By Alibaba Cloud Enterprise Application Team
Li Bo, Senior Algorithm Expert at the Information Platform Business Unit of Alibaba Group, delivered a speech titled “Enterprise Smart Brain Organized and Empowered by AI” at the enterprise application forum of The Computing Conference 2018.
In this article, we will recap the main content of the speech. There are three main topics in the second half of Internet development: online and offline interconnection, artificial intelligence (AI), and enterprise services. The combination of these topics is significantly changing and optimizing the operation of enterprise functions. This speech elaborates on the true value that AI brings to enterprises, how to build an enterprise smart brain, and how to effectively implement the enterprise smart brain in human resources (HR), legal affairs, IT asset management, and other scenarios.
The Information Platform Business Unit is the team that builds the internal collaboration/operation platform of Alibaba Group. The unit is responsible for all internal systems, including the HR, administration, finance, procurement, IT business, asset and outsourcing management, workflow, and office network systems.
The enterprise data intelligence team has been established for more than a year. Conforming to the data-based strategy for information platforms, the team has explored the direction of “enterprise smart brain” and implemented the enterprise smart brain in many enterprise business fields.
The enterprise smart brain is an enterprise intelligent open innovation platform built based on the integration of new IT technologies such as AI and big data. It assists in intelligent decision-making and business automation, and drives intelligent upgrades of business systems, to achieve more personalized, customized, and refined enterprise production and operation.
“Information silos” mean that internal data of an enterprise is not interconnected. There are two reasons for information silos. One reason is imbalanced development of different types of business and sequential development of technical systems. In this case, two systems may use different data models to describe the same business concept, resulting in data interconnection failure. The other reason is offline storage of a large amount of data, such as paper documents, legal documents, reimbursement bills, and user behavior (for example, whether a conference room is being used). Therefore, we need to combine AI technologies, such as Natural Language Processing (NLP) and Computer Vision (CV), with low-cost Internet of Things (IoT) devices to break information silos to improve efficiency.
Deep integration into applications requires that we integrate AI into applications in the form of “Industry + AI.” The internal organizational operation in traditional industries is mature but still involves a lot of manual work, which is inefficient and error-prone. The involvement of AI can better improve operational efficiency.
Consumer to business (C2B) migration means migrating the successful AI experience of consumer applications to business applications. We can use the experience of successful consumer applications and consumer AI applications to effectively shorten the build path of the enterprise smart brain. This is also the key exploration direction for building the enterprise smart brain in the next five to ten years.
There are two frequently asked questions in the promotion business: Which candidates have the potential for promotion? Are these candidates better in line with promotion criteria than other candidates? These two questions are usually answered by supervisors and HR personnel, who inevitably include their subjective biases in the answers. Therefore, we use objective data to build an evaluation indicator system and an algorithm-based predictive model, which assist supervisors and HR personnel in making decisions during the nomination and review phases.
Interviewers’ interview skills and maturity directly determine the efficiency and effectiveness of recruitment. Unlike the promotion-assisted model, the interviewer evaluation model lacks objective historical data. For this reason, we need to apply active learning to the interviewer evaluation model and combine manual modeling with machine modeling. Although this model covers only 20% of interviewers, it can greatly help them with data-based operation.
Because each employee’s nickname is unique and ex-employees’ nicknames are retained, it is difficult for new employees to pick a nickname. The intelligent nickname system can randomly recommend nicknames, search for nicknames based on specified keywords, or even give nicknames based on a description or definition preference. For example, if you want to include the meaning of “the person who carves out and leads the way” in your nickname, the system can recommend nicknames such as “forerunner,” “pioneer,” and “trailblazer.” Since the intelligent nickname system was put into use, the proportion of employees who accept intelligent nicknames has exceeded 60%.
The automatic document review system automatically reviews agreements to detect potential risks and offer suggestions. It can further reduce risks on the Alibaba Information Platform. At present, the identification accuracy of automatic agreement review is about 98%. The system is able to detect 85% of violations. In addition, AI can also assist in the review of contract forms, including the content consistency review of contract text, contract amount correctness check (for example, whether the amount in words is consistent with the amount in figures), clause integrity check, serial number check, and typo check. The accuracy of contract form review is above 90%, with great flexibility in functional expansion.
A large number of legal documents, including contract documents, indictments, and evidence, mainly exist on paper. The method of interconnecting paper documents with information systems determines the efficiency of the entire business IT-based processing process. The intelligent document entry system can automatically convert offline text to online content, and use NLP information extraction to automatically extract and enter key information and classify clause types.
Contract search raises extremely high requirements for security and confidentiality. We have developed and deployed a set of encrypted search functions in the intelligent contract search system to effectively guarantee data security. According to the characteristics of legal documents, the system has also implemented a customized search and sorting process, which increases the overall search relevance to over 90%.
To learn more about Alibaba Cloud Enterprise Smart Brain solutions, visit www.alibabacloud.com/et