Commentary: How Japan’s ABEJA helps large companies operationalize AI, machine learning

How ABEJA solves problems for customers

I asked Aya Zook, head of enterprise business development at ABEJA, “What is the problem that ABEJA solves for its customers? Who is the typical customer?”

“Primarily, we help solve our customers’ problems through two means — ABEJA Platform and a Software-as-a-Service (SaaS) offering,” he said. “To successfully utilize AI, there are a number of issues to consider from data acquisition to retraining models.”

He continued: “ABEJA Platform streamlines the AI ​​implementation and operation processes by solving these issues, which allows engineers to focus on algorithm development. But it’s often not good enough to just provide the tools to deploy AI solutions into an operation. This is why we also provide comprehensive support from strategy planning to implementation to help companies realize true digital transformation. While we are customer agnostic as long as usable data is available and the problem is uniquely addressable by AI, over 50% of our revenue comes from manufacturers, logistics and infrastructure companies.”

Explaining why ABEJA developed a SaaS product, Zook said, “Our SaaS business focuses on retailers. Previously profitability of stores often depended on the experience and intuition of the staff instead of concrete data like visitor count and customer attributes.”

Delving deeper into how the SaaS product works for retailers, he said: “ABEJA Insight for Retail analyzes data from IoT devices such as network cameras and infrared sensors, and from actual store visits such as number of visitors, number of people passing in front of the store, age and gender estimation, repeat customer estimation, and flow line analysis. It visualizes customer behavior from entry to purchase, which enables the store to have a data-driven method to track the effectiveness of different marketing activities put in place to drive sales. To date, our SaaS solution has been implemented in over 800 stores across Japan.”

FreightWaves readers who have been following this #AIinSupplyChain series since July 7 are by now familiar with the difficulties that companies in traditional industries encounter when they try to incorporate AI and machine learning into their operations: First, obtaining data of sufficient quality and quantity with which to train models is difficult. Second, there’s an acute shortage of people with the knowledge and experience to implement AI and machine learning techniques within companies in mature industries. Lastly, executives find it difficult to quantify what it means for their companies to adopt the technology, and as a result such initiatives often fall by the wayside.

ABEJA’s approach emphasizes keeping humans in the loop

I asked, “What is the secret sauce that makes ABEJA successful? What is unique about your approach? Deep learning seems to be all the rage these days. Does ABEJA use a form of deep learning? Reinforcement learning? Supervised learning? Unsupervised learning? Federated learning? How do you handle the lack of high quality data for AI and machine learning applied to legacy industries?”

Zook said: “When helping implement AI solutions into companies, we recommend that humans supplement the AI until enough high-quality data has been generated in order to retrain the model and create a feedback loop to continuously improve its accuracy over time. From our deep experience of implementing AI into over 100 companies across multiple industries, we learned one of the keys to success is to operationalize it as soon as possible so you can start collecting real-world data quicker, thus reaching full or semi-full automation sooner. This approach allows companies lacking high-quality data to start accumulating data immediately.”

Emphasizing ABEJA’s flexibility, he added, “The approach we take, including what form of AI to apply, depends on the customer’s problem. Eradicating the problem most effectively is our goal, less so how.”

#AIinSupplyChain success is all about facilitating customer success

I asked Zook if there were any customer stories he could share with FreightWaves readers.

“Musashi Seimitsu Industry is a car parts manufacturer we helped implement a deep learning solution to automate their anomaly detection process,” he said. “They faced three problems. First, detection was done 100% manually by veteran workers, which wasn’t scalable since it depended on deep experience and intuition. The second problem they faced was inconsistency in the inspection quality due to varying degrees of experience. The third problem was the amount of heavy workload required to conduct visual inspection manually.”

Musashi exports its products to customers in North America, Europe, South America and Asia. It lists BMW, Daimler, VW, Ford, Fiat Chrysler, Jaguar Land Rover, Honda, Kawasaki, Suzuki, Daihatsu, Mitsubishi and Subaru among its customers.

Zook added, “By helping Musashi collect and annotate 86,000 photos, we managed to help them achieve 97-99% accuracy at two-second intervals per inspection, drastically reducing inspection time and human labor involved while transforming the process into a highly scalable operation with almost no depencies on time-intensive experience-building.”

One of the things I did not expect to find once I started studying ABEJA is the depth AND variety of its experience. As mentioned earlier, the company does a lot of work in manufacturing, logistics and infrastructure. It also has customers in transportation, energy, health and pharmaceuticals, agriculture, finance, property development, and e-commerce.

ABEJA’s customers include Daikin, the world’s largest manufacturer of air conditioners; Komatsu, the world’s second-largest manufacturer of heavy construction and mining machinery; SMRT, the leading multimodal transportation operator in Singapore; DENSO, the world’s second-largest manufacturer of automotive parts; CHUBU Electric Power, Japan’s second-largest energy company; and AEON, Japan’s largest retail chain.

#AIinSupplyChain success is also about walking in customers’ shoes

There’s a third part of ABEJA’s business that we have not highlighted as much so far, but that I think is just as important as ABEJA Platform — the Platform-as-a-Service (PaaS) offering, and ABEJA Insight — the Software-as-a-Service offering, that Zook described earlier.

ABEJA also offers an AI consulting service that enables it to work with customers to examine problems, and then to develop tailored solutions for that customer.

The reason this is important is that each customer that considers using ABEJA’s PaaS or SaaS offerings is unique from every other company in the way that its operations and supply chain are organized. So it makes sense that a set-it-and-forget-it approach is unlikely to be effective for very long, if at all.

In this sense, ABEJA AI Consulting functions as a thought partner, helping executives at large, mature corporations seeking to implement AI within their extended supply chain and operations sort through which approach makes the most sense for each company specifically, and then implementing that approach on ABEJA’s SaaS or PaaS offerings. The analogy that comes to my mind is that of the training wheels that are sometimes used when young children are learning to ride a bicycle; in a sense, corporations that want to become AI-enabled can rely on ABEJA AI Consulting in much the same way as children rely on training wheels as they learn to become proficient at riding a bicycle.

Ultimately, the goal must be that these corporations become nearly self-sufficient, requiring no permanent, ongoing assistance from ABEJA’s consulting services, but becoming ongoing customers of its PaaS and SaaS products.

Moreover, like some of the other startups that we have already encountered in this series, ABEJA has developed proprietary tools to ensure that its AI Consulting engagements do not metastasize into never-ending ordeals. Data that can be used to train AI and machine learning models is usually the stumbling block that startups like ABEJA and the customers that wish to work with them encounter. To this end, ABEJA has developed tools and processes that speed up the process of developing and annotating data that the customer can use to develop and train its models.

The advantage of AI Consulting for ABEJA is that this enables it to learn directly from its customers, becoming better at incorporating the insights that it accumulates from these relatively small, highly tailored consulting engagements into general frameworks, methods and models that it then bakes into the PaaS and SaaS products

Many venture capitalists are quick to express their dislike of startups generating revenues from consulting services. The complaint is that revenue from consulting does not scale efficiently. However, for startups like ABEJA, a services or consulting offering is a highly effective way of gaining deeper insights into the problems that customers need to solve with the technology that the startup seeks to bring to market. Of course, the goal is always that at a foreseeable point in the future, consulting revenues should be dwarfed by sales generated through other revenue models in the startup’s business model.

The competitive landscape

I asked Zook if there were other startups doing something similar to what ABEJA does.

“ABEJA is unique in that we have two distinct business portfolios, the PaaS business and the SaaS business,” he said. “While there are existing competitors like Element.AI and H2O.ai on the PaaS side and RetailNext in SaaS, our deep experience from implementing AI solutions into 100-plus companies across multiple industries gives us a unique advantage because at the end of day, the key to getting AI right is speed to market to learn what works — and doesn’t — quickly.”

ABEJA’s investors induce envy

During our conversation, Zook pointed out that there aren’t many startups that can count both Nvidia and Google as investors. Nvidia is notable because it is the leading designer and manufacturer of the processing units on which a lot of AI and machine learning systems are deployed. Readers of FreightWaves are undoubtedly very familiar with Google but are likely much less familiar with Nvidia.

Toshiba is an investor too. Regular readers of this column may remember that on Feb. 17, FreightWaves ran Commentary: Toshiba’s simulated bifurcation machines may optimize your supply chain in which I described how new technology from Toshiba might find its way into systems designed to optimize industrial supply chains. In that article, I also pointed out that the new technology was already available on AWS for anyone who wanted to test it. So, from my perspective it is interesting to note that ABEJA counts AWS as one of its notable business partners.

Salesforce.com is an investor as well. This is interesting because of Salesforce’s Einstein AI platform that seeks to empower salespeople in various industries with AI-enabled insights with which to empower them to close more sales.

The other major investors listed on ABEJA’s website are no less impressive — altogether, 17 major shareholders are listed.

For example, Itochu Corp. is one of Japan’s largest general trading companies. I have already highlighted that Daikin is a customer, but Daikin is also an investor. Musashi Seimitsu, the automotive parts manufacturer, is an investor as well as a customer. Mitsubishi UFJ Financial Group, one of the world’s largest banking and financial services groups, is an investor too. The other investor I will highlight is Topcon Corp., which operates in positioning, smart infrastructure and eye care.

I am fascinated by ABEJA’s roster of investors because of the observation Zook has already made that “the key to getting AI right is speed to market to learn what works — and doesn’t — quickly.” Obviously, customers who are investors too have a deep-seated incentive to see ABEJA succeed.

This is an approach I have described to some of my friends who are independent venture capital fund managers, and who ordinarily have less than charitable opinions about collaborating on investments with corporate venture capital funds.

In the effort to bring nascent innovations and technologies that can transform industrial supply chain operations to market, I think corporate venture capital is a plus more often than a disadvantage. In making that assertion, I am assuming the independent venture capitalists will exercise enough vigilance to ensure that startups do not get hoodwinked into accepting predatory terms from corporate venture capital funds.

What’s next for ABEJA?

ABEJA has three offices, one each in Tokyo, Singapore and San Francisco. The company has raised about $60 million so far. In addition to being highlighted by CBInsights in 2019, ABEJA was named by the World Economic Forum to its 2020 Cohort of Technology Pioneers.

It seems obvious to me that having established a relatively strong foothold in its home market of Japan, ABEJA is now poised to pursue global expansion. It can take the lessons it has learned while helping large Japanese companies operationalize AI and machine learning to pursue customers in North America, Europe and other parts of Asia.

If you are a team working on innovations that you believe have the potential to significantly refashion global supply chains, we’d love to tell your story in FreightWaves. I am easy to reach on LinkedIn and Twitter. Alternatively, you can reach out to any member of the editorial team at FreightWaves at [email protected].

Dig deeper into the #AIinSupplyChain Series with FreightWaves.

●     Commentary: Optimal Dynamics – the decision layer of logistics?

●     Commentary: Combine optimization, machine learning and simulation to move freight

●     Commentary: SmartHop brings AI to owner-operators and brokers

●     Commentary: Optimizing a truck fleet using artificial intelligence

●     Commentary: FleetOps tries to solve data fragmentation issues in trucking

●     Commentary: Bulgaria’s Transmetrics uses augmented intelligence to help customers

●     Commentary: Applying AI to decision-making in shipping and commodities markets

●     Commentary: The enabling technologies for the factories of the future

●     Commentary: The enabling technologies for the networks of the future

●     Commentary: Understanding the data issues that slow adoption of industrial AI

●     Commentary: How AI and machine learning improve supply chain visibility, shipping insurance

●     Commentary: How AI, machine learning are streamlining workflows in freight forwarding, customs brokerage

●     Commentary: Can AI and machine learning improve the economy?

●     Commentary: Savitude and StyleSage leverage AI, machine learning in fashion retail

Author’s disclosure: I am not an investor in any early stage startups mentioned in this article, either personally or through REFASHIOND Ventures. I have no other financial relationship with any entities mentioned in this article.