
LoyJoy Platform
Build AI Agents that actually resolve customer requests.
Your business team builds, tests and runs agents for chat and phone, visually in the LoyJoy Manager or in conversation with its AI assistant. Independently or supported by our delivery team, live in 2 to 4 weeks. IT keeps governance, permissions and integrations.
Why Agentic AI now?
Companies have identified Agentic Customer Service as a strategic priority. Yet rule-based chatbots fail on complex processes, LLM wrappers hallucinate, and do-it-yourself builds with cloud toolkits deliver demos instead of end-to-end, auditable processes. LoyJoy closes this gap: our AI Agents connect AI securely with your knowledge and your back-end systems. Customer requests turn into completed tasks, traceable and measurable.
The 3-layer platform for Agentic AI
What is LoyJoy?
- AI Agents (Act)
Generative AI answers questions, runs tools and works with your enterprise data, on the AI model of your choice.
- Process Orchestration (Orchestrate)
A no-code BPMN engine models dialogs and processes, with rules, routing, experiments, and analytics.
- Integration Layer (Connect)
Any MCP server and native connectors link CRMs, DWHs, PIM, payment, and custom REST/GraphQL services.



Build
Visually or in conversation with your AI assistant
Your business team builds agents itself and keeps control of knowledge, processes and models.
- AI Agents & Knowledge
Curated sources (PDF/DOCX/XLSX/PPTX, websites, knowledge bases), guardrails and prompts you control. Plus direct access to complete documents such as insurance policy terms.
- No-Code BPMN 2.0
Design dialog flows and real processes: conditions, loops, webhooks, human-in-the-loop.
- LoyJoy Headless
Your business team builds, optimizes and monitors AI Agents directly from Claude, ChatGPT, Microsoft Copilot or other AI assistants that support MCP. Included in all plans.
- Latest AI Models
Gemma 4 on our own hardware by default. Plus always access to the latest AI models, from GPT and Claude to open-source models like GLM and Kimi. Selectable per tenant, no lock-in to a single model.
- Smart Product Advisor
GPT Recommender builds intelligent product advisors without programming, with follow-up questions, product suggestions, availability, and add-to-cart.
- Avatar Integration
HeyGen, D-ID, and Moonscale avatars integrate seamlessly into LoyJoy agents, for personalized and engaging customer experiences.





Channels
One agent for chat, phone and your customers' AI agents
Build once: the same agent works by text, voice and phone, and on your website it also serves your customers' AI assistants.
- AI Phone Agents (Voice)
AI agents on the phone that resolve requests end-to-end: low latency, natural voices, barge-in, and SIP telephony connection. Same agent as in chat, 40 languages, seamless handover to live agents.
- Voice in chat
With one click, customers switch to voice mode in the web chat and talk to the agent like on the phone, without calling. Same agent, same knowledge, same processes.
- Multi-Channel Deployment
Build once, deploy everywhere: web chat, WhatsApp and messengers. The same agent in every channel, responsive, accessible and screen reader compatible.
- Live-Chat & Handover (async)
When a request needs human empathy or expertise, the AI agent hands over to a live agent with full context. Customers are notified and can continue later.
- MCP: Client & Server
Every LoyJoy agent uses external MCP servers as tools and is one itself. Connect systems without code, and let your customers' AI agents use your services directly.
- WebMCP: your website for AI agents
External AI assistants such as ChatGPT discover approved services on your website, such as tariff recommendations or appointment booking, and call them directly. Curated and approved by you.
Coming soon






Test and improve
Quality you can measure
Test before go-live, improve continuously in operation.
- AI Testing Automation
Upload a question catalog, including follow-up questions and multiple languages, and check every answer automatically. Ideal before each go-live and after model or knowledge changes.
- AI Quality Check
AI reviews your dialogs and highlights optimization potential and missing data, for better deployment quality.
Enterprise and above- AI analysis of conversations
Conversations are summarised automatically and rated by sentiment, topics are detected and made filterable. See what customers ask about and where content is missing.
Enterprise and above



Enterprise
Governance, security and control
For organizations where many teams work on the customer interface.
- Versioning & Control
Staging and live operation are separated. Changes are approved and published, and older versions can be restored at any time.
- Enterprise Role & Rights Management
Fine-grained user permissions and roles for large organizations, full control over access and authorizations.
- Privacy, Security & Compliance
EU hosting and features designed to support GDPR, the EU AI Act, DORA and accessibility law. The chat LLM Gemma 4 runs on our own hardware in Germany, telephony in its own EU Data Zone. Encryption in transit and at rest.
- OAuth Authentication in Chat
Secure customer identification with OAuth providers such as Microsoft Entra ID or Keycloak. The agent then works with customer-specific data, or you restrict access to registered users.
- Multi-Language & Localization
40 languages available including German, English, Spanish, French, Chinese, and Arabic (RTL), global rollouts and local adaptations.





Process Engine
Customer-facing AI Agents, built without coding
The LoyJoy Platform connects agentic AI and process automation at the customer interface. With over 100 business process modules, processes are realized. AI can execute these autonomously, within defined guardrails.
Excerpt from the 100 process modules:
Appointment booking, PDF creation, NPS, questionnaires, surveys, product gallery, photo upload with AI image recognition, scanner (e.g., barcodes, QR codes), opt-in collection, loyalty program, coupon codes, contests, quizzes, Google Maps, and many more.

AI Phone Agents
Voice service that actually resolves requests, not just routes them
The same AI Agent you build for chat also answers calls. It understands natural language, accesses your existing systems, and resolves requests end-to-end on the phone. When a human is needed, it hands over with full context to your live agents.
Natural conversation
Low latency, high-quality voices, and barge-in, customers can interrupt and are understood as in a real dialogue.
Your telephony, your data
Connect to your phone system (SIP) and existing systems via MCP and connectors. EU hosting and features designed to support your GDPR and AI Act requirements.
One agent, all channels
Phone, chat, WhatsApp, and messengers from one model. 40 languages, built once, available everywhere.

EU Data Sovereignty
AI models on our own hardware in our German data center
For our chat agents, we run the AI model Gemma 4 on our own hardware in our data center in Germany, not in a rented cloud. This removes external AI subprocessors from chat inference: no chat data is sent to external US-based model providers during this processing step. Based on the feedback we receive, customers rate Gemma 4 as comparable with current large language models from well-known providers for accuracy, speed and reliability. Other current models remain available by choice. For telephony, the speech-to-speech model additionally runs in its own EU Data Zone.

Templates & Use Cases
Ready-to-run templates
For Insurance, Banking/Finance, Energy & Utilities, B2B Manufacturing/MedTech, E-Commerce, and HR.
Phone Agent for Insurers
Claim intake, address change, status inquiry, and pre-qualification resolved end-to-end on the phone, with secure identification (OAuth) and handover to the right case handler.
Service Agent
Address change, appointment, claim/incident intake, ticket deflection with AI Quality Check.
Smart Product Advisor
GPT Recommender for intelligent product advice, customers describe needs, agent asks follow-up questions and suggests matching products. Optional with avatar integration.
HR & Careers
Candidate Q&A, role matching, and pre-qualification with automated test simulation.
Onboarding & CX Journeys
Welcome flows, ID capture, consent, and follow-up with complete version control.
Built for teams
Platform for Every Role
- Business & CX
Build agents without code, own the copy and optimize with A/B tests. AI Quality Check shows where to improve.
- IT & Security
Standardized integrations, identity, observability, SLAs, policies. Enterprise role & rights management for large organizations.
- Ops
Multi-tenant rollout, localization in 40 languages, change control, releases. Complete versioning with history and OAuth authentication for secure user management.



Outcomes you can measure
Measurable Business Impact
Case closure
Automated resolution rates and time-to-resolution. 98% automated handling at RTL+.
Deflection
Reduced call/chat volume; improved self-service share. 70% less email volume at 1. FC Köln.
Conversion
Qualified leads, applications or purchases, around the clock. TRUMPF qualifies leads worldwide, including on its US website.
CX
Satisfaction and NPS right after the conversation, resolved on first contact.
FAQ
Both, in the best sense. Classic chatbots follow fixed paths and only answer. LoyJoy AI Agents understand natural language, run processes on their own, access back-end systems and resolve requests end to end. You model processes with BPMN and connect external systems via MCP.
The default for chat agents is Gemma 4 on our own hardware in a data center in Germany. In addition, you choose other models per agent such as GPT, Claude or Mistral with processing in the EU, open-source models or your own LLM. Your knowledge base is independent of the model, so switching is a setting, not a rebuild.
The AI answers only from the sources you approve per agent, and guardrails allow only grounded statements. AI test automation checks your question catalog before every go-live. Probability-based models offer no 100 percent guarantee, but deviations are rare in practice.
The first agent typically goes live within 2 to 4 weeks, and with templates a first version is often ready in 15 minutes. Integrations with back-end systems and the role and permission concept take the most time. Your business team builds on its own or supported by our delivery team with project leads and Forward Deployed AI Engineers, directly from the vendor.
When a request needs human empathy or expertise, the AI agent hands over to LoyJoy Live Chat with full context. Customers are notified and can continue asynchronously. You decide when the handover happens.
With EU hosting, privacy by design, configurable guardrails, roles and permissions, audit logs and encryption in transit and at rest. Which requirements apply depends on the specific use case and configuration. Evidence is available in our Trust Center.
Yes. With LoyJoy Headless you build, optimize and monitor AI Agents from Claude, ChatGPT, Microsoft Copilot or other AI assistants that support MCP. LoyJoy keeps managing processes, permissions and publishing. Included in all plans. More about LoyJoy Headless
In 30 minutes, we will show you which customer requests your agents can resolve starting next month.
Free demo with concrete examples from your industry.