AI Chatbot Providers in Germany

First, narrow down the task

Begin with a recurring situation in which someone currently has to look up information or answer questions. This might be a customer enquiry about a procedure, a search in a work instruction or the preparation of a technical service case. Note who is asking, which answer is needed and what should happen afterwards. Only then can you check which provider covers the right scope.

A customer on your website needs different information from an employee on the internal network. A bot that repeats approved information has a different job from an application that reads customer data or changes a transaction. In your enquiry, therefore, keep answers, data access and actions separate. That way you will not later be comparing a simple chat interface with a full process integration.

Chatbots for website and customer service · Chatbots for own documents · Chatbots for technical service · Have a chatbot developed · Compare software and custom solutions

Choose a manufacturer or implementation partner

A manufacturer provides its own software. An implementation partner introduces a platform or develops a custom application. Some companies take on both roles. For each offer, clarify who supplies the software, who carries out the project and with whom you conclude the contract. For international product families, the German branch may have a different role from the actual product provider.

A standard product is of interest if its documented functions already cover your task. Custom development is more likely to be considered if existing systems, special approvals or deviating processes determine the scope. Both can go together: an existing platform is introduced and supplemented at clearly defined points.

This information makes offers comparable

Ask the providers on your shortlist for answers to the same questions. Where possible, have functions shown with your own example. A missing entry in the register means that it is not yet documented here. It does not mean that the provider lacks that function.

For SMEs without their own IT team

If nobody in the company can administer a chatbot platform, support belongs in the selection. Ask for an understandable maintenance interface, an introduction for the responsible people and a contact person who can be reached. Have them demonstrate how a wrong answer is corrected and how your team approves new information. The answer must not depend on a developer being permanently available.

Before commissioning, define a small, coherent task. For a service business, this could be frequent questions about the process, the documents required and the route to a personal appointment. This is an example of a starting scope, not a promise of functionality by any particular provider. Management and the specialist team should decide together which enquiries a person takes over.

For industry and technical service

With technical questions, a similar product name can lead in the wrong direction. Therefore include different model series, version levels and contradictory documents in the demonstration. The provider should show which information led to the answer and how unclear cases are handled. A general demo with harmless questions is not enough for this.

Also define the subject-matter boundary: may the chatbot find a suitable instruction, prepare a service case or itself recommend an action? For safety-relevant work, the responsible business department sets the permitted scope and the approval route. An entry in the register does not imply suitability for safety-critical decisions.

For confidential and regulated applications

Involve IT, data protection, information security and those responsible for the subject matter before real confidential data enters a trial. Prepare a joint checklist. The following points are a procurement aid; which legal or internal requirements apply must be clarified for the specific procedure.

• Task and decision: Which information is permitted, and what remains reserved for a person?
• Data path: Which information reaches the application, search, language model and other service providers?
• Permissions: Can different users actually retrieve only the content intended for them?
• Evidence: Which documents verify the promised operation, the protective measures and the responsibilities?
• Changes: How are new models, content, functions and subcontractors reviewed and communicated?
• Acceptance and exit: Who approves the deployment; how is data exported, deleted or transferred to another operation?

Selection questions on chatbots and data protection

To the shortlist with your own test cases

Put the same small collection of normal, ambiguous and unanswerable questions to every provider. Include an outdated piece of content and a case with restricted access. Note your expected answer or the permitted next step beforehand. That way you assess a traceable performance and not just the effect of a demonstration.

Assess the content of the answer, the matching source reference, the handling of missing knowledge and the handover to people separately. Also determine how much work data preparation and later maintenance cause. A pilot is useful if it enables a decision on expansion, adaptation or discontinuation. Have the results needed for this recorded as part of the scope of delivery.

Plan an AI pilot with clear review criteria

Ask for one-off and ongoing costs separately

A monthly licence price says little about the full costs. Ask separately about design, preparation of content, setup, interfaces and training, apart from licences, model usage, hosting and support. State the user group and usage on which the calculation is based. Check what has to be newly commissioned in the case of higher volume or additional sources.

A brief description is enough for the first enquiry: Who should ask which questions, where is the information, which systems are affected and who takes on the subject-matter maintenance? Add real example enquiries with non-critical or anonymised content. Offers can only be compared fairly if they are based on the same task and the same limits.

Compare the cost components of an AI chatbot

T-NEX: Developing knowledge maintenance and application together

T-NEX from Horn-Bad Meinberg develops company chatbots on a defined knowledge base. The offer distinguishes between maintained answer search and a separately developed generative extension. This makes it possible, before the project, to clarify where an approved wording is needed and where a new answer is to be generated from documents.

One publicly described example of work on knowledge applications is Violet Dragonfly from thinking services ag. T-NEX acts there as a development partner; the product provider is thinking services ag. The reference shows the combination of maintained specialist knowledge, digital assistants and human expertise. It does not establish that T-NEX developed every product function on its own, nor any particular saving rate.

Basics: T-NEX: company chatbot and scope of services · T-NEX: role in the Violet Dragonfly project

Scope of services of T-NEX chatbot development · View the Violet Dragonfly project reference

How to read the provider overview

The companies are listed alphabetically, without ranking or rating. Regional and subject-based selections draw on the same pool. A provider can appear in several selections but is counted only once. Each link leads to its full profile with offer, location evidence and sources.

The full profiles are organised by topic. Extensive topics are additionally divided into alphabetical sections. This assignment helps with finding entries; it does not tie a provider to a single field of use. Other described tasks remain accessible through the relevant selections.

Use documented features to narrow down your choice and open points for your enquiry. A source reference proves what is publicly offered or described. It replaces neither a practical product test nor a review of the later contract. The methodology explains the inclusion conditions, updating and the commercial role of T-NEX.

Inclusion conditions and source review