Nordvay Relay LLM
Not a rented, general-purpose API — our own large language model, developed and operated by Nordvay’s engineering team.
Nordvay Relay doesn’t call out to a third-party AI provider — it runs on our own AI model, hosted on Nordvay Cloud. Your data never leaves our infrastructure. It answers requests from WhatsApp, Microsoft Teams, web chat and other channels by connecting live to your ERP and CRM systems.
Many "AI-powered" products on the market simply call a third-party provider’s general-purpose model behind the scenes — your questions, prices and customer data leave your own infrastructure in the process. Nordvay Relay works differently: it runs on our own large language model, Nordvay Relay LLM, developed by Nordvay and hosted on Nordvay Cloud.
Not a rented, general-purpose API — our own large language model, developed and operated by Nordvay’s engineering team.
The model runs on Nordvay’s own private cloud infrastructure, or — if you prefer — within your own organization’s data center (on-premise). Either way, there’s no dependency on an external AI provider.
Your requests, pricing and customer data are processed only within Nordvay’s infrastructure — never sent to an external AI provider.
Data sovereignty is part of the design from day one, aligned with KVKK and GDPR requirements.
The model isn’t a static, generic chat engine — it continuously learns from operator corrections and your own business process.
The figures below come from our own test and production measurements.
This isn’t a performance claim — it’s about where your data goes and whose infrastructure it runs on.
| OpenAI (GPT) | Anthropic (Claude) | Google (Gemini) | Nordvay Relay LLM | |
|---|---|---|---|---|
| Model architecture | Provider-owned, general-purpose model | Provider-owned, general-purpose model | Provider-owned, general-purpose model | Our own model, built by Nordvay |
| Hosting | Microsoft Azure infrastructure | Amazon AWS / Google Cloud infrastructure | Google Cloud infrastructure | Nordvay Cloud or on-premise |
| Where data is processed | On the provider’s servers | On the provider’s servers | On the provider’s servers | Only within Nordvay’s infrastructure |
| Third-party data sharing | Subject to the provider’s terms of use | Subject to the provider’s terms of use | Subject to the provider’s terms of use | No data is shared with any third party |
| Customization to your process | General-purpose, prompt-driven | General-purpose, prompt-driven | General-purpose, prompt-driven | Continuously learns and adapts to your operation |
Your enterprise customers wait outside business hours, repeat the same questions to different people, and spend more time than they should to get a simple answer.
Phone queues and email response times get in the way of customers expecting a fast answer.
Every request that arrives outside business hours is either waiting or going to a competitor.
A significant share of team time goes to automatable questions — price, stock, order status.
Different operators can answer the same question differently; manual system lookups carry error risk.
Headcount needs grow with demand; sudden spikes degrade service quality.
Whichever channel the customer uses — the system queries your business processes, AI responds, and an operator steps in when needed.
Submits a request through their preferred messaging channel.
Drawing on the workflows and past scenarios it has learned, it understands the request and decides which process applies.
Live data is pulled from ERP, CRM and other systems in real time.
The response is written in your organization’s tone and natural language — it reads like a message from a person.
Every conversation is visible in the panel; an operator can step in at any time.
Once deployed, Nordvay Relay LLM learns the business systems, workflows and past scenarios of the organization it runs for — and starts behaving like that organization’s own AI model.
Over time, it comes to understand the data structure and business logic inside your ERP, CRM and other systems.
It learns organization-specific processes from operator corrections and past conversations.
Not a generic engine that gives everyone the same answer — a system shaped by your organization’s knowledge.
Customers get a natural exchange that reflects your organization’s tone, not a templated reply.
The difference between the two approaches at the same service volume.
| Traditional Support | Nordvay Relay | |
|---|---|---|
| Service hours | Limited to business hours | Uninterrupted, 24/7 |
| Response time | Minutes to hours | Within seconds |
| Concurrent capacity | One conversation per operator | Unlimited concurrent conversations |
| Cost structure | Headcount cost scales directly with volume | Fixed platform fee + low variable usage cost |
| Channels | Phone and email | WhatsApp, Teams, web and more |
| Data accuracy | Operator looks up systems manually | Live system connection, always current |
| Scaling | Headcount needs grow with demand | Grows without added cost |
| Consistency | Answer quality varies by operator | Same quality standard on every channel |
| Integration | Manual lookups, copy-paste | Live connection to ERP/CRM |
| Onboarding & training | Repeated training process for every new hire | Configured once, keeps itself current |
| Peak-period performance | Queues grow during campaigns or holidays | No performance loss as demand spikes |
| Freshness of information | Depends on training material, can lag behind | Reflects business-system changes instantly |
| Impact of staff turnover | Risk of losing institutional knowledge when an operator leaves | Learned process knowledge stays on the platform |
| Reporting & visibility | Manual or delayed reporting | Real-time operator panel and analytics dashboard |
| Multilingual support | Requires separate staff per language | Responds in multiple languages from the same model |
Every added channel extends your reach; all of them are managed from one platform and speak to every business system you run.
Communication Channels
Business Systems
Examples from different industries show how the same platform adapts to very different workflows, from logistics to healthcare, aviation and automotive. In a live deployment, responses are based on data pulled from your own business systems.
Account, history and agreement terms are recognized automatically.
Sales, support, finance and logistics run from a single conversation window.
Same quality of service, whichever channel the request comes from.
Order status lookup, stock check and creating a new order.
Issue reporting, automatic diagnosis and support ticket creation.
Transaction lookup, statement request and credit limit increase process.
Lab result lookup and doctor appointment booking.
Flight status lookup, delay notification and baggage tracking.
Maintenance booking, plate-based record lookup and automatic appointment creation.
Not just simple Q&A — it covers operational processes from customer service to finance, logistics and platform management, using business logic it has learned specifically for your organization.
Technical support, product information and request handling run automatically, with no business-hours limit — holidays and weekends included.
Status, owning team and estimated resolution time of a support request can be queried instantly — no need to call back.
An operator can take over at any time with full context preserved — the customer never notices the handoff.
Responses are generated in your organization’s tone, in natural language that doesn’t feel templated.
Live stock, per-warehouse quantity and customer-specific pricing/discounts delivered instantly.
Orders are recorded into your business system instantly with a unique tracking number and automatic confirmation.
Instant, secure access to past invoices, date-range filtering and order details.
Authorized access to account balance, overdue amounts and payment plan information.
Shipment stage, carrier information and estimated delivery date delivered instantly.
Return requests, process tracking and approval notifications, with automatic routing to the right team.
Conversations from every channel are monitored, filtered and prioritized in one screen.
Traffic, automation rate, performance and per-channel breakdown in one panel.
Automatic satisfaction survey after a request is closed; results roll up into the analytics panel.
Recognizes the distinct processes of different industries — from appointments to statements to fault tickets — from automotive to fintech.
We’re building real-time voice processing so customers experience a natural conversation over the voice channel.
In DevelopmentSpeech converts to text instantly, AI responds, and the reply is converted back to speech.
Tone, pacing and context tracking aim for a fluent experience.
Voice conversation history is being designed to sync with other channels.
Carries concurrent conversations without interruption; performance doesn’t degrade as your volume grows.
Nordvay Relay can be used as a fully managed service on Nordvay Cloud, or deployed within your own organization’s data center (on-premise). For on-premise deployments, Nordvay’s engineering team provides end-to-end advisory support.
The model and data layer run fully managed on Nordvay Cloud — the fastest way to get started.
Isolated, single-tenant resources running in a dedicated environment on Nordvay infrastructure.
Nordvay Relay LLM is hosted within your own organization’s data center; data never leaves your premises. Nordvay provides advisory support for setup and operations.
Can be configured with the AI engine on-premise while channel and integration layers run on Nordvay Cloud.
Varies with scope and integration complexity, but a standard rollout typically follows this shape.
Existing business processes are analyzed; channel and process priorities are set.
Connections to ERP and other systems are established; AI is configured for your organization.
Controlled testing, team training and fine-tuning.
Go-live with real customer traffic; close monitoring begins.
Customer data and business system access are protected with layered security controls.
Our security controls are designed around the access-control, audit-trail and incident-management principles found in international frameworks such as ISO/IEC 27001 (information security management) and SOC 2; our data processing is designed to align with KVKK and GDPR requirements. This is a design-principle statement, not a claim of formal certification.
There are many chatbot products on the market. Concrete reasons to choose Relay:
Not a rented third-party API — our own large language model, developed and operated by Nordvay.
Requests and commercial data are processed only on Nordvay Cloud — never sent to an external AI provider.
A platform that connects live to your business systems, not one built on templates.
Every messaging channel managed from a single operator panel.
Account balances, discounts, customer codes, lead times — B2B concepts built in.
Every conversation is recorded; an operator can take over at any moment.
Not a generic chat engine — it continuously learns from operator corrections and your own process.
Runs on Nordvay’s own private cloud infrastructure, within Turkey.
An engineering team that knows the Turkish market and local business systems.
No. Nordvay Relay runs on our own large language model, Nordvay Relay LLM, developed and hosted on Nordvay Cloud. Your requests and commercial data are never sent to an external AI provider.
It integrates with your ERP, CRM and other business systems through a standard API or database layer. Integration scope is planned together based on your existing architecture.
The system only responds based on data from connected business systems — it never guesses or assumes. If it can’t find an answer, it hands off to an operator; every response is visible in the panel in real time.
The system is configured with a tone specific to your organization. The goal is a service that differs from a traditional operator mainly by being faster and more consistent.
As AI handles repeat questions, your team shifts to higher-value work — the same team generally ends up handling a much higher volume of demand.
It depends on scope and integration complexity. See the "Rollout" section for a sample timeline.
Nordvay Relay runs on Nordvay’s own private cloud infrastructure for both the AI model and the data layer — or, if you prefer, within your own organization’s data center (on-premise). Either way, data is never sent to a third-party AI provider and is processed with KVKK requirements in mind.
Yes. Nordvay Relay LLM can be deployed within your own organization’s data center. Nordvay’s engineering team provides advisory support throughout setup, integration and operations.
The architecture is designed not to cap concurrent conversations at a fixed limit; capacity is planned around your needs.
It combines a fixed platform fee with usage-based variable cost; a clear cost projection is worked out together during the conversation.
The voice assistant module is in active development. A move to the voice channel can be planned once messaging channels are live.
Let’s evaluate together, in a demo environment connected to your business systems, how the AI performs against your real data.