keeps copy-pasting between five systems.
★★★★★
“The bot said my parcel was delivered. It wasn't.”
New review · 2 min ago
Most AI failures don't crash. They answer politely, and you hear about it in a one-star review.
More hires Another chatbot
A chatbot answers: it hands your customer a link and waits. A twin resolves: it checks your systems, follows your rules and closes the ticket.
Typical chatbot
My order #48213 was supposed to arrive yesterday. Where is it?
You can track your order using the link in your confirmation email. Anything else I can help with? 😊
Customer writes again tomorrow
PandaChat twin
My order #48213 was supposed to arrive yesterday. Where is it?
It left our warehouse on Monday and DHL is holding it at the depot. New delivery: tomorrow before 17:00. I'll message you if that changes.
Resolved in 38 s
A digital twin, not a chatbot
Fitted to how your company works: the people who answer, what they know, the systems they check, the rules they follow, and the checks a good team lead runs every day.
In your support operation
In your PandaChat twin
A team of AI specialists
One routes, one knows your documents, one works in your systems, one fact-checks, one writes in your tone.
Your documents, rules and tone
Plus a running list of what is missing, sent to the person who knows the answer.
Shop, ERP, marketplace, invoicing
Connected, with limits a person sets on anything that moves money.
Followed, every time
What goes to your people, and what never gets an AI answer, like legal threats or chargebacks.
Morning briefing & control room
Every promise made to a customer followed until it is kept.
Every answer checked first
Against your data before it is sent. Top questions re-tested every night.
A value ledger in euros
What was handled, priced, with assumptions your CFO can check.
Live right now
Real numbers from July and August 2026, on web chat and email. Counted conservatively.
Everything else went to their team, untouched.
7,600+
conversations answered
in two months
€7,000+
of support capacity
the hires they didn't need
600+ h
agent time given back
about two full-time agents a month
2,700
answered after hours
nights and weekends, 6 languages
From more than 11,000 real e-commerce conversations we handled between July and September 2026.
4 in 10
ask where an order is or when it will arrive.
6 in 10
are one of five questions: status, delivery, payment, changes, returns.
1 in 3
arrive outside business hours, when nobody is at the desk.
1 in 13
want to change an order, and only before it ships.
And the regional details generic tools miss: R1 invoices, bank transfers that haven't shown up yet, and the quirks of local carriers.
How the twin is built
Layer 1
Shop, ERP, warehouse, carriers and invoicing, connected. A person sets the limits on anything that moves money.
Layer 2
A team of AI specialists, each an expert in one area: orders, returns, carriers, your documents, your tone.
Layer 3
Every answer cross-checked against your data before it reaches a customer. Anomalies freeze for a person.
Layer 4
Email, live chat, WhatsApp, Messenger and Facebook, in every language your customers write in.
How we work with you
tickets · last 90 days
stack
Shopware · ERP · WMS · 3 carriers
01 · Weeks 1–3
We learn how your support really works: your tickets, your team, your policies, your systems.
Before it talks to a single customer
02 · Weeks 3–8
We build the twin on the PandaChat framework and test it on your real questions.
handled end-to-end
week 1 → week 12
03 · Monthly, capped
We go live channel by channel and make it better every week. You see what it was worth every month.
Everything else stays in Zendesk, Freshdesk or whatever you use today, untouched. We start with the messages we're sure about and take on more as the twin proves itself. No big-bang switch, no ticket left behind.
PandaChat twin
Order status
Delivery
Address change
Invoices
Your helpdesk
Complaints
Legal threats
Chargebacks
Everything new
You stay in control
You don't log into a dashboard. We come to you. Our team in Slovenia watches every briefing and steps in when something needs a person.
Anomaly watcher
just now
Refund of €1,200 requested on order #3391. The order total is €89.
Mitja · PandaChat team
Monthly value ledger
June · Nordic Living
€2,323
Support capacity delivered
999 tickets handled end-to-end, about 0.7 of a full-time agent.
€1,840
Losses prevented
Refunds that didn't match their orders, rejected by a person. Only what was stopped counts.
244
After-hours conversations
Nights, weekends and holidays in 6 languages, nobody on call.
Every figure comes with its assumptions, ready for your CFO.
Proof, not promises
Counts alone don't survive a budget meeting. So every month we put it in money: the capacity you didn't have to hire, the losses we stopped, the hours nobody had to cover.
If PandaChat prevents just two multilingual support hires, it has already paid for itself.
PandaChat is our framework. We're the three people who map your operation, build its twin and run it with you.
CEO & Builder
Builds the AI team: a specialist for every system, a supervisor on every answer.
Engineer & Customer Partner
Integration engineer, hands-on with customers every day.
Chief Performance Officer
Value ledgers and anomaly intelligence that catch problems before customers do.
AI Innovation of the Year
SDC Awards London
Golden Innovation Award
Go Digital GZS
Best Tech Park Innovation
DIGI-SI Challenge
Best Use of Generative AI
SRIP GoDigital
No. A chatbot copies one agent. On the PandaChat framework we build a digital twin of your whole support operation: specialists for each system, a supervisor on every answer, your rules and escalations, and a morning briefing. And our team runs it with you.
Mapping and building take 4–8 weeks. The twin is tested on your real questions before it talks to a customer, then goes live channel by channel.
A one-time setup, typically €25k–35k for 3–5 connected systems, then a capped monthly fee, typically €4.5k–6k. Simpler scopes start lower. Never per ticket.
No. The twin handles the repetitive volume. Your people handle the judgment calls, in a clean inbox with everything already looked up.
Every answer is checked against your data before it is sent. When it can't verify, it hands over to your team with full context. We fix mistakes at the source: the document, rule or policy.
Established e-commerce with thousands of conversations a month, answers spread across several systems, and customers writing in more than one language. Not for a simple store with a few hundred FAQs.
“Over 20 years ago, I was on the other end of the line. That's why we built PandaChat.”
A 30-minute review with one of us. We look at a sample of your tickets and your stack, and tell you what we'd automate first.
Predictable capped pricing · Setup in weeks · A real team in your timezone