Suzuki embraces Azure OpenAI Service, supercharges AI use with five apps, and sparks ideas for AI in business

Becoming an AI early adopter, responding to complex linguistic needs

Founded as Suzuki Loom Works in 1909, Suzuki has led the motorization of Japan since the conclusion of World War II. Today, the company develops, manufactures, and distributes automobiles, motorcycles, and marine products among a wide range of goods loved across the globe. In January 2023, it announced its growth strategy toward FY2030, aiming to create customer-focused, value-packed products and building a carbon-neutral society based on the three actuals principle which prioritizes actual place, actual thing, and actual object over theoretical discussions.

Generative AI is an increasingly powerful tool in this endeavor.

“We started working on AI around 2018,” recalls Yoshihiro Ugai, Managing Officer and Executive General Manager of Suzuki’s Global IT. “In 2019, we became keenly aware of the importance of data analysis and sent about 10 members on an Osaka University data analytics course. We initially used AI in technical divisions, but usage has spread across other departments.”

To further these efforts, Suzuki established the Data Analysis Promotion Group in January 2022, creating a core team to drive the use of AI throughout the company.

“One major problem was that despite our high linguistic needs, we didn't know exactly what to do,” says Daisuke Hata, Group Manager of the Data Analysis Promotion Group. The company had already tried AI but found it difficult to create something tailored to the needs of user departments.

“User departments wanted to leverage historic knowledge stored in a text format; to sort and organize files written by multiple people in various parts of the company; and to summarize survey results,” Akira Nonaka, Department General Manager of Suzuki’s Digitalization Promotion Department. “Our surveys allowed open-ended answers, and we received around 10,000 responses per month to customer surveys alone, which limits manual analysis.”

Despite these hurdles, Suzuki forged ahead with ChatGPT after a broad investigation.

“We examined a variety of other text applications but noted the rise of ChatGPT in December 2022 and saw great potential,” says Hata. “So, we decided to use it first.”

ChatGPT proved to be quite useful. However, according to Hata, there were several problems for corporate use, including security concerns over the information entered (prompts) being shared outside the company.

In March 2023, Microsoft released a preview version of Azure OpenAI Service, an Azure solution delivering the same generative AI capabilities as ChatGPT. Suzuki immediately adopted the new service.

Adopting Azure OpenAI Service for data security, driving company-wide use with five multipurpose apps

“Using Azure-based services meant we wouldn’t have to worry about data protection,” Hata says. “And solving two additional challenges would open a path to deployment across the entire company.”

The first of these challenges was to formulate and disseminate guidelines on company-wide use. These guidelines emphasized avoiding copyright infringement and scrutinizing the replies of generative AI. “We knew about hallucinations, or generative AI lies,” explains Hata. “Therefore, instead of making decisions based solely on generative AI replies, we decided to use the tool on the premise that our staff could independently verify answers under the principle of three actuals¹.”

^[1] Known as the genba, genbutsu, genjitsu principle in Japan, this style of behavior prioritizes actual place, actual thing, and actual object over theoretical discussions.

The second challenge was to prepare a user interface (UI) for employees without advanced IT literacy to use easily.

“We decided to create a web application and run it on the Data Analysis Promotion Group’s on-premises server,” says Shintaro Oashi, Assistant Manager of the Data Analysis Promotion Group. “We wrote the front-end in HTML5 and JavaScript, accessing the Azure OpenAI Service API with Python. Microsoft released the API on March 9, 2023, and we completed our first app in about two weeks."

Meanwhile, at an executive committee, they reported that they were developing and would like to release a secure generative AI environment using Azure OpenAI Service. After having approved the establishment of guidelines emphasizing three actuals and confirmation of the environment’s data security capabilities, Suzuki commenced a company-wide rollout on March 21.

Oashi’s first application, NAGI, accompanied the rollout. One week later, Suzuki released NAMI, an interactive app that answers direct prompts in the same way as ChatGPT. According to Oashi, “Our major focus was extreme ease of use. For example, NAGI is a question-answer chat app featuring user-friendly buttons for common prompts such as 'translate' and 'summarize'.”

In addition to NAGI and NAMI, Suzuki has created several web apps, including one that gives ChatGPT pseudo-personalities that can debate each other.

 
“All of our apps are useful, but the most interesting one features ChatGPT pseudo-personalities that debate each other,” says Ugai. “There are 26 personalities consisting of different nationalities, genders, and occupations, such as start-up executives, lawyers, consultants, housewives, elementary school students, and YouTubers. When you place them in virtual meetings, they often generate surprising perspectives. You can have one-on-one conversations with them, using them as a foreign language teacher or to receive advice. We’re also expecting generative AI to produce fresh ideas.”

Suzuki is also working to train executives in digital transformation through generative AI. “We had generative AI create program code after having executives give instructions in a dialog-based format,” Ugai explains. “If the instructions were wrong, the code wouldn’t work correctly. This taught the executives that giving instructions to subordinates works in the same way.”

The names NAGI and NAMI are based on gods that appear in Nihon Shoki, the second-oldest book of Japanese history. Oashi hopes the app will give birth to various ideas and innovations, similarly to how the book describes Izanagi (NAGI) and Izanami (NAMI) as the parents of all Japanese gods.

Company-wide deployment before end of fiscal year, Microsoft support to engineer more ideas for use

Although Suzuki has only created all-purpose apps to date, the company is already applying them to specific work. One such app streamlines recruitment in India.

“Before recruitment interviews in India, we have about two days to screen documents, but we sometimes get over 100 applicants for a few positions,” says Nonaka. “We were consulted by the recruiter about using generative AI, so we quantified the suitability of leading applicants and mapped the values against other applicants. We were able to screen the documents, narrow the candidates down to three to five people, and recruit exactly the kind of employee we wanted: the recruitment manager was astounded.” 

Suzuki is also using generative AI in several other initiatives, including searching internal documents with a question-based format and creating meeting minutes by converting and summarizing audio to text. By having employees exchange ideas on mastering generative AI, Suzuki expects to cut labor on simple and routine tasks by 80 percent.

“We’re coming up with many ideas internally regarding use of generative AI, and support from Microsoft to realize these ideas is proving to be another major advantage of Azure OpenAI Service,” says Hata. “Microsoft has provided several key tips on searching internal documents.”

According to Nonaka, another appeal of Azure OpenAI Service was an absence of additional implementation costs: “We greenlit the company-wide rollout in March 2023. Our annual budget is usually gone at this point, so we carry out new initiatives in the next fiscal year. Azure OpenAI Service was available within our existing Azure license, so there was no need to budget for a new service. We were able to begin the rollout immediately.”

Suzuki expects generative AI to steadily become an entrenched part of daily work, with collaboration between humans and AI dramatically boosting productivity. Soon, the company will move generative AI out of the office and into vehicles to provide customers with information through connected services. One example is a car navigation app that provides voice-guided assistance to tourists. Azure OpenAI Service supports many languages—including Hindi—making it easy to deploy in India, one of Suzuki’s main markets.

“The automobile industry is confronting a once-in-a-century upheaval,” says Ugai. “We must tackle many novel challenges, but there aren’t enough personnel to do so. To complete new initiatives with a limited workforce, we must streamline work, identify jobs needing human workers, and reduce unnecessary tasks. AI is an effective tool for accomplishing this. Suzuki will master new technologies to enhance human potential.”