Take AI out of the black box
You can see why the avatar reacted the way it did — which conversation phase was active, which rule applied, and what changed in the conversation. No guesswork.
Whitebox Creator
AI conversation simulation where you can see why the avatar reacts that way.
AI simulations for soft skills training
Immersive conversational exercises with AI avatars for sales, HR, customer service & leadership.
The Whitebox Creator is an AI-powered authoring tool: You describe a conversation scenario, and from that an avatar is created that your employees can practice with.
The avatars respond to what you say and how you phrase it: content, choice of words, conversation management, and argumentation. And they have a mood of their own — patience, skepticism, and trust change over the course of the conversation.
Traditional courses convey knowledge. The Whitebox Creator trains what comes after that: the conversation.
The result: Real conversation experiences instead of passive learning.
Why it is called “Whitebox”
No black box: you can see and control how the avatar thinks
Correct the avatar in the middle of the conversation
You describe the conversation scenario — that’s how the avatar is created. If it doesn’t behave as desired, you tell it so in the middle of the exercise: “You conceded on price too quickly.” The correction is applied, recorded, and remains revocable at any time.
Your knowledge, not general internet knowledge
Upload your discussion guides, product documents, or training videos.
The avatar then argues using your content instead of general knowledge from the internet.
Applications
Examples from practice
The Whitebox Creator shines wherever conversations determine the outcome.
If there is a recurring conversational situation that can be practiced, it can be recreated — the following four are examples from real-world practice:
Sales
“Too expensive!”
Practice until you can deliver the answer without offering a discount.
Handling objections · Needs analysis · Closing confidence
HR
The employee or termination meeting that you can’t conduct twice.
Separation meetings · Conflict moderation · Feedback meetings
Service
The call that starts with anger.
De-escalation · Complaint management · Active listening
Leadership
The change decision the team doesn't want.
visionary to concrete & always motivating · critical conversations · negotiation management
Your use case?
Onboarding & training · complaint and claims conversations · job interview and selection conversations · citizen contact in public administration · consulting in retail · informational and family conversations in healthcare · change communication · compliance and audit conversations — and your situation?
Three ways to build conversational skills
Only one combines realistic conversation scenarios with scalability, repeatability, and didactic traceability.
Classic eLearning is scalable and available at any time – but usually remains passive.
Anyone who wants to develop conversational skills needs to hold conversations, make decisions, and respond to different reactions.
In-person role plays come closest to reality. They enable genuine interaction and immediate feedback, but are complex, expensive, and can only be repeated to a limited extent.
General AI chatbots do make conversations possible at any time, but as a rule they do not follow a defined instructional concept. Why the chatbot responds in a certain way often remains unclear to learners, and learning progress is difficult to track.
The Whitebox Creator combines both: the realism of a role play with the scalability of eLearning. Learners can conduct conversations as often as they like, try out different situations, and receive understandable feedback – including an explanation of why the avatar responded that way.
| What matters | Role-playing in face-to-face training | General AI chatbote.g. ChatGPT role-playing | Whitebox CreatorAI conversation simulation from CREATE |
|---|---|---|---|
| Body language, eye contact, and spatial presence | available | not intended | not intended |
| Speak freely instead of clicking — responses to word choice and tone of voice | available | limited or dependent on the setup | available |
| Conversation dynamics: Patience, skepticism, and trust change over time | available | limited or dependent on the setup | available |
| Available at any time, repeatable as often as desired | not intended | available | available |
| Scalable across all locations without travel or trainer costs | not intended | available | available |
| Behavior is regulated, assessable, and the same for all learners | limited or dependent on the setup | not intended | available |
| Understandable why the avatar reacted that way | limited or dependent on the setup | not intended | available |
| Create scenarios yourself — without prompt know-how, without an agency | limited or dependent on the setup | limited or dependent on the setup | available |
| The avatar only argues using the content you have approved | limited or dependent on the setup | not intended | available |
| Feedback on every conversational turn with specific alternative wording | limited or dependent on the setup | not intended | available |
| Evaluation aligned with your learning objectives — visible to learners and trainers | not intended | not intended | available |
| EU model operation, your conversations do not train any AI models | limited or dependent on the setup | available |
available limited or dependent on the setup not intended
What distinguishes the Whitebox Creator from an AI chatbot?
The avatar’s behavior is not a single prompt, but a structured model: persona, conversation phases, rules, states, and transitions. A control system reliably determines what is permitted in each phase — the AI only formulates the response within this framework. This keeps the behavior stable, verifiable, and repeatable.
What applications is the tool suitable for?
For any recurring conversational situation that can be practiced. Common examples are sales, HR, customer service, and leadership — but onboarding, retail consulting, citizen contact, or complaint conversations work just as well. There is no fixed list: the scenario is created from your description.
Is its use GDPR-compliant – and where is the data stored?
The application runs on our own infrastructure: simulations, conversation histories, evaluations, and uploaded documents are stored in our database. For the AI services — speech recognition, speech synthesis, and the avatar’s responses — we work with contracted service providers; the models for conversation management are operated within the EU. Your conversation content is not used to train AI models.
Do I need experience in prompting to create AI training courses?
No. The editor guides you through the creation process. You describe the situation, role, and learning objectives, the tool asks follow-up questions and generates a finished, immediately playable avatar from them. A first functional draft is ready in just a few minutes — after that, you refine it by speaking with it yourself. You can use advanced parameters, but it is not necessary.
What if the avatar doesn't behave the way I want it to?
Then you tell it to him — even in the middle of the authoring session. “You conceded too quickly on price” becomes a structured, versioned change that you review as a before/after and approve. No prompt tinkering, no invisible changes.
Can I contribute my own content — product information, guides, policies?
Yes. Documents and videos are uploaded, facts are extracted from them, which you review and approve. Only approved content is incorporated into the simulation — so the avatar speaks about your products and your standards.
Can the learners' progress be viewed?
Yes. After each simulation, there is an automatic evaluation: feedback on each conversational turn, specific alternative wording, and an assessment for each learning objective — visible to learners and trainers.
Any questions left?
Tell us about your use case — and we’ll tell you honestly whether the Whitebox Creator is the right tool for it.