In recent years, projects related to AI and data analysis have attracted attention in many companies, and while efforts are being made in various places, there are many projects that fail because they cannot connect to business.
This is because the skills and knowledge related to data science and business are not connected. So, what exactly are the “business skills ” that are the key to success in AI and data analysis projects ?
This time, Mr. Oshiro, the supervisor and author of the book “All About AI and Data Analysis Projects” released in December 2020, and Mr. Masked Analyze, the author, shared important “ We interviewed him about “ business power ” and the AI human resources that will be needed in the future.
- “Business power” is the key to success
- What is “business strength”?
- What are the pitfalls of AI and data analysis projects?
- Pitfalls that haven’t changed in years
- This is the kind of human resources you need to advance your AI business.
- The 3 skill sets of data science
- Not ready to accept data scientists
- Harness the power of business
“Business power” is the key to success
ーーCould you give us a brief overview of the book “All About AI and Data Analysis Projects”?
Mr. Oshiro: The title is the same, but the main content is ” business power x technology power = value creation “. I feel that until now, AI-related books have only had the perspective of either technical books or business books.
Bookstores have separate corners for technical books and business books, but I decided to publish this book because I thought it was necessary to combine the two in order to actually provide value.
There are many books on new technologies and case studies, but I write about “how to actually turn it into a project” from the perspective of business and technology, and the entry and exit points . You can understand the flow and characteristics of the AI / data analysis project.
The content is aimed not only at junior data scientists, but also at middle and senior class data scientists.
Junior data scientist: In this book, it refers to a person who belongs to a company’s data analysis team and works under the direction of a senior data scientist.
Middle data scientist : In this book, it refers to a person who has several years of experience as a junior data scientist and is able to make decisions and proceed with analytical work based on the knowledge gained.
Senior Data Scientist : In this paper, it refers to a generalist who is not only a project player, but also has extensive management experience.
Mr. Masked Analyze (hereinafter referred to as Mr. Masked): As a book with the same concept, “Machine learning that starts at work”, “A book that understands the project of artificial intelligence system from planning and development to operation and maintenance”, “Data analysis and data analysis that does not fail” Introducing AI into Business: From Project Progress to Organizational Development” has been published.
These books are the origin of books that cover both business and technology, but they also cover non-technical topics such as common mistakes in projects, starting up teams within the company, advice for those looking to find a job or change jobs, and contracts and legal affairs. There was a part where I felt “I want to know more deeply” in elements etc.
Also, the contents of the books published in the past were strongly focused on the side of companies that are promoting data analysis projects. On the other hand, this book is written for people who are going to work on data analysis, people who manage outsourcing, students and juniors who aim higher, and what we have learned through our experiences. .
There was an idea to aim for a book that can be used as a cane before falling down with this one book.
Mr. Oshiro: As I wrote at the beginning of “All About AI and Data Analysis Projects”, I wrote the contents of this book with the feeling of ” secret sauce “. Originally, it was a secret, and everyone brought together the content that they only wanted to teach to their disciples.
It is not about what one company has thought about, but what is written is a common understanding when bringing together the problems of data analysis projects that can be said from each company. I thought that by writing a book, I would be able to widely convey what I would have learned from my seniors when I got a job at a company.
What is “business strength”?
Data science skills have methodologies and study methods have been established, but business skills are difficult to put down in textbooks because there are some aspects that cannot be understood unless they are learned through onboarding.
ーーWhat kind of skills do you specifically mean by the “business skills” mentioned in this book
Mr. Oshiro: In a broad sense, it means ” everything other than technical skills .” The same goes for the “internal coordination ability” that Mr. Masked mentioned earlier. It is also a “businessman’s skill” to do.
It’s not necessarily limited to AI projects, it means “the ability to carry out projects as a businessman”.
In writing this book, as a recent issue, data science is becoming more research-oriented, so it is better to leave business tasks to others and engineers to handle only the technical part. The tide may be strong.
From around 2010, there was no data scientist profession at that time, and people who started out as businessmen and IT engineers learned data science. The data scientist had some practical experience in IT projects and internal coordination.
On the other hand, if someone who graduated from a university or vocational school where they learned AI education is in charge of in-house consulting, they may not have the necessary experience to promote the project, such as such personal connections and preliminary groundwork. Therefore, I wrote that it is better to have both data science knowledge and business implementation skills.
In summary, “business power” means “everything other than the technology needed to move a project forward.”
Mr. Masked: I think that “business power x technological power = value creation” has a wide range of meanings, but in layman’s terms, it is the point of ” how to generate profits, how to produce results, and how to be evaluated as a company. ” think.
I think we can create value by using AI to reproduce tasks that only veteran humans can do, and by using AI to reproduce tasks that place a heavy burden on humans. For example, if you look at a crack in a tunnel and use AI to reproduce a judgment that requires many years of intuition and experience, such as “This is dangerous if left unattended,” it will eliminate dependence on individual skills and lead to value creation.
At the same time, when proceeding with AI and data analysis projects, if you team up with a company that does not have the know-how to utilize technology in business, or a company that understands the business but does not have the technical capabilities, there is a high possibility of failure. .
In order to avoid it, it is necessary to grasp both “business” and “technology”. It would be difficult to focus only on business and consulting, or only on technology and research and development. In that sense, I think that “bridging power” is business power in a broad sense.
ーーHow was it when you actually wrote about “business power”?
Through writing this time, I was able to reorganize “what business skills are necessary to carry out AI and data analysis projects ? ” It was an opportunity.
By the way, the definition of a data scientist announced at the
In fact , it is not enough for a data scientist to simply have analytical and
Mr. Masked: Since the forms required by various industries, industries, and companies are different, the contents of this book are generalizations.
On top of that, I would like you to delve into your own issues and know-how, and create a “multiplication” that thinks about how to apply it to AI in your company .
What are the pitfalls of AI and data analysis projects?
Pitfalls that haven’t changed in years
ーーMr. Oshiro and Mr. Masked, what kind of pitfalls did you run into in your AI and data analysis projects?
Mr. Oshiro: In 2014 or 2015, there was a trend that often occurred , “We will collect data for the time being .” There was no other person with experience in analysis, but we created a DX department, so we started with deep learning for the time being, and this is still the case. In such cases, the problem is not clear and the goal is not clear, so in the end, collecting data is useless.
Then, after two or three years, people would say, “What is this business for?”
Also, in the end, we were able to come up with a good measure, but in reality, it was a business that was at odds with the parent company, so there was a chance that the measures that we had worked so hard to come up with would go bankrupt…why didn’t you tell us that in the first place! I feel that there are really a lot of traps, such as the story.
Such cases are still seen here and there. The reason for this is that the use of data has progressed relatively in the last 10 years in the Internet and consulting industries, but there is a trend to start working in areas other than Tokyo and other industries from now on. My honest impression is that the same problems that have occurred in our industry in the past are occurring frequently in various places.
Mr. Masked: Examples of failures include temperature differences and gaps within the company. Since only the person in charge is proactive, he or she may get ahead of the others, and may be seen coldly by the people in the field or upper management.
As a result, it takes time to obtain the approval of the entire company, and the person in charge becomes demoralized and loses enthusiasm for the project, which may disappear naturally.
Then there’s the matter of money. Since the budget is limited, even if you say “I want to do that” or “I want to do this”, there is a limit to what you can do. Sometimes we fail because we don’t have enough budget or we make a mistake with the part that should be budgeted for.
Doing so will increase distrust of AI and data analysis within the company, and negative impressions will become stronger. Even if it’s not a problem with the person in charge, AI, or data analysis, I think the problem is that only the bad side is emphasized because the expectations were high.
If you fail the first time, the second time will be difficult, so for better or worse, you will be cautious. However, it is not good to be too cautious, so it is difficult to moderate.
Oshiro: 10 years later, some things still haven’t changed, but 10 years ago I didn’t have this book, so it was difficult (laughs).
“All about AI and data analysis projects” is a book that describes the results of our 10 years of trial and error. If I had this book 10 years ago, I honestly think I could have saved some projects.
It’s not only about this book, but there are now experienced data scientists in various places. First of all, I don’t think it’s a bad first step to learn the joseki by attending study sessions like this.
This is the kind of human resources you need to advance your AI business.
The 3 skill sets of data science
ーーWhat kind of thinking and skills do you think people will need in future AI and data analysis projects?
Mr. Oshiro: I think there is still a high demand for data scientists with all three skill sets , just like before . Whether there are such unicorns is another matter.
Currently, “AI human resources” are further subdivided into “AI human resources” that are called by various names such as data engineers, machine learning engineers, and AI engineers.
Mr. Kida of Mitsui Sumitomo Insurance wrote “Become a data analysis human resource. Aim is “Business Translator”” introduces data analysis human resources called digital translators. This is a liberal arts job that connects the field and the data scientist.
In other words, I think there is a high demand for people who know both the field of AI and the data scientist, and who can act as a bridge.
Therefore, I believe that AI and data analysis projects need to work together with such people to advance the business.
Mr. Masked: I also think that the role of bridging and the position of leading the project will become important.
A person like Mr. Oshiro who understands both business and AI, and who, as a leader, can persuade his superiors to take care of his subordinates and assume a supervisory position, would have seen fewer failures.
The three skills of a data scientist are “ business skills , ” “ data science skills , ” and “ data engineering skills .”
When embarking on an AI or data analysis project, it is possible to proceed smoothly if there is a person who can act as the central axis.
Not ready to accept data scientists
ーーAs the number of students studying data science is expected to increase, and data scientists are expected to be supplied to the market, do you have any concerns?
Mr. Masked: Most companies are not well prepared.
For example, large companies are slow to make decisions, have little experience working with data scientists, and cannot be said to be a company that is ready for students who have learned data science to start working immediately after joining the company. .
There is concern that the number of students and young working adults who are troubled by these gaps will increase.
Mr. Oshiro: I feel that the gap is likely to become apparent from here on out.
For example, when a company considers entrusting a project to a consultant with only one year of experience, it is difficult. In the first year of new employees, most of the cases are under the supervision of senior employees.
A data scientist must also have a part of the consulting area in terms of business ability, so it will be necessary to experience the basics.
The consulting industry has a history of 50 to 100 years, but the data science industry has only been around for about 10 years, so there is no system in place for accumulating practical experience. And many companies are groping.
Harness the power of business
ーーIn what situations should student data scientists entering the market acquire business skills?
Mr. Masked: I think it varies depending on the company you belong to, but if it’s a company with a long history and tradition, it’s necessary to be like a newcomer and not be afraid to do cumbersome work.
On the other hand, in an ongoing IT startup, you are expected to give your opinions, ask questions, and participate in new projects, so I think it depends on the case.
What is important in common is how to get along with people inside and outside the company. Since we are human beings, if we get along well, we can communicate smoothly and proceed with the project without stress.
I think that making friends with people who are older than you or who have different values than you will be an opportunity to develop your business skills as a new member of society.
Mr. Oshiro: At first, any position is fine, such as an engineer or a planner in the field you want to work in, so start from where the industry has a human resource development scheme, and after two to three years after acquiring business skills and engineering skills, etc. We recommend that you pivot.
By doing so, it will be very easy for companies to hire, and the frontage will be wider.
In the first place, I think that data science and engineering essentially do not change what people want to do. Therefore, I think it would be good to pivot from the middle after acquiring strength as a career path.
Also, I think it would be nice to participate in a lot of study sessions and accumulate output. During my 10 years of study sessions in various communities, I met a student. That student was presenting his findings on data analysis in Lightning Talk every month at a study group, and he entered Recruit and Accenture as a new graduate. Therefore, I think there is also a way to participate in study groups, take action on your own, and make it part of your achievements.
Also, although it’s difficult to hold events offline these days, social gatherings and other places where members of the public gather can receive real-world information, so I think it would be effective to use such places. It is recommended because you can get information on internship recruitment at companies, and you can also make connections with seniors who can consult with you when something happens, and you may have the opportunity to gain practical experience there.
Mr. Masked: I also agree with actively outputting. I think it’s important to make announcements and show up, because if you have connections, you’ll have more chances.
It’s hard to go out now due to the coronavirus, but I think it’s important to talk to people on social media and experience it.
Only a few years have passed since the job of data scientist was recognized. As a result, the training system within the company may not be in place.
People aiming to become AI human resources may be worried, but now there are more opportunities for input and output such as long-term internships, lectures, and presentations.
In order to become an AI human resource like a data scientist and play an active role, it is important to take a step forward, such as participating in various events, making connections, and experiencing practical work.
Mr. Oshiro and Mr. Masked Analyze, who cooperated in this interview, published “All About AI and Data Analysis Projects”, which is divided into levels from junior to senior. If you are thinking of trying an AI / data analysis project, please read it once.