Data×AI — Forerunners Reveal the Secret to Success

As suggested by the expression "Data is the new oil," data drives the future of business and produces profits, and AI is a tool to make use of massive data. Here, we explore the challenges of data utilization by adopting a business perspective to examine data and AI.
[Fujitsu Forum 2019 Frontline Session Report]

In this session, Fujitsu's Mr. Hirofumi Watase, Mr. Akira Shibata of AI startup DataRobot Japan, and Mr. Yuta Kikuchi of AI startup ABEJA gave presentations on the utilization of AI in business. After the presentations, Dr. Akie Iriyama of Waseda Business School joined the panel to moderate a discussion on the challenges and prospects of data and AI utilization from a pioneering perspective.

Designing a Human-Centric, Trusted, Bright Future

Hirofumi Watase
Head of the AI Services Business Unit
Head of the Data Business Development Office
Fujitsu Limited

"Data is the new oil" - This expression has been used all over the world. The global market for data is estimated to reach 10 trillion yen - or even as much as 175 trillion yen - during the next several years. The era of data ubiquity in every aspect of business and daily life has begun.

However, merely gathering data does not create value. The DIKW Pyramid model shows that data by itself is meaningless. This model explains that gathered data becomes information, information creates knowledge, and knowledge is applied as wisdom. We can collect only a limited amount of data in the world, and it is said that only about 15% of collected data is used effectively. Meanwhile, implementation of AI in society has not progressed much.

The judgment criteria differ between humans and machines or numerical formulas. For example, you are probably glad to see a bus arriving 2 minutes late if the bus usually arrives 5 minutes late. This is because always arriving 5 minutes late is your judgment criterion. However, numerically speaking, a delay of 2 minutes is still a delay. The important thing is to understand such differences and let people analyze AI's answers and derive the correct solutions.

Megacorporations such as GAFA possess enormous amounts of data, and want to use that data and technology to create a better world. Setting a purpose, preparing data for said purpose, removing noise from the data, and applying tools to utilize the data - they iterate this cycle to achieve their goals. The most important thing in this cycle is to clearly define what you wish to achieve at the beginning.

Goals Are the Key to Data Utilization

Based on this idea, Fujitsu proposes "Design the Trusted Future by Data×AI" to create a human-centric, trusted, and bright future through integrated utilization of data and AI.

Recently, Fujitsu designed and released a framework that consists of a process for achieving customer goals, as well as the services and technologies required to build said process.

Overview of "Design the Trusted Future by Data×AI"

Emphasis is placed on the first item, "Purpose Setting and Hypothesis Design" and the last item, "Process Implementation and Value Creation." Data and AI are valuable only when applied in a business process. So, how can we apply data and AI in business? If goals and challenges are not clearly defined at the start, the process that results becomes ambiguous. By firmly agreeing on goals and challenges, the final implementation step will be achieved, thereby generating value.

The world of AI includes various technologies. Deep learning is just one of these. In the "Design the Trusted Future by Data×AI" framework, customers can pick and choose which technologies are required to successfully achieve their goals. Customers may even include technologies that Fujitsu does not possess via collaboration with other companies, including startups.

Fujitsu operates research facilities globally and works jointly with research institutions, including universities. For example, Inria, a French national research institution, and Fujitsu are working together to analyze time-series data obtained from various IoT sensors and other devices. This study applies an advanced form of mathematics called topology to data analysis. It enables valuable information to be extracted from data that is difficult to handle using conventional methods. Using this Topological Data Analysis (TDA) technology, the rate of false judgment when detecting arrythmia from electrocardiograms has been reduced by 70%. We will incorporate such new technologies and further promote utilization of data and AI.

By viewing technology as a means to an end, Fujitsu will help our customers achieve their goals and solve their challenges in order to support the creation of a bright future.

The Need for a Framework to Enable Everyone to Utilize Data and AI

Next on stage was Mr. Shibata from AI startup DataRobot Japan, who is working to automate AI-based data utilization.

Akira Shibata
Chief Data Scientist
DataRobot Japan

When talking with people from various Japanese companies, what I feel as a data scientist is that although the fields of business differ, like manufacturing and distribution, their problems have much in common.

Many have told me that they would like to apply AI to use data effectively to solve problems. However, at present, there is an overwhelmingly short supply of data scientists in contrast to the high demand. So, we at DataRobot provide services that help people, even without knowledge of AI or data science, to develop AI-driven businesses.

To exploit data for business, you must carry out the process of preparing data, formulating a management method, deploying a model, and establishing post-implementation operation rules. Many people understand the importance of establishing this process before starting a project.

In particular, a large number give importance to data preparation. Some even say data preparation accounts for 80% of the entire process. I disagree. People with such opinions tend to be satisfied when a model is completed and pay little attention to the aspect of data utilization. Preparing data is not easy. However, finding the value in prepared data and exploiting it are both more difficult and more important.

That is why automation of this process is necessary. Once the process is automated, everyone can use AI freely and businesspeople in the workplace will become able to do things that only data scientists can do at present. These new users may come up with novel ideas that data scientists are not able to.

For example, banks are adopting AI-based machine learning to predict defaulting debtors and account balances, as well as to prevent illicit use of credit cards and money laundering. In agriculture, AI is being used to predict consumer demand and pest infestations, as well as to select seeds.

For example, banks are adopting AI-based machine learning to predict defaulting debtors and account balances, as well as to prevent illicit use of credit cards and money laundering. In agriculture, AI is being used to predict consumer demand and pest infestations, as well as to select seeds.

AI's Importance Increases with the Shift from "Selling Things" to "Selling Services"

Mr. Kikuchi of AI startup ABEJA, a company that supports businesses in realizing the social implementation of AI, took the stage to speak about the process of implementing AI in business.

Yuta Kikuchi
Executive, ABEJA
Vice President and Director, CA ABEJA

According to OECD data, the U.S. GDP continues to grow year after year. By contrast, Japan’s GDP remains flat; about one-third that of the U.S. When we look at the rankings of companies by market value, Japanese companies used to rank highly during the bubble era, but now American and Chinese companies top the list. One factor that brought about this difference is the amount of AI investment. High ranking companies such as Amazon and Google invest heavily in AI, incomparably more than what Japanese companies invest. According to the Gartner Hype Cycle, which is research that indicates the maturity, adoption, and social application of specific technologies, technologies such as PaaS and deep learning are presently at the peak of inflated expectations, and their practical applications in society are expected to accelerate going forward. Japanese companies should make use of these technologies, too.

Why does business need AI?

Today, the focus of sales is moving from "selling things" such as hardware to "selling services." AI plays a major role in the latter. For example, a body weight scale is a "thing" to measure weight, but what users seek is "good health." So, instead of selling scales, we can provide a service to support good health, which is what users truly seek, by using AI to collect data that contributes to the promotion of good health and analyzing such data to create a service.

Just like the human brain, AI accumulates data and becomes smarter through repeated learning. Therefore, the process of continuously accumulating data via a subscription model, improving the accuracy, and returning new value to users is crucial. If data that contributes to the promotion of good health, such as body weight, is accumulated over an extended period of time, the service will improve further based on that data. AI supports such services behind the scenes.

What is important here is that AI is a means, not a goal. First, you must clearly define what you want to do with AI. If you ignore this step, you will certainly fail in the next step.

I would like to introduce some of ABEJA's efforts. In the retail distribution industry, businesses obtain store visitor information by setting up in-store cameras and using AI to analyze the images from the cameras. Businesses can now analyze previously unavailable in-store customer purchase behavior data, such as customer demographic attributes and the number of visitors per store. As a result, hypothesis testing for issues at individual stores becomes possible, such as increasing the number of staff members or changing the in-store layout. In the manufacturing field, AI-based behavior analysis of video of skilled workers visualizes specific inefficient operations, thereby identifying areas that require improvement.

AI utilization in business has only just begun. At present, attention is limited to AI technology and vendors that provide said technology. However, that is not what we want. By offering products such as ABEJA Insight for Retail, a store analysis service for the retail distribution industry, and ABEJA Platform, a labor-saving AI development and operation platform, ABEJA promotes the democratization of AI and helps to create more AI-based businesses.

The Hurdle to Clear to Use Data and AI in Business

Waseda Business School professor Dr. Iriyama summarized the presentations. "Incidentally, the three presentations made the same point; namely, that it is important to clearly define why you wish to use AI." He then opened the panel discussion by asking, "Businesses are keen to implement AI. What are the difficulties and key points for implementing AI?"

Akie Iriyama
Professor
Waseda Business School
(Graduate School of Business and Finance)

Watase: Fujitsu is engaging in a large project in which the client’s entire company is transforming itself using digital technology. Some projects are very difficult in terms of how to proceed with a big theme and create value. After various instances of trial and error, we have realized that once the purpose is made clear, the project tends to proceed smoothly. It is important to discuss with customers about matters such as narrowing down the scope and proceeding in phases, and to clarify challenges based on their business goals.

Shibata: At DataRobot, we hold workshops with customers to find challenges and themes. Based on the customer company’s projected demand, our data scientists present various proposals. However, in workshops, businesspeople from the workplaces and managers tend to come up with more interesting ideas.

What is important here is that the discussion should avoid general brainstorming and focus on what AI can do. Then, we evaluate the ideas that were generated in the workshop from the viewpoint of business impact, prioritize them, and tackle them starting from those with high priority. This is our process.

Kikuchi: It may sound harsh, but you should know that insofar as AI is concerned, hardly any projects involve rigid requirement definition as is done in general system integration. To succeed, it all comes down to finding the challenges and moving the cycle toward AI implementation as quickly as possible. Taking time to define requirements only delays the implementation of AI in society.

Shibata: To accelerate business, it is necessary to proceed by repeated trial and error. However, one should not think, "We must be able to do something with the massive data we have." It is important to clarify "what for" at the start.

Watase: To quickly iterate the implementation cycle, it is desirable to establish a separate organization that is in charge of the new initiative. Alternately, it is also effective to consider incorporating the initiative into the existing business process. In other words, for the sake of starting a new initiative as well as for sustaining it, you cannot take half-hearted measures.

Kikuchi: I ask customers to do two things. First, to establish a dedicated team directly under the top management. If you leave all the work to an outside expert, the expert will not know how you think about AI, or you will have no control over the project. Second is to get the data ready. In consideration of secondary uses of data for AI and other purposes, you should prepare data in advance, such as deciding how to configure data storage and how to maintain the data.

Taking up the theme of "AI utilization requires an organization," Dr. Iriyama asked the panelists: "What type of people are suitable for running such an organization? What type of people are not?"

A Passionate Panel Discussion
"How do you use AI?"

Shibata: When a separate organization is established, and since people from the company must be involved to achieve implementation in society, it is important to have those who are driving business in the company participate.
New technologies appear one after another, and the way to create a model is different from the past. In terms of personal attributes, an open-minded way of thinking is a must.

Kikuchi: Skills are important, too. They may not need to understand the theory behind deep learning, but they must acquire the skills to utilize it in business. In order to repeatedly iterate the cycle from issue discovery to solution implementation, acquiring and training human resources is absolutely necessary.
However, such people - those who are open-minded, understand technology, and have a managerial perspective - are hard to find. When such an individual cannot be found, a team of people should divide up the roles. For example, one formulates a strategy, another builds a model, and another promotes implementation in society.

Finally, Dr. Iriyama asked: "What kind of impact do you think AI will have on business?"

Kikuchi: AI will definitely penetrate into every industry. Eventually, AI will work behind users' tools. What I mean is that people will be able to routinely use AI without being aware of it. As AI technology develops, medicine will also advance, and people will live longer. The roles of companies in such an age will be to sell services instead of selling "things" like hardware.

Shibata: Like other technologies, there are both good and bad ways to use AI. We often hear people question whether AI is good or bad for humans. The answer varies depending on who the user is. Maybe we are asking the wrong question. Either way, AI has the power to change society and can be used to help people in many ways.
AI becomes smarter by accumulating data. This invigorates economic activity. Meanwhile, we must be cautious about people who cunningly use this technology for nefarious purposes. We must think of ways to prevent AI from being used for criminal purposes.

Watase: We have released the Fujitsu Group AI Commitment, which is a code of ethics. Based on a Human Centric vision, Fujitsu conducts all corporate activities from the standpoint of "technology brings happiness to people."
We do not assume that we can achieve something by randomly trying. We always think of how we can use specific technology to help our customers achieve their goals, and we provide products and services to this end.

Dr. Iriyama concluded the panel discussion as follows: "Fujitsu has been supporting IT for Japanese companies for a long time, while DataRobot and ABEJA are startups. It was striking that they share a greater common awareness of the issues than I imagined. The next step is to utilize data in practice."

The presentation content was summarized by graphic recording.

The Will to Implement "Data × AI Utilization"

How do you use AI?

Presenters

Akira Shibata
Chief Data Scientist
DataRobot Japan

Yuta Kikuchi
Executive, ABEJA
Vice President and Director, CA ABEJA

Akie Iriyama
Professort
Waseda Business School
(Graduate School of Business and Finance)

Hirofumi Watase
Head of the AI Services Business Unit
Head of the Data Business Development Office
Fujitsu Limited