In 2020, digital tools are being used in various fields due to the movement restrictions due to the spread of the new coronavirus and the abolition of stamps by the government. At the same time, attention to DX is increasing, and in the AI field, an increasing number of companies are introducing AI-OCR and demand prediction AI.
Amid the buzz of terms such as DX and AI , in order to increase the superiority of companies in the future, it is necessary to incorporate AI and DX into strategies and promote transformation throughout the company .
On April 14, 2021, Mr. Hajime Hotta, founder of Cinnamon Co., Ltd., a leading AI venture company in Japan, and Mr. Kazuhiro Ohara, who is also known as the author of “Survive in an era without after-digital offline”, announced ” Double Harvest Victory”. Strategic Design in the AI Era (Double Harvest), which creates a mechanism to continue, has been published.
This time, we interviewed Mr. Hotta, who has been involved in AI and DX strategies for many large companies, about the winning patterns of DX strategies that he has seen from his experience.
- Thoughts I want to convey with “Double Harvest”
- The trend in the DX era is “Natural Language Processing”
- Combining strategy and natural language processing
- Double Harvest Success Story
- Purpose driven learning from successful cases
- Two organizations failing with their DX strategy
Thoughts I want to convey with “Double Harvest”
Double harvesting is not a “one harvest and that’s it”, but by correctly designing a strategy that incorporates AI and creating a self-propelled mechanism (loop structure), you can continue to harvest two or three times. refers to (Quoted from this book)
–How did you come to publish ” Double Harvest ” with Mr. Ohara, an IT critic?
I thought that by using the concept of ” loop ” when the two of us were reorganizing our thoughts, we could explain various things.
ーーWhat kind of people would you like to read “Double Harvest”?
The person in charge of the DX project does not necessarily need to be familiar with AI technology. This is because if you have some knowledge about business, you can develop strategies for AI and DX businesses.
“Knowing business patterns ” is more important than “knowing technology”, and if you consider all possible patterns in the project, you will be able to see the whole picture. It would be great if we could enlighten people in that way through Double Harvest.
Also, I would like not only those involved in DX but also entrepreneurs and product managers of venture companies to read it.
I think there are many people in the management of venture companies who think that the DX strategy has nothing to do with it, but even if AI is not used, if you think that the time will come when you can store all the data in the company and use it in the future, the relationship I don’t think there are any entrepreneurs or managers who think that they are not.
For example, let’s say you run a fitness business. Just by installing one camera in the gym, you will be able to see what kind of conversation there is and what kind of training menu you are working on.
Even if you can’t analyze it right away, you can move on to the next action just by having the data. In any case, it is important to try to accumulate data and try to set things up, but there are many managers who are unable to take action and are unable to make the most of their company’s strengths .
Through this book, there is a strong social feeling that it would be nice if such people could be stimulated.
The trend in the DX era is “Natural Language Processing”
ーー5 years have passed since Cinnamon Co., Ltd. was established. How do you think the trends in the industry are changing?
On the technical side, the resolution of natural language processing has increased.
For example, image processing is a macro task for each scene, so the market does not expand and AI utilization does not progress. However, as the field of natural language processing develops, it will be possible to connect it with various technologies, so the range of applications will expand considerably.
In the field of natural language processing, with the announcement of the natural language processing model “BERT” that can read the context and the sentence automatic generation AI “GPT-3”, its potential is attracting attention, and social contribution is expected. I’m here.
For example, when trying to deal with a complaint at a call center, if an AI with a natural language processing function presents a solution for a case similar to that trouble, I think it will overwhelmingly exceed the use case of “search”. increase.
Since it can reduce the burden on people, I think there will be more opportunities to use natural language processing in the future.
Combining strategy and natural language processing
ーーDoes the combination of strategy and natural language processing become important when working on DX?
I think that in the future, people will probably pay more attention to what kind of meaning they provide rather than what functions are useful to everyone .
If you compare it with job change activities, you can find ways to present value from various angles, such as rooted in emotional value such as ” matching is better than other services ” and ” you can devote yourself to job change activities without anxiety “.
If you think purpose-driven and adopt a strategy of “differentiating from other companies by emotional value,” you will measure and raise emotional value in communication with service users. In such cases, we sometimes come to the question of ” how human-like communication is included .”
In that case, instead of full automation, by actually having people respond and accumulating data together with customer reactions, we will eventually be able to promote AI learning with emotional data such as when the customer feels relieved. Become.
In the medium to long term, it may be possible to scale with the power of AI while valuing each customer with the strength of “emotional understanding” as a result .
So I think I need to go back to purpose after all. Once the differentiation point is decided, the tactics after that should change for each differentiation point, so by collecting the data necessary in the discussion and multiplying it with highly versatile natural language processing, it is possible to win. It’s about connecting with strategy .
Double Harvest Success Story
ーーAre there any successful examples of companies that practice double harvesting?
With the “hospitalization passport”, you can apply for actual cost-type medical insurance from the 2D barcode, and you can complete the procedure in about 5 minutes . In principle, premiums can be received within 24 hours, or in as little as 30 minutes.
This example focuses on two points: ” How satisfying is the UX when acquiring data ” and ” Automation of the variation process “. Both of them store data, so the accuracy increases as the user uses it.
Even if other companies develop similar services, I think Sompo Japan will be able to do it in 5 minutes while other companies take 30 minutes.
Once you start a DX business that attracts attention, it is often imitated by other companies. The point of double harvesting is how much you can keep securing the first-mover advantage at the timing of being imitated.
At first, you may be able to differentiate yourself from other companies, but after that, it is quite possible that you will be overtaken by other large companies, so it is important to pursue UX to prevent that from happening . By doing so, even if a similar service is released, you can continue to differentiate yourself from others on the grounds that you can feel safe and have no worries.
In order to raise the first mover advantage to a system that continues to win , rather than one-time You will need a mechanism to store it in. I think it will be a strategy to win by creating a topic with a new service at first, and then advancing further research while other companies are developing similar services.
Purpose driven learning from successful cases
ーーI think that “utilizing data while taking advantage of the strengths of our own services”, which was also a “double harvest”, is the basis of DX. After grasping the company’s strengths, will it become important to further utilize AI in a data-driven manner?
When working on DX, if you think “I value the emotional value of customers ” or ” This is how it should be in the future ” down to the purpose level, if you talk about sales etc. for the purpose of demand forecasting, it will not be possible. I think.
However, there are cases where such things are done on a department-by-department basis, so it is necessary to emphasize purpose . The purpose of working on DX is not to maximize sales by forecasting demand, but to deliver the original value of the company’s services to users .
By raising the flag of DX based on the value that the company originally wants to deliver, it will be possible to allocate the necessary cost to the necessary place. Doing so will increase user satisfaction.
Therefore, when thinking about what to improve, instead of just thinking about the KPIs in front of us, we should be purpose-driven and stoically ask, ” What do we really want to achieve? ” I think that is important.
ーーPlease tell us the points when designing AI strategically.
When trying to maximize the value of AI, there are cases where the idea is to “make the work more efficient with full automation,” but it is better to think in terms of lines such as 50% automation and 70% automation. I believe that the application areas of AI will increase .
For example, when considering the flow of “calling to a website and then guiding to a chatbot”, 90% of users may go to a call center before AI is used. Because this is not considered human-in-the-loop, AI cannot be utilized .
It would be more impactful to say that work can be replaced 100%, but at the current stage, such cases are limited, so if we say that it is 100% replaceable, the range of utilization will be narrowed. increase.
AI is not a substitute for humans, but the effect of AI can be demonstrated by utilizing it in a way that assists humans, so we also value the concept of human-in-the-loop.
ーーIt is important to break down the work into tasks, define tasks to be entrusted to AI and tasks to be handled by humans, and create a human-in-the-loop.
Originally, you want to release 10 products in a year, but this way of thinking is effective when you can’t scale, such as compromising with 7 products. I believe that productivity will increase, which will lead to lower prices and an increase in the amount of output.
Even in such cases, it is important to store data . For example, recording a Zoom meeting can improve communication. Regardless of AI, the accuracy of improvement is greatly different if there is evidence data.
By collecting data even for detailed tasks, it can be used for services that automatically provide feedback, which can lead to improved communication and sales, so it is important to break down the tasks of humans and AI .
Two organizations failing with their DX strategy
ーーWhat kind of organization does not work well with the DX strategy?
If we only communicate the message that AI will be used, or just the fields and potential of AI, no change will occur. The DX strategy will be effective if the management layer thinks about 5 or 10 years from now, when they coexist with data, and creates a big picture. I really feel that if we can’t draw a vision down to that level of granularity, the site that implements AI won’t move.
Next is an organization whose vision is not flexible .
For example, even if you propose a change to natural language processing to an organization that has been processing tasks without using natural language for the past two years, they will say, ” This policy has been decided for two years, and it is already working. There are cases where it is said that it cannot be changed . In such a case, we, who provide AI, will be unable to move, so we will not be able to produce results.
The characteristics of organizations that do not work well with DX strategies are that they do not have a vision for the future and that they cannot change their vision or strategy anymore .
Technology has changed a lot in the last few years. In the past, a large amount of data was required, but now it is possible to perform machine learning with a small amount of data, and with the announcement of BERT, attention has shifted from image processing to natural language processing.
At the same time as the technology changes, the winning pattern of the DX strategy changes rapidly, so the flexibility of the organization is the key point.
When promoting DX in-house, it is important for management to put into words a vision of the future five to ten years from now in a purpose-driven way and convey it to members.
If you leave it to the site, the value of AI will not be demonstrated and the cost will be wasted, so when designing a strategy, it is better to break down the tasks of humans and AI before working on DX initiatives. .