Deloitte Tohmatsu Consulting LLC, a member of the Deloitte Tohmatsu Group, is promoting the implementation of AI technology to draw out the tacit knowledge hidden within various human resources as a measure to dramatically advance data-driven management in companies. This time, we have developed an “AI Interview Agent” that consolidates that technology. This app uses multiple AI technologies to program a series of processes, from natural voice dialogue by the AI interview agent to data structuring and data search.
While the recent spread of generative AI technology has expanded opportunities to utilize corporate databases, there are many cases where necessary information remains dormant in people’s heads and has not been digitized, which is an obstacle to data-driven management. In order to digitize such tacit knowledge, the general method is for the person who has the information to transcribe it into a document themselves, or for a third party to elicit the information through interviews, but there are issues such as being too busy to find the time, not knowing what information to elicit, and the quality of the information that can be elicited being dependent on the ability of the interviewer.
In this app, AI conducts voice interviews according to a pre-set scenario and organizes the information extracted in a structured manner. This app makes it possible to obtain a large amount of information through human-AI dialogue, 24 hours a day, anywhere, with a certain level of quality. The information obtained is stored in a database in a data structure called “graph data,” which is suitable for organizing knowledge. Graph data is a data structure that is composed of nodes (vertices) and edges (sides), and is suitable for expressing complex relationships between data. When searching for information, information can be obtained in natural language, and searching accuracy can be improved by following the data structure.
Main functions of Deloitte Tohmatsu’s “AI Interview Agent”
The following features enable information acquisition through AI-assisted voice interviews and information search using natural language.
Voice interview: Voice AI conducts interviews based on pre-set scenarios and extracts information from people in a natural way (can also be used in conjunction with text input)
Conversation log display: Displays the contents of conversations between humans and AI in real time
Graph data display: The conversation content is analyzed by a large-scale language model (LLM), converted into graph data, and visually displayed on the screen.
Data editing: Users can directly correct any content that has been misrecognized by the AI on the screen.
Data search: Obtain the information you need by asking questions about graph data by voice (can also be used in conjunction with text input)
AI agent: Depending on the content of the dialogue with the user, it decides whether to conduct an interview, convert data into graph data, or perform a data search, and automatically switches between multiple processes.
SOURCE: Deloitte
