Merkur Xtip: AI support for the casino floor, website and voice.
Three products, one platform. A chat and voice support system for website and app, a machine troubleshooting agent that guides floor staff through manufacturer-approved procedures, and a voice support agent that expands customer support coverage.
We gave Merkur Xtip’s casino floor an AI that helps fix machines — and a voice agent that never stops answering.
Merkur Xtip needed more than a customer-support chatbot. For its Serbian operations, the company needed a system that could support players across chat and voice, help floor staff troubleshoot gaming machines in real time, and keep sensitive documentation and player data inside the client environment. We built all three products on one Growww AI platform and grounded the troubleshooting layer in manufacturer technical documentation.
Merkur Xtip
Large German Casino Operator
Serbia · Phase 1
Chat + Voice · Troubleshooting · Voice Support
When a machine goes down on a casino floor, every minute costs money.
A casino floor is a high-pressure environment. When a slot machine or gaming terminal malfunctions, floor staff need to react quickly. Every minute a machine is down is a minute it is not generating revenue.
The problem: floor staff are not technicians. Manufacturer documentation contains the correct repair procedures and error-code references, but that knowledge usually lives in manuals, PDFs and training materials - not at the exact moment it is needed on the floor.
On the customer-support side, players need answers across website, app and phone, often outside the practical capacity of a human-only team. At the same time, casino-industry data handling, call recording and player information require a deployment model with far more control than a typical SaaS tool.
Three products. One platform. Full control.
We built three products for Merkur Xtip on the Growww AI platform, deployed on the client’s infrastructure. The core differentiator is the shared RAG pipeline: manufacturer technical documentation - including repair procedures, error-code references and diagnostic workflows - is indexed so the AI can retrieve the right procedure in real time.
That means a floor worker can describe the problem or enter an error code and receive step-by-step guidance. If the issue can be resolved on the floor, the machine returns to service quickly. If it requires a technician, the system creates a ticket for follow-up.
Website chat · Mobile app chat · Voice support line · Floor staff troubleshooting interface
Chat + Voice Support Agent · Machine Troubleshooting Agent · Voice Support Agent - all running on the Growww AI engine
Manufacturer technical documentation, repair procedures, error-code references and diagnostic workflows - indexed and retrievable in real time
Deployed on Merkur Xtip’s infrastructure. Full data control. Player data and call recordings remain inside the client environment.
What we built.
Chat & Voice Support
A unified support layer across Merkur Xtip’s website and mobile app. Players can ask questions through text or voice, while answers stay grounded in Merkur Xtip’s knowledge base and aligned with service standards.
- Text chat on website and app
- Knowledge-base grounded answers
- Handles inquiries at scale
- Voice support integration
- Brand-aligned responses
- Reduces load on human support
Machine Troubleshooting Agent
Manufacturer technical documentation is transformed into an on-demand troubleshooting assistant. The agent guides floor staff through diagnosis and repair, and escalates to a technician when the documented procedure requires it.
- Manufacturer documentation in RAG
- Error-code lookup and resolution
- Automatic technician ticket creation
- Step-by-step diagnostics
- Repair guidance for floor staff
- Designed to reduce machine downtime
Call Center Performance Agent
This agent watches every call. It compares setter and closer conversations against predefined scripts and the BANT qualification framework. It checks whether reps are following instructions, sticking to the script, qualifying properly, and maintaining the right communication tone and emotional register. The result is a systematic, data-driven view of call quality that was previously impossible to track at scale.
- Monitors adherence to predefined scripts
- Tracks BANT qualification compliance
- Analyzes communication tone and emotion
- Identifies reps needing coaching
- Flags script sections that aren't working
- Data-driven call quality at scale
Voice Support Agent
A dedicated voice agent for inbound customer-support calls. It understands the inquiry, retrieves an accurate answer from the knowledge base, and routes complex issues when human handling is required.
- AI voice agent for inbound support
- Grounded responses
- Routes complex issues
- Handles high call volumes
- Extends support coverage
- Consistent service quality
How the machine repair agent works.
A slot machine or gaming terminal stops working or throws an error code.
The worker describes the issue or enters the error code.
The RAG layer retrieves the relevant manufacturer diagnostic procedure.
The worker follows the documented process step by step.
If the documented fix works, the machine can return to service quickly.
If specialist intervention is needed, the system creates a follow-up ticket.
One workflow. No guessing.
Manufacturer documentation becomes real-time guidance, while unresolved issues automatically move into a clear technician escalation path.
The AI system gave our team instant access to the knowledge they need, whenever they need it. Issues can now be identified and resolved faster, while 24/7 support helps us keep operations moving without unnecessary delays.
CTO
How we deployed it.
Operations audit & knowledge mapping
Mapped customer-support workflows, floor operations, machine types and manufacturer documentation sources. Identified the three products with the highest potential impact for Serbian operations.
Manufacturer technical docs into RAG
Majstori Grada's team uploaded their services, repair methods, pricing, terms, conditions, and sales scripts into the shared knowledge base. Structured for retrieval across all four agents.
Four agents built & tested
Lead Gen Chat, After-Hours Call Center, Knowledge Agent, and QA Agent — each built, tested against real scenarios, and tuned for accuracy before deployment.
Website, telephony & sales interface
Chat widget deployed on website. Voice agent integrated with telephony for after-hours routing. Knowledge agent integrated into sales team interface. QA dashboard connected to call recording system.
Full system live with ongoing support
All four agents went live. Ongoing support includes monitoring, agent tuning, and additional development as Majstori Grada's needs evolve.
Not a chatbot. Not a call center service. A system.
Most companies that sell AI tools sell one piece. A chatbot vendor gives you a chatbot. A call center service gives you call center coverage. A sales enablement tool gives you a knowledge base. None of them talk to each other.
What we built for Majstori Grada is a system where every agent shares the same knowledge base, the same infrastructure, and the same data pipeline. The lead chat knows what the call center knows. The knowledge agent feeds the QA agent. The QA agent's insights can improve the scripts that the lead chat and call center use. It compounds.
This only works because it's PaaS. The platform runs on Majstori Grada's infrastructure. Their call recordings, their lead data, their sales scripts — all of it stays inside their environment. A SaaS chatbot vendor can't offer that. A third-party call center can't offer that. This is the PaaS difference.
Want a complete AI system for your sales operation?
We build multi-agent systems that cover your entire lead-to-close pipeline. Lead capture, after-hours coverage, sales knowledge, and call quality — all on your infrastructure.
