CHECK24 Sophie: Build Your Own AI Customer Service Agent

CHECK24’s digital assistant Sophie answers customer questions around the clock, takes personal cases into account after sign-in, accepts documents, and connects to customer advisors when needed.

With the LoyJoy Platform, companies apply this same principle to their own customer service: with their own knowledge, their own customer data, their own service processes, and full control over handovers and permissions.

AI Agents from LoyJoy do not just answer questions. They capture data, check requirements, call backend systems, and carry customer requests through to a defined outcome, by chat and phone.

See an AI Customer Service Agent live · Assess your automation potential

What CHECK24 Sophie really is

Sophie appears on numerous CHECK24 pages as a shared digital assistant. The publicly documented functions, however, show different domain-specific implementations, for example for hotels, internet contracts, and tires and car parts.

Sophie is best understood as a shared interface for several specialized service assistants:

1. A single name and entry point for customers
2. Domain knowledge for each product and service area
3. Access to personal cases after sign-in
4. Structured capture of requests and documents
5. Execution or routing of defined service processes
6. Handover to the appropriate customer advisor

This model is more relevant for companies than the idea of a single universal chatbot. Complex customer service comes from specialized capabilities, data access, and reliable process control.

What Sophie demonstrably does

1. Help around the clock

CHECK24 describes Sophie as an instantly available digital assistant. When a request cannot be resolved automatically, she connects the customer with customer advisors or offers a further point of contact.

2. Answer domain-specific questions

The hotel assistant, for example, helps with questions about bookings, payment, cancellation terms, rebooking, and issues during a stay. The Sophie variant for tires and car parts helps with orders, delivery times, invoices, returns, and complaints.

3. Incorporate personal cases

Once a customer is signed in, CHECK24 can tie the chat to a specific comparison, contract, or provider switch. Status information and contract-related communication then sit within the context of that personal case.

4. Prepare callbacks and handovers

Sophie can capture a phone number and the specific request. During a handover, the customer advisory team receives the prior context so the customer does not have to repeat their problem in full.

5. Bring documents into a process

For the internet cashback, the customer uploads the invoice directly through Sophie. The document is checked automatically, and the customer receives immediate feedback. On successful verification, payout to the bank account is initiated.

6. Capture service requests in a structured way

Billing changes, complaints, returns, and issues with partners can be captured through Sophie and handed over to the responsible team.

7. Involve human decisions

Not every request is decided autonomously. In various cases, Sophie first gathers the necessary information. The actual review or approval is then carried out by customer service or an external contract partner.

An AI Customer Service Agent is more than a FAQ chatbot

A classic FAQ chatbot answers recurring questions. An AI Customer Service Agent connects five layers:

1. Understanding: What does the customer actually want to achieve?
2. Knowledge: Which product, contract, or service information applies?
3. Context: Which customer, contract, order, or case is this about?
4. Action: Which data needs to be captured and which systems called?
5. Escalation: When is an employee, department, or partner required?

Only by connecting these layers can high automation and a seamless customer experience be achieved.

How LoyJoy implements the Sophie principle

Sophie capability	Implementation with LoyJoy
Single entry point	Home view or a central agent as the entry point to specialized services
Domain specialization	Dedicated agents and processes for products, contracts, orders, and service areas
Answering questions	RAG over websites, PDFs, DOCX, PPTX, knowledge articles, and API data
Identifying customers	One-time code by email or integration with Auth0, Keycloak, and existing identity systems
Loading personal cases	API access to CRM, ERP, ticketing, contract, or order systems
Capturing data in a structured way	Questionnaires, forms, addresses, IBAN, free text, and selection fields
Accepting documents	File and image upload directly in the dialogue
Reviewing documents	Handover to OCR, a vision model, or the customer’s own review service
Executing a service action	GET, POST, or PUT calls to the responsible backend system
Creating a ticket or case	Integration with Salesforce Service Cloud, CRM, or a ticketing system
Preparing a callback	Capturing and handing over phone number, request, and context
Live chat	Handover to available service staff within LoyJoy
Domain-specific escalation	Routing by topic, priority, language, or customer group
Service by phone	Same agent and knowledge logic as the LoyJoy Phone Agent

One central agent, or several specialized agents?

For a larger company, a single agent for every topic is rarely the best solution. Different service areas have different data, risks, permissions, and definitions of success.

The better architecture consists of:

1. A central entry point that recognizes requests and orients customers
2. Specialized agents for clearly scoped tasks
3. Shared knowledge and governance rules
4. Separate permissions for sensitive customer data and actions
5. Unified analytics and quality assurance across all agents

An insurer, for example, can run separate agents for product questions, claims, contract service, and broker search. The experience still stays consistent for the customer.

Example process: Reviewing an invoice or proof document

A document process comparable to Sophie’s can look like this with LoyJoy:

1. The customer signs in or identifies their case.
2. The agent checks whether the claim is generally valid.
3. The customer uploads an invoice, photo, or proof document.
4. LoyJoy hands the document to a review service.
5. The review service returns a structured result and extracted data.
6. On successful verification, the process continues in the backend.
7. If uncertain, a case is created for customer service.
8. The customer receives a clear status message and, where applicable, a case number.

The final business or legal decision still rests with the designated backend system or the responsible employee.

Personal customer service after sign-in

Without identification, an agent can answer general questions. After secure sign-in, personal service becomes possible:

1. Displaying order and delivery status
2. Explaining contracts and booked services
3. Checking personal deadlines and requirements
4. Displaying existing tickets and cases
5. Capturing changes or applications in a structured way
6. Taking service history and existing data into account

LoyJoy supports email verification via one-time code. Alternatively, the agent can be integrated into the company’s existing login and permissions architecture.

Controlled automation instead of blind agent action

Not every customer input should automatically trigger a legally binding action. For certain travel processes, CHECK24 explicitly notes that change or cancellation requests are in some cases only forwarded and only take effect after confirmation by the respective contract partner.

LoyJoy therefore supports different levels of automation:

1. Information: The agent answers a question.
2. Pre-check: The agent checks data and requirements.
3. Preparation: The agent captures a complete task for processing.
4. Execution: A connected system executes an approved action.
5. Human approval: An employee reviews and decides a sensitive case.

Companies define, for each process, which level is appropriate on business, technical, and legal grounds.

RTL+: Up to 98 percent automated handling

RTL+ uses LoyJoy in customer service to offer users instant support outside of service hours and to automate high request volumes.

The AI Agent combines natural AI responses with controlled processes. Standard requests are handled directly in the dialogue. Only cases that require human review go to customer service.

RTL+ shows how a chatbot becomes a productive service channel with measurable automation. Comparable principles are also used by Vaillant, R+V, and 1. FC Cologne, each tailored to its own industry.

Read the RTL+ case study · See more customer service cases

Four stages to your own AI Customer Service Agent

Stage 1: Automate knowledge. The agent answers common questions from approved knowledge sources and links to relevant content.

Stage 2: Qualify requests. The agent recognizes topic and urgency, captures the necessary details, and routes to the right service area.

Stage 3: Integrate customer data and systems. After sign-in, personal cases are loaded. The agent creates tickets, updates data, or triggers defined processes.

Stage 4: Resolve requests end to end. The agent carries out clearly approved processes end to end and hands over only exceptions to customer service.

Frequently asked questions about the AI Customer Service Agent

What is CHECK24 Sophie?

Sophie is CHECK24’s digital assistant. She supports customers around the clock with questions and service requests and connects them with customer advisors when needed. Depending on the CHECK24 area, specialized functions and data are available.

Can you build a chatbot like CHECK24 Sophie yourself?

With LoyJoy, companies can build their own AI Customer Service Agent with comparable functional principles, including knowledge search, identification, customer data, documents, backend actions, and human handover.

What is the difference between an AI Customer Service Agent and a FAQ chatbot?

A FAQ chatbot mainly answers questions. An AI Customer Service Agent understands requests, takes personal context into account, captures data, and can execute concrete service processes through integrations.

Can LoyJoy display personal contract or order data?

Yes. After suitable authentication, LoyJoy can retrieve the required information via APIs from CRM, ERP, shop, contract, or ticketing systems.

Can customers upload documents and images?

Yes. LoyJoy supports file and image uploads within the dialogue. The content can then be handed over to OCR, vision, or the customer’s own review services.

Can the agent process changes or refunds?

Yes, if the responsible backend offers a suitable interface and the process is approved. For sensitive cases, LoyJoy can instead run a pre-check and obtain human approval.

What happens if the agent cannot resolve a request?

The agent can hand over to live chat, email, callback, ticketing, or a specialized department, passing along the topic, prior history, and information already captured.

Does the agent also work on the phone?

Yes. LoyJoy AI Agents can be deployed by chat and phone. Knowledge, rules, and backend functions can be used across channels.

Start automating customer service with LoyJoy

First identify the most frequent requests with high volume and a clear definition of success. LoyJoy then connects knowledge, process, and the required systems into a production AI Customer Service Agent.

Assess your automation potential · Request a personal demo

For a related read, see how companies build their own AI Shopping Assistant inspired by Amazon Alexa for Shopping.

Note: CHECK24 and Sophie are trademarks or designations of their respective owners. LoyJoy is not affiliated with CHECK24. This article is provided for factual comparison of different functional principles.

Sources:

1. CHECK24: Help and contact
2. CHECK24: Hotel service and digital hotel assistant Sophie
3. CHECK24: Internet cashback via Sophie
4. CHECK24: Sophie for tires and car parts
5. CHECK24: Data protection for chatbots, login, and contract communication
6. CHECK24: Terms and conditions of the travel planner
7. LoyJoy Docs: Retrieval Augmented Generation
8. LoyJoy Docs: API Client
9. LoyJoy Docs: Sign In
10. LoyJoy Docs: Data Collection and Upload
11. LoyJoy Docs: Live Chat