parksha technology
PKSHA launches document search function on “AI Help Desk for Microsoft Teams” to the public, demonstrating 210% automatic response compared to previous models using a hybrid of proprietary AI and generated AI Utilizes PKSHA LLMS to generate answers using AI generated from company-specific internal documents and manuals – Proposing 3 steps as a new standard for help desks
……
PKSHA Technology Co., Ltd. (Representative Director: Katsuya Uenoyama, hereinafter PKSHA) is a group company, PKSHA Workplace Co., Ltd. (Headquarters: Bunkyo-ku, Tokyo, Representative Director: Tetsuya Sato), uses PKSHA LLMS (*1), We are pleased to announce that the document search function of “AI Help Desk for Microsoft Teams (hereinafter referred to as AI Help Desk)” has been released to the public on February 7, 2024. The published document search function utilizes our proprietary AI PKSHA LLMS, ChatGPT, and Azure OpenAI Service.
PKSHA Workplace provides a variety of AI SaaS products that utilize natural language processing technology, including an AI help desk, to realize a new relationship in which workers and software interact and evolve together, helping companies We support DX promotion.
Summary of new feature release: Demonstrates automatic resolution rate of 210% compared to previous models by linking company-specific knowledge – Can be used immediately on Microsoft Tems, which is used by 320 million people worldwide every month
[Image 1: https://prtimes.jp/i/22705/138/resize/d22705-138-e7a7df822acf575ecc66-0.png&s3=22705-138-6ed97964d8f72f5dedcb034e91325b01-1280×670.png ]
AI Help Desk is an enterprise knowledge management system that uses AI chatbots to automatically respond to employee inquiries on Microsoft Teams, a business chat tool used by many Japanese companies, as well as smooth human coordination and automatic FAQ generation. It’s a system.
This time, we have released a new document search function that allows you to search for knowledge from documents such as company-specific materials and manuals. This function utilizes our proprietary AI, PKSHA LLMS, and Azure OpenAI Service to achieve high response accuracy. PoC (proof of concept) tests were completed during the beta version at a number of major companies with over 2,500 employees, and all results demonstrated an automatic response rate of over 210% compared to conventional AI chatbots. PKSHA Workplace proposes a new standard for AI help desks in three steps: traditional FAQ search, document search, and manned support.
3-step support for AI help desk using PKSHA LLMS
[Image 2: https://prtimes.jp/i/22705/138/resize/d22705-138-8840327a3469abf0fcb7-1.jpg&s3=22705-138-68e42338336e66589de2f26f0421edd9-1920×953.jpg] Step 1. AI references knowledge and automatically generates answers from registered FAQs (*2)
AI automatically answers internal inquiries based on FAQ information registered on the system.
Step 2. Knowledge reference/automatic generation of answers from internal documents (new feature)
If we cannot provide an answer based on the registered information, we will read the contents of document files owned by the company, such as manuals, minutes, company regulations, etc. uploaded to the system, and generate them on the spot in response to the inquiry.
It will be answered automatically.
Step 3. Coordinate responses to employees and provide advanced inquiry response If you are still unable to answer, please contact the relevant department via online contact. Also, people respond
Regarding your inquiries, if they can be generalized, they will be automatically turned into FAQs and registered. on the other hand There are a certain number of inquiries that are highly
individualized and difficult to resolve, such as “changing the salary transfer destination.”
The ability to collaborate with other people is also important for smooth resolution of problems within the company.
By combining a conventional chatbot with a document search function using generated AI and human connection, we have created a chatbot with a higher resolution rate (*3) while strengthening response automation. For the success of an in-house help desk, it is most important to improve the overall resolution rate, including not only chatbots but also manned responses, rather than the rate of correct answers to questions for which answers exist.
Development background: Since generative AI, the standard of AI dialogue engine has been updated – Toward a new form that utilizes LLM and corporate knowledge (wisdom/knowledge/information)
With the emergence of generative AI such as ChatGPT in the second half of 2022, the demand for improving work efficiency and productivity using AI technology is increasing, and PKSHA Workplace is based on the vision of “AI Powered Future Work”. and developed AI SaaS to support productivity improvement, and the number of customers had grown 12 times in four years.
In order to further improve productivity, we have utilized the group’s PKSHA LLMS technology to update our AI help desk, including an internal document search function. It was said that 100% automation of in-house inquiries was difficult as some inquiries were highly individualized, but the introduction of document-based search has greatly improved the range of answers, leading to full-scale deployment in the market. It has arrived.
◆Future prospects:
After release, we will improve the product through feedback from an operational perspective from companies that are already implementing experimental implementation. Furthermore, by advancing this research, we aim to realize AI that not only improves operational efficiency but also proactively supports employees in onboarding and specific work areas.
PKSHA Workplace will create the best case for collaboration between AI and humans based on its track record of implementation in the employee field and the PKSHA Group’s technological capabilities, including natural language processing, to accelerate companies’ unique knowledge management. I am willing.
*1: “PKSHA LLMS”: A solution that provides an environment that allows multiple LLMs to be integrated and customized in March 2023
(https://prtimes.jp/main/html/rd/p/000000090.000022705.html) *2 : “FAQ”: AI automatically analyzes the interaction log and generates it in the form of FAQ (Frequently Asked Questions and Answers). *3: Regarding the difference between a chatbot’s correct answer rate and solution rate: Correct answer rate indicates the ability of a chatbot to accurately answer when an employee asks a question for which there is an answer. On the other hand, resolution rate indicates the ability of employees to finally resolve their queries. Even if the correct answer rate is high, if the coverage rate for inquiries in the FAQs and documents that the chatbot is trained on is low, or the quality of the answers is low, the inquiries will not be resolved and the burden on human responses will be high.
Comments from Satoshi Asano, Executive Officer, Managing Director, Partner Business Headquarters, and General Manager of ISV Business Headquarters, Microsoft Japan Co., Ltd.
Microsoft Japan would like to sincerely welcome that PKSHA Workplace Co., Ltd. has implemented a document search function in “AI Help Desk for Microsoft Teams” and has started making it available to the public.
“AI Help Desk for Microsoft Teams” features natural language processing technology with extremely high accuracy in Japanese, and by implementing it on Microsoft Teams, it can be smoothly linked to manned response, making it a popular choice for many companies. It has contributed to improving productivity in help desk operations. The implementation of the document search function that incorporates this generation AI technology has been proven to dramatically expand the scope of automatic response and lead to further improvements in the efficiency and accuracy of inquiry response in help desk operations. This is PKSHA Workplace Co., Ltd.’s innovative approach to the challenge of improving internal work productivity, which many Japanese companies are facing, and is expected to significantly improve productivity through the use of generative AI.
Microsoft Japan will continue to work with PKSHA Workplace Co., Ltd. to provide products and services that make full use of AI technology, and to promote initiatives that can contribute to the promotion of DX in Japanese companies.
◆About “AI Help Desk for Microsoft Teams”:
AI Help Desk automates in-house help desks on Microsoft Teams by seamlessly linking multiple functions such as an AI conversation engine that utilizes PKSHA’s deep learning and natural language processing, automatic FAQ generation, human collaboration, and inquiry management. It’s a service. We improve productivity by supporting the various inquiry operations of the back office department, which responds to internal inquiries, and the support departments, which respond to inquiries about products and technical information. * OpenAI’s GPT-4 model is used for some functions.
https://aisaas.pkshatech.com/aihelpdesk/
◆About PKSHA Workplace
With our business vision of “AI Powered Future Work: Turning employees’ wisdom and connections into power for companies,” we develop and provide AI SaaS products that realize future work styles. Centering around -AI Help Desk for Microsoft Teams-, we aim to improve the sophistication of knowledge management and communication within companies. Aiming for a work environment that directly leads to maximizing the value of products and customers, we support a work style that integrates AI technology and promotes co-evolution between employees and between employees and software.
◆About PKSHA Technology Group
With the mission of “giving shape to the software of the future,” we develop AI solutions and provide AI SaaS using algorithms in the machine learning/deep learning field developed in-house to create future relationships between companies and people. I am. In addition to providing solutions tailored to customer issues based on a wide range of technologies such as automatic responses using natural language processing technology, image/video recognition, and predictive models, we also develop AI SaaS that solves common issues. We provide multifaceted support for the social implementation of software and aim for a prosperous society where people and software evolve together.
Company name: PKSHA Technology Co., Ltd.
Address: Hongo Segawa Building 4F, 2-35-10 Hongo, Bunkyo-ku, Tokyo Representative: Representative Director Katsuya Uenoyama
URL: https://www.pkshatech.com/
◆Inquiries regarding this matter
pr@pkshatech.com
*Microsoft, Microsoft Teams, and Azure are registered trademarks or trademarks of Microsoft Corporation in the United States and other countries.
*Other company names and product names mentioned are trademarks or registered trademarks of each company.
More details about this release:
https://prtimes.jp/main/html/rd/p/000000138.000022705.html