In March 2021, Dr. Wanli Min, Mr. Tyler Wang, and Mr. Fabian Wong were invited to join in Master Club. The session was hosted by China Experience in Europe and the audience included professionals and CXOs from 15 companies in different industries.
During the session, Dr. Wanli Min, Mr. Tyler Wang, and Mr. Fabian Wong shared their insights in terms of data-driven development and digital transformation in intelligent manufacturing, new retail, and digital venture. The followings are their advanced experience and comments in digital transformation and the latest trends in AI.
Dr. Wanli Min
As for AI (Artificial intelligence), I would re-interpret that A as actionable, accessible, and affordable because essentially all these algorithms together with the data streaming are used for real-time actions to prevent something bad from happening in the first place or even take some informed smart decisions in the business flow.
We have witnessed a huge retail revolution in the past five years. I call this D-TAIL (digital retail), for which everything is driven by personalized consumer understanding, the detailed understanding about putting consumers first. It is not channel-centric but consumer-centric.
Having 4G, 5G or 6G definitely will make a difference but having mindful discernible eyes on identifying tangible business problems is more critical.
Once we have well-defined business pain points with quantifiable business impact in place, we can start with three Vs, “Value first, technology second”, “Vertical first, then horizontal”, and “Velocity (speed)”.
Today, there are two kinds of companies: One kind of companies think it is a recession period, so they have to cut costs and budgets to wait and see. The other kind of companies, on the contrary, take a very bold decision to say, “oh it is the right time for us to make a jump”.
Things are a little bit more online. The world is going more digitally right now and it is a less touch or no touch economy. There are some great things or interesting observations in China’s marketplace. I believe there is an opportunity for us to share a lot of new models during the pandemic, in terms of how they can be transferred or transcended to the European markets and vice versa.
The video above shows a very traditional industry in China, the footwear industry. It is known that China produces 56% of the shoes in the world, where 7 billion people cannot move without shoes in a history of 5,000 years. So, essentially it is a very green industry with a huge persistent demand and production capacity in China. However, what if we have implemented digital technology in the footwear industry?
From the picture above, you can see that the industry is labor-intensive because there are so many workers in the production line. If you go to a typical production workshop or factory a decade ago, you will see many workers lined up, manually doing the production. Now, what the video just showed is that robotics can automatically do stitching but who is going to control the robotics behind it? Since this is a process manufacturing, there are so many procedures. For different procedures, you have different robotics with different operational schedules. How are you going to reconcile and coordinate across different procedures, consecutive procedures in such a way that they act as a well-coordinated army with a commander in chief? Who is going to be the commander in chief? It is going to be AI.
Intelligent manufacturing will not do quality control and process the lean production from the traditional six Sigma approach. We know that the six Sigma approach was developed by GE, General Electric and was applied to Toyota, a Japanese car manufacturer to achieve big success. But today, with AI in place, it looks as if each production line has a commander in chief and they can think based on the real-time operational data from adjacent upstream and downstream operational status. The question is how are we going to put the capability of thinking into this production line? That requires a lot of data as you can imagine.
Fortunately, for a typical production line, there are so many sensors that can understand the status, like the process parameters, the control parameters, and the environmental parameters of different procedures. The entire data flow corresponding to the production flow has been there and has been archived somewhere in the database. What we need to do is just to deploy AI to understand those data to create causality or quantify the relationship between different procedures in such a way that we can figure out what is the best way to tweak, to coordinate the process parameters so that you’re the final product has the best quality ever. To do a better way of quality control, the yield of the production line will also be maintained at a high level but with very little fluctuation.
Please see Figure 3, it’s just another example to show how we do this. Is it possible to apply this kind of technology to the very traditional footwear industry? The answer is yes. In a very typical workshop or factory, different things have to be scheduled for each order, i.e., the number of volumes, the sizes, the colors, the materials, the style, the design, and the SKU as you can imagine. So, suppose different orders are in the queue and each order has different sizes, how are we going to deploy the production sequence in such a way that the orders will be fulfilled in time? That’s not easy because, for each order, different types of raw material suppliers are required which is contingent upon the stocking level of the supply chain. How are we going to reshuffle different orders and generate this production sequence? That could easily explode as hundreds of possible combinations are beyond human brains. Obviously, we cannot do it with human brains but how are we going to do it? Again, AI because we all know that the best thing about AI is that it can exhaust all the possibilities very fast if you design the algorithms smart enough.
In the old days without AI, there are eight procedures on the screen, and each procedure is labor-intensive. Many manual watches are involved, and the production scheduling is cumbersome and sub-optimal because there are so many switchovers, handovers, or switchgears. Each switchgear from order A to order B requires a different type of materials and a different group of workers because their skill sets are different. So, with gears on and off, off and on, a huge waste of productivity is created.
But what if we deploy AI and streamline the entire process to a very simple approach? There are only two steps, number one and number two. We deploy a commander in chief which is AI who automatically looks at the current supply chain, the stock level of the materials, the urgency of each order, the amount of each order, and also the skill set required for the workers and has all those factors considered. The AI is going to automatically create a production plan and also generate the delivery request because once you fulfil the order, you are in the rush to customers which requires delivery. For the supply chain, the AI has to do a prediction procurement, where they can anticipate huge demand of a certain leather, for example, demand for genuine leather in three days but usually has to be placed one week in advance. So, the procurement officer will have predictive procurement support rather than just wait for the order to be issued and then do this procurement immediately which often requires at least seven days of waiting time.
Having all those AI commanders in chief in place, we will be able to create a lot of savings as you can see on the next page. The calculation of added value is very simple. The order fulfillment rates have increased by 20% and order fulfillment time is reduced by 21%, so this is very important. In the old days, one cannot fulfill the order once because the order is too big, so you have to split them into multiple batches. For each batch, essentially there exists a waiting time. Now, however, one can somehow bundle them together as a holistic order and generate and create products altogether continuously. So, it is better to finish this bulk order all at once as suppose you have enough supply and thus, we could reduce the order fulfillment time by 21%. More importantly, for the process acceleration part, all decisions in the production schedule that used to take a couple of days now can be refreshed and updated within a second, so the decision time is almost close to real-time. This translates to very fast, very intelligent commanders in chief for the production scheduling system in such a way that they can find the best combination of multiple orders to maximize utilization of the current supply of the material at the stock level and also to increase the fulfillment rate and reduce the fulfillment time which further translates to the service level guarantee, the satisfaction level of customers and more businesses for this production factories.
Christina: There is a question for you, what is your view on when you should proceed with AI in your business model and some other key points about when it is a good time to approach implementing AI? This is a participant from the financial industry who asked this question.
Dr. Min: First of all, you have got to understand your business pain points. You have to have a clear objective, say having to solve this business problem because it creates one million dollars loss or waste per year. So, this kind of mathematics has to be done even before AI kicks in. Once we have well-defined business pain points with quantifiable business impact in place, we can start with three Vs, first of which is “value first, technology second”. So, you don’t have to use the most complex neural network and you don’t have to use the most complex image recognition algorithm of AI. Rather than that, you want to find the most appropriate, suitable, reliable, and robust AI. So, “value first” means the business value should be quantifiable. The second V is “vertical first and then horizontal”. What I mean by “vertical first and then horizontal” is that first of all, you want to narrow down and focus on the problem. Don’t try to solve many problems at once. Focus on one problem, drill down to the root cause vertically and tackle that root cause to solve this problem fundamentally, rather than superficially. And by “horizontal”, I mean if you happen to solve the first business pain point, there will be more to come. Even from a big organization in the financial industry perspective, for instance, the business pain points used to be very conservative for risk compliance. But if you somehow show the power of AI with the marketing group and they might just get on board and say “I got some problems with the risk control or whatever compliance. Can we leverage AI?” By that time, you are expanding horizontally across the business domain inside of your big corporation. The third is “velocity” by which I mean speed. You have to act fast and conclude fast. Thus, don’t do this AI implementing the project as a two-year project because for innovation, the patience is very slim. I would suggest that you should never try a project that goes beyond six months, especially for the very first project with AI. Do it as small as possible but as tangible and concrete as possible. Do it quickly, and make sure you get positive output. Once you have positive output, you can accelerate.