Recent advances in AI allow for realistic communication between a person and a computer-implemented model via a computer-implemented application. Due to the growing need for effective and efficient meeting outcomes, the idea has arisen to use such an application to manage meetings (managing a meeting is also called "leading" or "moderating"). The technical challenge here is that such applications currently assume a person is communicating with the model, whereas meetings naturally involve at least two people (referred to here as "participants" or "TPs"). Having both TPs share a single, simple
user interface system, as currently exists, is not a viable solution, because the model would then be unable to distinguish who made which statement.Nor is it a solution to have each participant use their own
user interface, and for these interfaces to be completely independent of each other. Furthermore, a meeting cannot be controlled by a single response from a model, as a meeting is always a longer communication process between the participants. This requires a technical method for transmitting instructions for the specific implementation of the leadership / moderation (including control if the meeting has multiple objectives and / or topics) and for providing the model with
relevant information from the meeting's progress so far. All of this is technically solved by a computer-implemented
system product with inherent methods as described below: For all participants in a meeting (6), an environment is set up within a (computer-implemented) communication
system (1). The participants in this meeting can perceive what each participant says in the meeting. At the same time, this provided environment is separated from other environments to ensure
confidentiality. Within the provided environment, a computer-implemented "bot" (2) controls the meeting. (It could also be called a "virtual assistant," "virtual moderator," or similar.) This bot transmits all predefined instructions for leading the meeting (3), the first
utterance of a participant, and an identifier for that participant (!) to a model (4) (e.g., a pre-trained Large
Language Model), which then responds accordingly. The bot outputs this response in the communication system and stores the original
utterance of the participant and the model's response in a computer-implemented storage system (5). With each subsequent
utterance from a participant, the bot retrieves the previous utterances and responses from the storage system and also transmits them to the model. If basic communication rules are not followed by TPs, the bot can react in a specific way. Additionally, a separator is used if the TPs in a meeting use a shared
user input channel. For a successful technical system product, it is also important that two environments deployed in (1) during the same period have identical properties. Therefore, such environments are created automatically after a request (7). If desired, each deployed environment can later be used for further meetings. The invention described here achieves the overcoming of the aforementioned technical problems, enables meetings to produce more effective and efficient results with less resource expenditure, and offers further advantages.