Large language model (LLM) integration for scenario generation in a risk management platform

A data processing system adapted for scenario generation, the system comprising:

• a host computing platform comprising one or more computers, each with memory and one or more processing units including one or more processing cores;
• a business continuity management system executing in the host computing platform and comprising both a database of scenario description records for an organization and an asset database of the organization, and also a user interface providing access to the database of scenario description records and the asset database;
• and,
• a scenario generation module coupled to the business continuity management system and comprising computer program instructions enabled while executing in the memory of at least one of the processing units of the host computing platform to perform:
• establishing a communicative coupling over a computer communications network between the user interface of the business continuity management system and a selected large language model (LLM);
• specifying an adverse event in the user interface and querying the database of scenario description records with the specified adverse event in order to receive a vector of relevant data to the specified adverse event;
• tuning a pre-determined LLM prompt with the vector of relevant data to produce a specific LLM prompt for the specified adverse event;
• submitting the specific LLM prompt to the selected LLM from over the communicative coupling, the selected LLM returning to the user interface over the communicative coupling scenario text referring to a plausible instance of the specified adverse event;
• persisting the scenario text into persistent storage in connection with a specific scenario for use in a scenario simulation engine;
• retrieving the scenario text from the persistent storage;
• extracting from the persistent storage a vector of infrastructure asset characteristics;
• tuning an additional pre-determined LLM prompt with the vector of infrastructure asset characteristics to produce an additional specific LLM prompt for the specific scenario;
• submitting the additional specific LLM prompt to an additional LLM from over the communicative coupling, the additional LLM returning to the user interface over the communicative coupling, a list of infrastructure asset types sharing the infrastructure asset characteristics so as to be affected by the specific scenario;
• querying an asset database of the organization with the asset types in the list to retrieve a concrete set of infrastructure assets affected by the specific scenario;
• and,
• persisting the concrete set of infrastructure assets in association with the scenario text in the persistent storage for use in the scenario simulation engine.

A computing device comprising a non-transitory computer readable storage medium having program instructions stored therein, the instructions being executable by at least one processing core of a processing unit to cause the processing unit to perform scenario generation by:

• establishing a communicative coupling over a computer communications network between a computer-generated business continuity management user interface and a selected large language model (LLM);
• specifying an adverse event in the user interface and querying a database of scenario description records for an organization with the specified adverse event in order to receive a vector of relevant data to the specified adverse event;
• tuning a pre-determined LLM prompt with the vector of relevant data to produce a specific LLM prompt for the specified adverse event;
• submitting the specific LLM prompt to the selected LLM from over the communicative coupling, the selected LLM returning to the user interface over the communicative coupling scenario text referring to a plausible instance of the specified adverse event;
• persisting the scenario text into persistent storage in connection with a specific scenario for use in a scenario simulation engine;
• retrieving the scenario text from the persistent storage;
• extracting from the persistent storage a vector of infrastructure asset characteristics;
• tuning an additional pre-determined LLM prompt with the vector of infrastructure asset characteristics to produce an additional specific LLM prompt for the specific scenario;
• submitting the additional specific LLM prompt to an additional LLM from over the communicative coupling, the additional LLM returning to the user interface over the communicative coupling, a list of infrastructure asset types sharing the infrastructure asset characteristics so as to be affected by the specific scenario;
• querying an asset database of the organization with the asset types in the list to retrieve a concrete set of infrastructure assets affected by the specific scenario;
• and,
• persisting the concrete set of infrastructure assets in association with the scenario text in the persistent storage for use in the scenario simulation engine.