Customer support for licensed gaming operators
Support that answers every player, and notices the ones at risk.
Juno is a complete support desk built inside Casino Elite, a casino licensed in Belgium. Its AI answers players from your own help articles. Every message is checked for signs of gambling harm. Your team works chat and e-mail from one console, with quality scoring, shifts and reports included.
Live in production at Casino Elite
Try it
Write to Juno as a player. Watch what your agents see.
Pick a question or type your own, in English, French, Dutch or German. Then try the last button to see how Juno handles a player who might be at risk.
The conversation appears here in English, the agent's language, as the player writes.
This demo runs entirely in your browser, with three sample help articles and a small part of Juno's harm-phrase list. It is not connected to Juno or to any AI, and nothing you type leaves this page. The real product answers from your own help articles and understands far more than these examples.
Chat widget
A chat window players can use on any page of your site.
Add one script tag, or use the React component. The widget sits in an isolated layer, so your site's styles can't break it and it can't break your site.
Knows who is logged in
Your site tells Juno who the player is. A logged-in player skips the form and the chat opens in their language. Logging out or switching account starts a fresh chat.
Short pre-chat form for guests
Guests pick their language first, shown with native names and flags, then give a name, e-mail and date of birth.
Safe on shared devices
When a guest's chat ends, the browser forgets who they were and what they wrote, so the next person on that device sees nothing.
Live replies
- Answers appear as they are written
- Typing indicators and read receipts
- Unread counter on the minimised button
- "Not sent, tap to retry" on a weak connection
Photos and documents
- JPEG, PNG, GIF, PDF, MP4 and MOV, up to 7 MB each
- Paste a screenshot and preview it before sending
- Take a photo with the camera inside the chat, useful for ID checks
Help & FAQs tab
Players can search your public help articles without leaving the chat, and ask a question in chat if the article doesn't answer it.
Quick questions
Up to six one-tap questions, a greeting and an FAQ link, written by you for each language.
Opening hours
Outside your live-chat hours, new chats are paused and players are invited to e-mail you, in your own words per language. Chats already open carry on.
Conversations & tickets
Players see their open and past cases, can search and filter them, and resume where they left off.
Satisfaction rating
A 1–5 star rating with an optional comment after each chat. It feeds agent reports directly.
Comfortable on desktop and phone
A draggable minimised button. On desktop the window can be enlarged or resized, and the size is remembered.
Your look
Your colours, corner style, font, avatar and position on the page. Only the websites you list may load your widget.
The widget's own buttons and labels are fully translated into English, French, Dutch and German. Conversations themselves can run in any of 57 languages; see At a glance.
AI assistant
Answers from your own articles, not from guesswork.
The AI searches your knowledge base for the passages that match the question, then answers from them. It already understands gaming support: deposits, withdrawals, bonuses, KYC, accounts, casino, sportsbook and responsible gaming.
Grounded answers
Each question is matched against your articles by meaning, with keyword search as a backup. The assistant is told never to invent policies, contact details, or financial, legal or medical advice.
Hands over when unsure
Every AI reply carries a confidence score. When confidence is low, the case goes to your team's escalation queue.
A human persona
Replies appear under a name you choose, "Support" by default. You can add your own instructions to shape tone and rules.
Answers in the player's language
The conversation language comes from the player's choice, your site or their first message, and the AI replies in that language.
Two-way translation for agents
Agents read every message in their own language and write in it too. Juno translates the reply into the player's language and flags anything it could not translate fully.
Sorts and reads every case
Each case is filed into a category (deposit, withdrawal, bonus, KYC, account, technical and more) and tagged. Frustrated players trigger a live alert to agents.
Suggest a reply
One click drafts a reply for the agent from the conversation and your knowledge base. The agent edits it and sends it.
Chats become articles
Turn a resolved conversation into a draft help article. Drafts wait in a review queue until an admin approves them.
Switch it off anytime
Team leads can turn the AI off from the home screen. Players then see that an agent will be with them shortly.
AI runs on AWS in the EU by default. Each knowledge base can be set for the AI, for agents only, or for both.
Player protection
Every message is checked for signs of gambling harm.
The check runs on every chat message and every e-mail, even when a person is already handling the conversation. It recognises indirect phrasing and negation, so "I'm not addicted" is read correctly, while any mention of self-harm always raises the alarm.
| Level | What it recognises | What Juno does |
|---|---|---|
| Critical | Self-harm or suicide. Requests to self-exclude, including national registers such as CRUKS and the German player block. | Immediate caring reply with your helpline, a self-exclusion offer when asked, critical escalation and a console alarm. |
| High | Addiction, loss of control, debt or borrowing to gamble. | Caring reply with your helpline, high-priority escalation and a console alarm. |
| Medium | Chasing losses, rent money, desperation, asking to raise limits or for a pause. | A review ticket for your team. |
| Low | Frustration such as "it's rigged" or "I always lose". | Recorded on the message. |
The alarm keeps sounding
A critical or high alert shows a banner with sound on every console page until someone acknowledges it.
Silence makes it louder
An escalation left unanswered is raised to critical automatically.
Nothing closes it by accident
Auto-close never shuts a case with an open harm escalation.
Everything is on record
Every detection is saved on the message, and the escalation queue doubles as your responsible-gambling log.
The phrase library covers English, French, Dutch and German, and the AI also flags distress in its own replies. Juno supports your responsible-gambling process. It does not replace your policies, staff training or the obligations of your licence.
Your support inboxes, worked as tickets.
Player e-mails arrive in the same console as chats. Each one gets a ticket number, the same AI and harm checks, and a reply that goes out from the address the player wrote to.
- Connect Microsoft 365, Amazon SES, SendGrid, SMTP, or a generic webhook. Every incoming connection is verified.
- Addresses Several support addresses, each with its own language, sender name and signature, for example one for French and one for Dutch.
- Threads Replies join the player's existing ticket only when the sender is verified. Otherwise the agent sees an "Unverified sender" badge.
- Auto-reply A branded confirmation with your logo and signature in English, French, Dutch or German, with a delay you set and per-sender limits.
- AI by e-mail The AI can answer e-mails from your knowledge base, or you can leave e-mail to people.
- Outbound Start a new ticket by e-mail, choosing the address, brand and category.
- Attachments Up to 10 files per e-mail, 10 MB each, including large messages.
- Never twice The same e-mail is never processed or answered twice.
Help centre
A public help centre built from the same articles.
Write an article once. The AI answers from it, agents search it, and players read it in your help centre and in the widget's Help tab.
Browse and search
Categories with article counts, and search across titles and text.
Learn what works
View counts, most-read sorting and a "Was this helpful?" vote on every article.
Only what you publish
Only articles you mark as public appear. Agents' personal notes are never shown.
Agent console
One place to find, answer and hand over every case.
Chats and e-mails share one queue. Agents pick up cases themselves, and every action is visible to the team.
Conversations
- Filter by channel, status, agent, date, category or brand
- Search by ticket number or name
- Claim, join, take over, release or transfer with a note
- Merge duplicate tickets
Replying
- Quick replies: type "/" to insert one
- Saved links, by brand
- Suggest a reply with AI
- Translation into your language and theirs
- Reply by chat or e-mail
Context at hand
- Customer panel with past conversations
- Internal notes only staff can see
- Tags and categories
- Badges for frustrated or waiting players
Tickets view
Every case by state: Ongoing, Unattended, Went quiet, Escalated, On hold, Missed, Solved and Closed. Filter by VIP tier, brand and assignee.
Escalations
A queue ordered by urgency. Reasons include distress, responsible-gambling review, player request, agent not replying and AI unsure. Escalations can go to a team.
Macros
One click to tag, set status, assign, send a message, add a note and set a category. Personal or shared.
Knowledge library
A shared team library by collection and brand, private "My notes" for each agent, and a review tab where suggested notes and AI drafts wait for approval.
Home for team leads
What needs attention now: unclaimed chats, overloaded agents and response targets at risk, plus today's volume and the week's trend.
Alerts that reach people
Desktop notifications and a sound for unclaimed chats, coordinated across open tabs so it rings once.
Built-in guards: messages are delivered once even after a dropped connection, and a filter blocks slurs and threats in agent replies.
Case lifecycle
Clear rules for when a case waits, ends or comes back.
Each case has one ticket number and one history, whether it is answered by the AI or a person, by chat or e-mail.
- 0 min
Opened
Ticket number, category and response target assigned.
- 5 min
Agent silent
A player left waiting raises a high-priority escalation, repeated every 15 minutes.
- 10 min
Player silent
The chat ends with a countdown shown to the player. On-hold chats are exempt.
- 1 h
Reopen window
The player can continue the same case for an hour after it ends.
- 24 h
Auto-close
Inactive cases close, except those with an open harm escalation.
- Wrap-up
Agent closes
A chat the player ended stays with the agent until they resolve it.
Statuses
The player sees each status change as a short line in their own language.
Routing
By default agents pick cases from a shared queue. Automatic routing to the least-busy agent is available as an option. Chats nobody ever answered are marked as missed, so none disappear quietly.
Team
Shifts, live activity and access, managed in the console.
Shifts and breaks
Agents set their status: on shift, tea break, lunch, meeting or end of shift. A shift ends by itself if the console is closed for 10 minutes, so hours stay accurate.
Agent monitor
See in real time who is typing, viewing a chat, idle, on a break or away. Leads can open the chat an agent is viewing.
Response targets by VIP tier
Set first-response and resolution targets per tier. Players get their tier automatically, and a dashboard shows compliance, breaches and overdue cases.
Roles
Owner, admin, supervisor and agent. A person can hold several roles. Billing is visible to owners only.
Aliases
Agents can show players an alias instead of their real name.
Opening hours
Time zone, weekly schedule including overnight ranges, and the offline message per language.
Quality and reports
See how each agent and the whole team are doing.
Performance, per agent
Conversations handled and resolved, resolution rate, response time, handling and resolution time split by chat and e-mail, satisfaction, rating response rate, quality score, escalations and harm flags.
Productivity
Time logged in and on shift, breaks, lunch and meetings, effective working time and days worked, with a day-by-day download.
Quality scoring
Leads score resolved chats from 0 to 100 on accuracy, communication, empathy, resolution and speed, with feedback. Agents can dispute a score. A leaderboard and an unscored queue keep reviews moving.
Satisfaction and exports
Average rating, the 1–5 spread and per-agent results. Conversation, agent and daily reports download as CSV or JSON. Agents see only their own numbers.
Brands
A casino and a sportsbook in one account.
Run several brands with one team. Juno recognises the brand from the website the player is on, or from your widget setup.
Separate knowledge
Knowledge bases for one brand or for all, so the AI never answers a casino question with sportsbook rules.
Brand on every ticket
Filter tickets by brand, choose the brand on outbound e-mail, and keep saved links per brand.
Your own console address
Your team signs in at your own subdomain, with your logo and colours on the console.
The widget's look and texts are set once for the account and shared by its brands.
Data and security
A support tool that can't touch the money.
Juno has no connection to your player database and no access to wallets, balances, deposits or withdrawals. It holds only what support needs, which makes it easier for your compliance team to approve.
GDPR export and erasure
Export everything held for a player. Erasure removes personal data from chats, e-mails and notes, and reports exactly what was removed and what is kept.
Retention you set
Choose how long conversations are kept, 365 days by default, with optional automatic deletion of closed cases.
Audit log
Sign-ins and failed sign-ins, assignments, merges, user changes, exports, erasures, article publishing and quality actions.
Protected access
Sign-in throttling, sessions ended when an e-mail address changes, signed visitor identities and a list of sites allowed to load your widget.
Credentials and connections
Stored credentials are encrypted. Every incoming connection is verified, and unverified requests are refused.
Isolation and storage
Each account's data is kept apart in the database. Attachments have a 1 GB allowance per account, and error reports are stripped of personal data.
Managed service
Run by Perzyst, with you in control.
Perzyst sets up and operates your account. Our access to it is limited, time-boxed and logged.
Onboarding
We create your account, brands, widget look and languages, and send your owner a one-time activation link.
Support visits on the record
When our engineers need to look inside your account, the visit needs a stated reason, ends after 60 minutes and is written to an audit trail.
Key changes without downtime
When your widget key is replaced, the old key keeps working for a grace period while you update your site.
Billing you can see
Owners see the plan, this month's usage, invoices and amounts due, and can request more seats, a new brand or a plan change.
A named contact
Each account has a person at Perzyst assigned to look after it.
Engineers who built it
Issues go straight to the people who wrote the product.
At a glance
Languages and channels.
| Languages | Count | Which |
|---|---|---|
| Conversation languages you can offer | 57 | English, French, Dutch and German on by default |
| Agent translation | 57 | Both directions, with an untranslated flag |
| Widget buttons and labels | 4 | EN, FR, NL, DE; others show English |
| Harm phrase library and caring reply | 4 | EN, FR, NL, DE |
| E-mail auto-reply and offline message | 4 | EN, FR, NL, DE, or your own text |
| Channel | Status |
|---|---|
| Website and in-app chat | Live |
| E-mail, in and out | Live |
| Public help centre | Live |
| Planned | |
| SMS, social and voice | Not offered |
Book a demo
See Juno on a real conversation.
We'll set up a test brand with your colours and a few of your help articles, walk your team through the console, and show the harm-detection flow end to end.
info@perzyst.comTell us your brands, languages, team size and current support tool.
Email us about a demo