AI, Data and Human shift the organization. The core of DX realized through “data democratization”

Speakers Host : Masaya Mori  Deloitte AI Institute Deloitte Tohmatsu Group Partner

Guest : Erik Almadrones Monitor Deloitte Deloitte Tohmatsu Consulting LLC.Partner

Facilitator : Yusuke Yoshizawa Monitor Deloitte Deloitte Tohmatsu Consulting LLC.Partner

The “democratization of data” has begun
Each member of the organization is an analyst
 

Yoshizawa:
The series is titled "AI Fueled Organization". It means an organization driven by AI, but we would like to think again about why data and AI are important for corporate management now. If there is a barrier to the realization of the "AI Fueled Organization", what would it be?

Erik:  
Technology has evolved a lot in the last ten years. In that sense, I think the "AI Fueled Organization" has become more realistic and practical. However, corporate executives still believe that the Data scientists are essential because data should be analyzed by highly specialized teams. Practitioners perform accurate data analysis in scale-conscious teams. However, not all of them are always necessary, and some personnel will go into sleep mode when they are not needed. In addition, they are a group of specialists who perform data analytics and are separate from other business units. Do you have such an impression?

But now we need the democratization of data. In order for data analytics to fuel the growth of companies, it is necessary to realize an environment in which more people of the organization, not just a few experts, can access the data widely. That is the ideal.

Mori:

Yes, indeed, I also feel that the evolution of technology is making it possible to democratize data. However, although there are more than a few companies that have introduced data platforms and provide AutoML (automated machine learning), it seems that they have not yet reached the democratization of data: Data is not accessible systematically to those who need it.

Erik:

Many companies don't seem to be doing well yet. I think every company would define ‘systematic' differently and handle data throughout the organization, but the point is that all analysis using data is systematic and link to the decision-making of the organization.

In fact, organizations where data is democratized have data analytics embedded somewhere in the decision-making process. As a result, it is a system in which human resources can be appropriately allocated and data can be utilized in the organization. Then, we can achieve a high level of data democratization, even if the tools are not that great. It is rather much preferable than the companies that do not use data properly or that do not use data for decision making in spite that they have an environment where they can use high-performance BI tools, etc.,. It means that tools and technology are not always important.

Yoshizawa:

In some companies, management still believes data and AI are just only for data scientists and engineers. Such companies are not able to utilize data and AI in decision making. What I would like to ask you is what is necessary to make effective use of data analytics and AI in a large-scale company.

Erik:

The first thing we need is to define the business process. When broken down into component, roughly 40-50% of many data analytics projects run by companies have process problems. They have not been defined as issues based on social background. And 30-40% is a data analysis question. Data polling (extraction), cleanup, and transformation (conversion) are not analyzed properly.

But if these are resolved, the data analytics process is 70-90% complete. However,　it's the last 10% or less of the process finally, finally that they require data scientists. Despite how important this 90% is in an organization, many managers don't realize it. That's because they are so preoccupied with the data scientist in the final 10%.

Mori:

Yes, I agree. Some executives of many Japanese company have strong expectations that data scientists will ensure accurate analysis and provide a great example. But what really matters is the customer strategy, business challenges, and the understanding the target of the market.

“Good Data”enhances the value of decision-making  
  

Erik:

In my experience, it's very rare to find a data scientist who can also effectively talk about business and analytics. But there are valuable people with these abilities, what we would call ’translators’. They can speak and understand both languages ​​effectively, getting between business and data science. Data science professionals often don't have deep insight into strategy or use cases. But strategists do not have analytics knowledge. Therefore, if there is a person who can act as a bridge effectively, understanding between both parties will deepen, and we can move forward with the project rapidly.

Yoshizawa:

I worked with you on a project four or five years ago. It was the case of an automotive company that we worked with US Deloitte. Deloitte Japan had assigned one data scientist for similar scope. But on the other hand, from US Deloitte, I think there were about 5 people? including Erik-san. I remember the team, that the1 data architect, 1 consultant familiar with the business of a car company, 1 engineer who can manipulate data, 1 analyst who creates a mathematical model, and you. I was amazed at how each of them had their own area of ​​responsibility. In Japan, we were in charge of these things alone.

The team consisted of the necessary personnel for the desired results, each organized according to their specialty. In Japan, people tend to think that everything related to data should be left to data scientists, but I think it confuse situation in various project sites.

Erik:
I generally agree to you. But I think Japan has its advantages.

I think the view in Japan is “Data speaks”. If you have a good data, and you listen to paying attention, you think you can make decisions from it. It's a little bit more in the DNA of Japan. It can be said you have ability to use data science effectively. People think, in some countries, that they do not need data and do not want to fall into dataism because " I've been doing this for 30 years”. 

Compared to that, people trust data highly, and the ability to make decisions based on data has matured in Japanese companies. Good data leads to good decisions. This is true.

Mori:

Many managers question what you said. “Then, how should we distinguish between good data and bad data?” What do you think about it?

Erik:

OK, So, when you get a data set, there are various check items such as whether the quality is properly guaranteed, whether it meets logical standards, etc. Now, away from it, judge from your point of view of business.

Good data supports managerial decision-making. In other words, decisions are made differently with or without data. The quality of data is determined by how well it connects to real-world behavior. No matter how good the data is, even if it is perfect data with high accuracy, if it is not involved in decision making, it will be of limited use and limited convenience.

I had a client years ago, that was very proud of the analytics efforts. An executive was priding about the number of data sources they have. He told 96 different data sources are aggregated in the data lake, and can be freely extracted and utilized using BI tools. Then I asked him. “What are you using the data for? What use cases are enabling? What kind of decisions are you making with the data? How does business change with and without data?” He didn't answer. I am not saying that collecting data is useless. We need such a data platform. But data that does not lead to decision making is worthless.

It is also important that the data is available when it is needed. Data that contributes to decision-making must be readily available and speedy doing it must be possible.

Mori:
Even if a client has a lot of data sets, even if they are data that has been scoured from various data sources, it is meaningless without a clearly defined cause and purpose, you mean. I would like to ask another related question. Recently, I have been thinking whether client companies should reconsider their ‘purpose’. They want to reconsider what to focus on and what value to offer to customers.  Some companies are building data by linking their purposes. But if they can link data strategies of the purpose and data use cases, is it correct to think that it will make the employees be able to utilize data while understanding their purpose, too?

Erik:

Yes. I do think in Japan it is important for companies to define an effective data strategy engagement before embarking on a large-scale analytics project, and to start with the key parts of the business. 

Finding all of the business questions and the use cases and how they would use data to make different decisions helps you to prioritize which data is the most important, right? Then, you have the data you want. Is it "nice to have" versus "you really need”?  You will understand them too. Many use cases are a very good way for many companies to kick-start their analytics.

Yoshizawa:

How should decision makers deal with data?

Conduct a PoC that makes business sense

A PoC that hasn't care even if it fails gives neither effect nor knowledge

Erik:

So, it may sound funny, but I would like to encourage those executives should stop to hang attention to the buzzwords of the latest data analytics. What is the difference between machine learning and AI, and which is the more comprehensive concept? Those buzzwords don't matter. Again, the executives should focus on the key question for them is what kind of questions should be asked of data analytics.

 You have talented people. You have the data.  And you can define the correct task, too. Then, you can come up with the answer even if the technology you use is different. Any methodology is fine. Executives should start by asking questions that get to the heart of their business, and prepare appropriate human resources, data, and tools in the organization, and delegate authority to next-generation human resources who can create new things. That's exactly what DX is.

There is one more thing that executives should be aware of. This may be hard in Japan. The Japanese way is to start small. It means to try small-scale PoC experimentally, and in that way, to try data analytics using AI. Then you can find what happens, and if it goes well, you can try another small PoC again. This is not a bad idea, but unfortunately large project plans are often spread out with small results, and there is a possibility that you will not be able to derive the big outcome that you originally wanted to find.

My advice to executives about kickstarting data analytics programs is: It's better to focus on two or three big areas in the business and try to find a solution to those problems they have, rather than a small piece by piece. 

Mori:

I agree. Small-scale PoC sometimes lead to best practices, but if they only do small-scale PoC, the process will also be small-scale. But a common mistake is, that you may end up with a small-scale PoC if you focus too much on a small PoC, then that drags down the vision and worldview you really wanted to achieve.

Erik:
The bigger the new question is, the more people will be involved, and the momentum within the organization will rise and become heated. And everyone involved can proceed with the project while considering it as "skin in the game".

But with a small PoC, no one cares if 10 fail. If it's a small, it's nothing, there's nothing lost. That's the PoC.

‘I want you to know me!’

Stay close to digital-based loyalty

Mori:

If companies could take the right approach, how would interaction with customers change?

Erik:

Expectations of modern consumers about their favorite brands are, “You know who I am, that you recognize me.” So, "Recommend me what's right for me, before I know it.” We can anticipate what they want and be able to provide a recommendation of what they want even before they know that they needed. All of those Modern consumer expectations are driven by data, right?

I think Mori-san's question was about how to respond to customers when data analytics is done correctly. However, the world is too big and too complex for everybody to know, so that it is not easy to accurately capture consumer trends. It is not easy to keep track of customers. What size product they buy, what color they like, when they buy it, and what method of payment they prefer. However, if a brand that they like knows about their tastes and ways of thinking and recommends it before they want it, they will feel grateful and they loyalty to the brand increases. Building relationships with these customers leads to a good way to enhance our customers. Data obtained in the process of repeat purchases can also be used to anticipate consumer behavior.

In this way, useful data for deepening relationships with customers will be the fundamentals of basics of marketing skill that meet the expectations of future customers, such as building systems and developing AI chatbots.

Yoshizawa:

Erik-san, you have worked in the Japanese and American markets and has insight into both, but you believe there is a big difference between the two. The biggest difference is that 70% to 80% of data scientists in Japan are from external vendors, whereas in the US it's the opposite.  Like Apple, Facebook, and GE, for example 70% to 80% are staff within the client company. Why is there such a difference? If there is a data scientist in the company, knowledge will be accumulated in the company, but in Japan, knowledge is outside and it is not accumulated in the company, which is worrisome.

Erik:

I'd start I guess, there are very few born data scientists, right? Oftentimes they are made and they are developed over long period of time. Really great data scientists to begin with are rare. I think it's very rare to find really, really good and talented data scientists, that would say even in, very mature markets, like the US. Therefore, I believe that it is important to hire and educate people who do not have much experience and who have room for growth, and to develop them into practitioners of data science and analytics. So I would actually be advising to start with less experience, less developed talent. And to be able to train and nurture that talent to be the sort of data science and analytics practitioner that you need. What you have to think about is what kind of data analytics and data science practitioners you want to have. Then let's give them the delegation of authority, that is adequate freedom and exposure to see other places so that they can in turn bring that sort of knowledge and the expertise back to the companies that they work.

Yoshizawa:

We should be conscious of that too.

Make up the diverse ecosystem

The key is work with alliances outside  

Mori:

Finally, is there anything you would like to add about the topic, "AI Fueled Organization"?

Erik:

What I feel is growing in importance now are “partnerships” and “ecosystems.” It is impossible for a company to complete the relationship with increasingly complex consumer behavior on its own. I think it's difficult even for the best companies.

The ultimate goal of a car company is to sell cars. Communicating effectively with consumers is not the purpose. It's just a means to an end. Therefore, it is extraordinarily important to work with partners and alliances that make up the appropriate ecosystem to increase the number of people who can better support what you are trying to achieve. As components of that ecosystem, I think we need big hyper scalars that are good at scaling businesses, and some Boutique Partners who specialize in various small unique things that are very crucial for your business.

Yoshizawa:
Thank you very much for your insightful talk. In fact, with regard to promoting DX within the company, we have sent out the message "Think Big, Start Small" with the image of ‘starting’ with a small PoC. However, only start ‘small’ was emphasized, and "think big" tended to be overlooked. This talk was a good opportunity to renew our awareness.