JapanManufacturing2023 – ongoingSankyo Tateyama
AI Futaba in-house generative AI assistant
Sankyo Tateyama's in-house AI assistant saw use grow about 1.5 to 2 times after version 2, AWS says
Problem
Sankyo Tateyama decided in June 2023 to use generative AI actively. It needed an internal assistant that could create minutes and answer questions from company documents.
Approach
The company built version 2 of its assistant, AI Futaba, in about one month on the open-source AWS project GenU. It uses Amazon Bedrock, Amazon Kendra for retrieval over internal documents, and some external models. It offers minutes creation with transcription and about 50 prompt templates for common business tasks. AWS Japan co-wrote the blog post.
Results
Takeaway
Inconsistent product names in internal documents needed extra handling, such as metadata, before document search gave usable answers. The 15,000 annual hours saved are a projection, not a measured result.
Categories
- Size
- Large enterprise
- Industry
- Manufacturing
- Business function
- Internal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Built in-house
- Evidence
- Vendor case study
Sources
Source summaries
An AWS Japan blog post co-written with Sankyo Tateyama. It described how version 2 of AI Futaba was built on GenU with Bedrock and Kendra. It reported growth in use of about 1.5 to 2 times and about 50 prompt templates. It also gave a projected saving of 15,000 work hours a year, which was an expectation only.
Full details are in the original sources.