Artificial intelligence in high fidelity
All trained documents go through an intensive quality assurance process. This ensures that the answers provided match the contents of the source documents. We call this AI HiFi.
AI custom knowledge bots must be trustworthy
It's not just about an AI bot finding an answer and, if possible, outputting it in natural language. The user wants to be sure that the answer corresponds to the truth, according to the knowledge acquired. This is the only way he can develop trust in the application.
AI high fidelity is the goal
We call this AI-High-Fidelity, AI-HiFi for short, meaning “high fidelity”. To achieve this, we have developed a multi-stage quality assurance process that monitors the quality of the current system status. We can run this backup process for every change in the configuration, every change in the trained knowledge base or every change in the software version. This is how we come to an overall statement about the system quality.
Measuring system quality
Using a complex testing process, we can determine the quality of the entire system and the quality of the underlying levels. (We would be happy to explain details in a personal conversation.)
This allows us to:
- monitoring the system quality with every change to the system
- the representation of the current system quality on the user side
- the constant improvement of the automated training steps
Importance for the user
- Trustworthiness assessment (“How much do I believe what the AI bot just answered me?”)
- Transparency about the current performance of the AI bot
- Assess the need to consult additional sources
