Workforce shortages, supply chain snarls, and overseas competition… these challenges are commonly among those we hear about from the manufacturers whose stories we tell in each edition of CompanyWeek's Manufacturing Report. While the COVID-19 pandemic has only worsened them over the last two years, Liaw believes that AI-driven automation may provide the much-needed solution -- if U.S. manufacturers are willing to embrace it.
In this Q&A, edited for length and clarity, we chat about her vision and how her company, Bellspring Automation, is working to bring it to fruition.
CompanyWeek: Analysts have often predicted that the economy of China -- the biggest competitor many U.S. manufacturers face -- will outpace that of the U.S. sometime in the next decade. Why do you think this is, and what role is China's investment in automation technology playing in this scenario?
Nancy Liaw: China, as we know, has really low labor costs. But even with that cheap labor, they have very actively been pursuing fully autonomous manufacturing since at least 2010. This means they can sell products at lower prices.
Here in the U.S., manufacturers have been reluctant to abandon the past way of doing things. They've taken shortcuts and quickly imported whatever was available globally at the lowest cost. Ocean freight was cheap, so they bought supplies from developing countries and used them. But since COVID, many have seen that is not a good way to operate now or for the future.
Once U.S. manufacturers determine to stop taking that shortcut and start focusing on how we need to be able to supply our own domestic needs at lower costs, they will also begin to pursue more automation as a tool for doing so.
CW: When many people think about companies investing in automation, they worry that the U.S. manufacturing sector will lose jobs. Is this a valid fear?
NL: No, that's actually wrong. For example, when my clients expand an operation from semi-manual to fully automated, it actually creates 25 percent more jobs in that region. Those jobs are not created by the automation itself but by the positive consequences of automation. As a result of investing in automation, the demand, market share, and value of the company increases -- thus increasing jobs. And not only do jobs increase by 25 percent on average, but those jobs are also generally higher paying.
CW: Traditionally, where have U.S. companies gone wrong when trying to incorporate automation in their manufacturing processes?
NL: They typically start with the hardware. They see videos of robots from Japan, like the huge yellow robotic arms, and they buy them and then have to recruit people from overseas to figure out the integration. Instead, they should start with software. We should use the latest artificial intelligence-driven software to first design and lay out the manufacturing plant to achieve the efficiency target we desire.
For example, let's say we want to cut our costs by 65 percent over the next ten years. How are we going to achieve it? We could use AI to look at our inventory system, logistics, and the manufacturing process. It could analyze the systems we have now and determine what we could adopt new. It could guide us and help us make a decision.
When I talk about artificial intelligence in manufacturing, I'm talking about a huge database with algorithms running behind the scenes. It's a fully autonomous database running constantly, 24/7, collecting data and computing statistical analysis, then applying those statistics to improve the efficiency of the machines. It leads all the systems -- and the manufacturing equipment involved -- in moving up to a higher level of efficiency at a lower cost.
AI can dramatically reduce our labor costs and -- as mentioned earlier -- create more and higher paying jobs. A lot of them will probably be on the quality control and management side. And I think it will all help U.S. manufacturers to be able to sell their products to Asia, to Europe. It's very possible. It may not happen right away, but hopefully three to five years from now we'll see it more and more.
CW: In general, where within their organizations would you advise manufacturers to first look at automation and why?
NL: Warehouses are the first place to automate using AI. If you integrate robots into a warehouse, you can build it high -- with many stories -- instead of spreading it out over a large piece of land. This actually generates energy savings in a number of ways. For example, during the summertime, a building like that is going to only be hot on the top floor. Anywhere under the top floor will stay comparatively cool. When it’s cold, you don’t need to warm up an entire 50,000-square-foot warehouse because only the bottom portion is touching the frozen land.
At one of our factories in Taiwan, we have invented a battery-operated, huge, track traversing robot. It's pretty heavy itself, but it can carry up to two tons along the inventory shelving. The robot climbs the shelving, and when it comes down, it recharges itself like an electric car engine. This particular robot is battery operated. It's especially good in a building with lots of levels because it can pick up heavy boxes, heavy items, and use the momentum, or gravity, to recharge its own battery on the way down.
The second place to automate is logistics and transportation between warehouses. One of the systems we're promoting right now is called Intelligent Routing. It constantly calculates the route for a truck carrying goods from Point A to Point B to Point C, including the gasoline price and the traffic. It's all written in the database with certain efficiency rules. This AI definitely reduces logistics costs, trucking hours, gasoline, distance, and time.
CW: What do you see as the biggest challenge to the adoption of AI-driven automation by manufacturers here in the U.S.?
NL: I would say the hardest part is not finding the money -- or investment in the technology. It's the system integration. System integration fundamentally means system change from the old to the new. How are we going to approach that? As I mentioned earlier, a lot of times people look at the hardware first. They see the fancy looking robot and think, 'Oh, that’s something we can use.' But what about their current system? Are they going to abandon it? Are they going to run both the old system and the new system separately? That would be very hard and not very cost efficient. Artificial intelligence -- and deep machine learning -- is needed to help us process this transition.
CW: How does Bellspring Automation work with clients to minimize these system integration challenges?
In a way, it's still conceptual but still very realistic. I let the clients know that they don't have to abandon their own systems, they just need to modify them. Start with the software and let it connect the old with the newer version of operation based on its statistical analysis.
For example, one of our clients is a pharmaceutical lab. I visited their site and noticed that they hire many laborers to take test tubes filled with testing solutions inside and put them into bags. I said, 'Oh, that's a lot of money and a lot of potential problems there with quality control, and consistency, and so many staff in the same place during COVID.'
So, I recommended automating the process as a system. We're now building them a production line for their lab packaging that can work with a variety of sizes of tubes and bottles. We're also trying to automate their very labor-intense capping system. We're building a machine that can handle different size caps and specific torque levels. They're also going through the process right now to get approval to let us automate their filling process as well.
All the young men and women -- scientists with degrees -- who were doing these tasks can then be moved to quality control, like monitoring the processes and testing. Instead of paying a lot of highly skilled personnel to do low level jobs, they can make better use of their talent.