Japan’s working age population has started to decline. To maintain prosperity into the future it will be essential to improve productivity. Expectation for Artificial Intelligence (AI) technology to improve productivity is immense. By using AI to perform sophisticated and complex work previously considered possible only by humans, humans can concentrate on higher-value-added work and produce more output than before. AI inside Inc. is a cutting-edge company whose goal is to spread AI throughout all parts of society. The company’s first AI-OCR product, “DX Suite,” has acquired many users and it is attracting attention as a rapidly growing SaaS company. UTEC partner Keisuke Ide, who was working alongside the company before its listing, hones in on the visions of Taku Toguchi, Representative Director, President and CEO of AI inside.
06
STORY
AI
inside
UTEC
Creating a world where AI is inside all people and things
Representative Director,
President and CEO, AI inside Inc.
Taku Toguchi
UTEC Partner
Keisuke Ide
President and CEO, AI inside Inc. Taku Toguchi UTEC Partner Keisuke Ide
PROLOGUE
SCROLL or CLICK
SECTION01 : Starting a business imagining the future 200 years from now.
SCROLL or CLICK
SECTION01
Starting a business imagining the future 200 years from now.
AI inside Inc. was established in August 2015. However, Mr. Toguchi envisioned he would develop an AI business more than 10 years prior.
- Ide:
- You have told me that you made a 200-year timeline for the future when you started your business. Can you tell me about the timeline again?
- Toguchi:
- Yes, to put it simply, I thought if one could predict the future, it would be easy to build a successful business. With that in mind, when I graduated from high school I made my own timeline for the next 200 years. The most significant themes that came up were AI and space.
But, when I started my first business in 2004, computer performance was still primitive that there was little I could do with AI. I was optimistic however, as Moore’s law stated computing performance would double roughly every two years. I guessed processor chip performance with the current architecture would reach its limit around 2018 to 2020. So even though I could not start an AI business right away in 2004, I started a business, basically in preparation to build an AI business when computing performance caught up with AI’s requirements.
My first startup was a gourmet restaurant directory website, with an AI essence. Neural network technology has been around for a long time and natural language analysis has been quite successful. Using that, we incorporated a function that is now called a chatbot.
- Ide:
- With the technical breakthroughs regarding deep neural networks in 2012, the possibilities of its applications expanded rapidly. At that time, many startups were in the image recognition field. How did you come to focus on character recognition within the domain of image recognition?
- Toguchi:
- We did conduct various tests around image recognition, to see what was technically possible, and decided to pursue character recognition business. The reason is that Japan’s working population had already started to decline, and improving productivity was becoming a big issue for companies. Our research suggested the work incurring the biggest outsourcing expenses for companies was data entry. It was clear there was a growing need to automate that work. Having made that decision, in 2014 we conducted R&D on character recognition and started to have customer visits to obtain actual work documents as training data to teach our AI engine.
SECTION02 : At first, Mr. Toguchi did not intend to obtain VC investment.
SCROLL or CLICK
SECTION02
At first, Mr. Toguchi did not intend to obtain VC investment.
Mr. Toguchi, as he and his team worked hard to improve the accuracy of his technology, met Mr. Ide at a conference.
- Ide:
- I first learned about AI inside at NVIDIA’s “GTC 2017” conference. AI inside had been certified by NVIDIA and participated in the startup pitch there. The first thing that surprised me when I watched Mr. Toguchi’s presentation was the accuracy of character recognition. My own handwriting is terrible, but the technology could recognize even worse handwriting than mine, which was amazing. (laughs)
But, more than the technology itself, the fact that there was clear priority in productization was very important for me. AI was attracting attention at that time and AI startups were being born one after another. However, most of them were consulting companies that provided AI technology to companies. AI inside, on the other hand, was focused on developing a product, and hence I really wanted to hear about it in detail. So, we exchanged contact information right away, but you were hesitant at first, weren’t you? (laughs)
- Toguchi:
- Haha. That’s because at the time I was thinking “Let’s not accept any VC investments.” Around the time I met you, we had already formed a capital and business alliance with an insurance company, and talks were also advancing behind the scenes regarding a business alliance with a large company in BPO. So we didn’t really need any funding.
To be honest, I actually did not intend to talk with you. If we partnered with a business company, even if the IPO came a little later we could help their business with our technology.I thought if we received VC investment, we would have to meet VC’s exit expectations. I was not confident I could handle that pressure, and I was also worried I would have to make impossible decisions to meet expectations.
But, after talking a few times, I thought UTEC was different from my perception of VCs. I imagined most VCs enter with questions like “What’s your capital structure?” or “Send me your DD material.” right? However, you mainly asked “What is your vision for your company?” and “Tell me how your technology gets you closer to your vision.” My impression had changed greatly to the point where I thought UTEC would support us in various ways in the areas we are lacking.
- Toguchi:
- Haha. That’s because at the time I was thinking “Let’s not accept any VC investments.” Around the time I met you, we had already formed a capital and business alliance with an insurance company, and talks were also advancing behind the scenes regarding a business alliance with a large company in BPO. So we didn’t really need any funding.
To be honest, I actually did not intend to talk with you. If we partnered with a business company, even if the IPO came a little later we could help their business with our technology.I thought if we received VC investment, we would have to meet VC’s exit expectations. I was not confident I could handle that pressure, and I was also worried I would have to make impossible decisions to meet expectations.
But, after talking a few times, I thought UTEC was different from my perception of VCs. I imagined most VCs enter with questions like “What’s your capital structure?” or “Send me your DD material,” right? However, you mainly asked “What is your vision for your company?” and “Tell me how your technology gets you closer to your vision ” My impression had changed greatly to the point where I thought UTEC would support us in various ways in the areas we are lacking.
NEXT
SCROLL or CLICK
PROBLEM
AI inside procured funds in its first VC round with UTEC as the lead investor in July 2018. While the number of users grew rapidly, the company needed to work on many issues for the next stage of its development.
SECTION03 : Laying the foundations for global expansion
SCROLL or CLICK
SECTION03
Laying the foundations for global expansion
After the release of its AI-OCR product, “DX Suite,” AI inside has expanded its business steadily, but what support did AI inside receive from UTEC?
- Ide:
- AI inside was the most hassle-free portfolio company I have ever been involved with. For example, there are quite a few startups that have superior technology and present themselves as “the solution for everything.” However, when you consider the costs of product development and sales, marketing in all business segments is high risk. In this regard, you had already refined your target segments to insurance and BPO. I called and organized an executive and sales retreat right after my investment, and I did no more than providing frameworks for productization and marketing strategies. It was the smoothest ran retreat.
I appreciate UTEC supporting AI inside in building the equity story for the IPO. An convinsing story is required to get attention from the market. Within that, global expansion in particular attracts interest from Japanese market participants. In order to build the story, Keisuke organized a bunch of meetings in Taiwan where two of us visited potential customers and partners.
- Toguchi:
- If you have the training data, expansion into other languages is not difficult so I thought from the beginning there was no option but to go global. But at that time we had just started in Japan so I could not draw a specific story. Going to Taiwan together and getting ideas on how the technology could be adapted in other languages was significant. That business trip was the trigger for us to expand our business into Taiwan, Thailand and Vietnam where we have business now.
You say that you did not do much, but you supported us in dealing with securities companies and the Tokyo Stock Exchange towards an IPO, and the HR team at UTEC has also supported us in recruitment work. It was encouraging to be able to receive advice from you as a board director in everything from marketing and finance to corporate matters.
- Ide:
- The IPO was in December 2019, as originally planned. I think the reason it went on schedule was because the management team understood from early on the difference between technology and product. If you look at AI-OCR from a technological perspective, it inevitably tends to become competition over recognition rates. However, in addition to improving the accuracy of recognition, you developed and patented a system around secure transmission of personal data. However high the recognition rate is, from the customer’s point of view they would not use products that do not provide guaranteed security. I think it was precisely because you had such business sense that you were able to grow rapidly. VCs do not really have much of a role to play with startups led by superior management teams. (laughs)
FINAL SECTION : From a SaaS company to an AI platform
SCROLL or CLICK
FINAL SECTION From a SaaS company to an AI platform
FINAL SECTION
From a SaaS company to an AI platform
AI-OCR is just one trigger for spreading AI around the world. The company’s vision is “AI inside X,” disseminateing AI to everybody and everything. As a mean of achieving this, the company has newly developed “AI inside Cube,” an edge computer that realizes AI operations on-premise, and the “Learning Center,” which enables non AI experts to develop AI without coding.
- Toguchi:
- Our goal is to be an AI platform, rather than a SaaS company. First, we will create an environment within which AI can be developed and operated simply by business domain experts, not AI experts. And then we will create a marketplace where people can share the AI they have developed with others. “AI inside Cube” and “Learning Center” are products for realizing a world where anybody can create and use AI.
Actually, we communicated this idea to UTEC before the IPO. Because “AI inside Cube” is a hardware, it is riskier than software. I was worried you might object to the idea and asked me to concentrate on the SaaS part of the business. However, you understood how the on-premise hardware would gets us closer to the vision, and approved the initiative as a board member. That was very encouraging.
- Ide:
- It is risky to pursue hardware in the year leading to the IPO, when SaaS metrics are scrutinized by the underwriter as well as the market. But we knew the AI-OCR engine was well built to be ported on hardware, and I was genuinely excited about the marketplace idea.
- Toguchi:
- It is not easy to become the world’s number one SaaS company; however, with an AI platform, we can create a market where non-AI experts can exchange AI engines from various business domains. I would like to hit the accelerator even harder now to realize a world in which AI spreads naturally throughout society.