【Press Release】Recursive Launches “FindFlow”, AI Platform for the Enterprise

This is an implied translation of a press release issued in Japan on April 11, 2023.

Control security levels and AI Supports Employees in Improving Work Efficienc

Recursive has developed and launched FindFlow (hereinafter "FindFlow"), which crowls files in the company network to collect information and respond to inquiries and writing requests from internal users. FindFlow is an artificial intelligence (AI) system that can collect information from files in the company network and respond to inquiries and writing requests from internal users.

FindFlow is a tool to enhance the abilities of Large Language Models*1 like ChatGPT with internal documents, images, audio and other data types. By analyzing these data, FindFlow-based AI solutions such as chatbots, business intelligence tools, or R&D assistants can combine internal and external data for maximum efficiency. In addition to the simple functionality of asking a question and getting an answer based on natural language, it also provides a quick summary of the document. For questions that cannot be answered in a single file, the system will research across multiple files, generate appropriate answers from them, and respond in natural language.

The system is also developed to allow for customization to build your own models, and access control at the user level within your company is possible. Data is kept private, secure, and compliant, and information obtained by internal users using FindFlow is not reflected outside the company, e.g., on the Internet.

In addition, file types, not only documents and slides, but also email, chat tools, and other platforms used to share information within the company are subject to search and analysis, efficiently supporting employee work. The system is available in two languages, Japanese and English. Its reasoning engine based on Large Language Models supports both private models (like GPT-4) and open source models, which can be run on-premises*2 for maximum data privacy.

*1: Large Language Models A computer program trained on a large amount of text data collected from the Internet. It uses advanced machine learning technology that can analyze patterns and relationships in natural language data and generate new text or perform other language-related tasks based on its understanding.

*2: on-premises A software and a hardware infrastructural setup deployed & running from within the confines of the organization.

■Technical Features

• User-level access control is available
• Supports a wide range of formats including Microsoft Word, Microsoft PowerPoint, PDF, and other documents, Microsoft Teams, Slack and other chats, email, Notion documents, Salesforce, etc.
• Support Large Language models such as GPT-4, and also private Open-Source models
• Connects structured (database) and unstructured data (documents, messages, emails)

■ Functions offered to internal users

• Enter a question and the AI will search, analyze, and respond to internal information
• Cross-research across different documents and synthesize answers to complex questions
• Summarizes long documents, etc.
• Available in English and Japanese

■Pricing/System Installation Varies depending on system specifications.