The reason why I tried to write such an article is that there is almost no information in Japan about the use of machine learning.
So, I was told by a good friend ** to investigate a case study on machine learning business improvement **, so I decided to summarize it.
By the way, ** Machine Re engineering ** in the title is an improvement by automating work using machine learning. What is it translated into Japanese? The range is quite narrow, but marketing automation may be in this area.
Actual Machine Re-engineering is a broader story and refers to the entire business process. Although it is a word that is not well known in Japan, such as exchanging emails with clients, ** churn ** prevention can be overwhelmingly improved by using machine learning. Let's take a closer look and summarize the areas that can be improved.
By the way, here is the blog that I run: Effort 1mm
Business improvement using machine learning can be roughly divided into the following three areas.
Let's look at some concrete examples. The following four examples will be introduced.
etc. All of them had surprising contents, so I will summarize them.
A case where a bank called Nuance in Canada used ** voice verification ** to recognize human voiceprints and reduced the response time per person by as much as 40 seconds.
The procedure is like this.
It seems that the response time has decreased by about 40 seconds per person. I'm very happy that I've done voice verification in a natural conversation, and when I notice it, I don't have to verify my identity. That exchange is very annoying. ..
There should be no reason why this is not possible, as voice verification is not a language, but a matter of being able to recognize human voiceprints. When it comes to speech recognition, it should be difficult to do it in Japanese (language structure problems, no excellent speech recognition server in Japanese, etc.).
Document creation and data entry that are common on a daily basis. If you take the time to die for such a simple task, 30 hours a day is not enough. What's more, this kind of wasteful and monotonous work is quite discouraging. I don't want to do it if possible.
The one who helps to create reports automatically. Even though it is automatically created, it does not mean that the text is written completely automatically. You should think that this service using machine learning will edit the text you wrote.
Write a highly complete document using disjointed and bulleted sentences as input data. It is said that it will create a proper document by inputting bulleted sentences. Customers have a wide range of industries, including healthcare companies, finance companies, and oil companies. I wonder if it is a service that is very welcomed by large companies with a lot of documentation.
Productivity seems to have increased by about 25%. Replacing it with familiar numbers, we have succeeded in reducing the time required for reporting tasks by 40 hours / month. I have about 2 hours of free time a day. it's amazing.
You shouldn't expect it to be completed in Japanese with the same quality as Latin. I often hear that I'm not very good at assembling sentences because the rules of writing in Japanese are broken. ..
Rather than quality control, it is necessary to closely monitor the manufacturing process. If an alert sounds across departments in a situation where each department cannot contact each other in a large warehouse, it can take time to investigate the cause.
For you who have such troubles! Don't give up, you have a Sight Machine! It would be great if the cause of the alert could be identified in a blink of an eye, and ** "it will tell you what's wrong before the alert sounds" **.
At the manufacturing floor, a single quality problem can trigger hundreds of alerts at the same time. It usually takes a lot of time to identify the problem that underlies the number of alerts, and sometimes even needs to stop the manufacturing process.
Sight Machine learns the alert pattern to help identify the problem. But next time it's even more amazing.
Furthermore, it seems that it is possible to learn the pattern of cases that occurred before the problem appeared and let them know that an error is likely to occur as it is. And the damage is minimized by dealing with the part where the error occurs in advance. Well, it's a common pattern of machine learning, but it's amazing to do this in the actual field. It feels like an evolved version of my favorite Komatsu IoT case.
I think many people know this, but as I summarized before, Singapore is an artificial intelligence power. A country where the minister is saying something like introducing artificial intelligence throughout the country and making it a laboratory for future society. Personally, I think this country like Singapore is really cool. It's nice to have a drastic desire for the future rather than being conservative.
Banking operations are fairly straightforwardly related to the desires of various people. If you know insider information in advance, you can make a sneak peek, and if you can get a loan from a bank, it will directly affect the survival of the company. Of course, there seems to be a stinking story in such a place.
Artificial intelligence that detects "people who are likely to be corrupt" by learning based on emails, internal behavior, and meeting behavior.
It seems that there have been attempts to detect bank fraud for some time. That's right, but the conventional method was research on external factors such as "fraud / yield". In other words, it seems that a third-party institution, not a company, was using artificial intelligence to monitor the market.
The DBS example should be pretty famous, but that's why. It seems to be rare as an example of being used for fraud detection in the company.
A large initial investment may be required to start data analysis. As anyone who has analyzed data knows, collecting and cleansing data takes a deadly amount of time, and it's so annoying and difficult that only an expert can do it. Moreover, it may fail. Data analysis cannot always give the correct results. I rarely see any successful examples of machine learning in Japan that have been talked about.
I wonder if Japanese companies can make such a decision, and it makes me feel very dark. Companies may be able to do it, but I feel that there are some societies that don't like the ** nice and rational irresponsibility ** of "let's leave it to machines."
I love learning and I do it quite seriously. That is why I feel the limits of human beings who have emotions. As long as emotions are blunting, it is impossible to behave 100% reasonably and ideally. I think it would be nice to be in a world where you can leave it to machines to increase productivity, make time, and spend time on your hobbies.
Click here for the blog that we run: Machine Re-enginereeing in the latest overseas cases
Recommended Posts