Retail AI War Begins: Japan's Drugstores Launches Chatbot to Discuss Horoscopes

(CNA Taipei, July 4) "I'm in a bad mood at work today and didn't eat lunch. If I go exercise after work, will I faint?" Japan's Drugstores, the largest Japanese drugstore chain in Taiwan, will launch its "AI Pharmacist 2.0" service this year. This chatbot can discuss horoscopes and MBTI with you, and most importantly, it can "prescribe the right medicine" to be a rational and empathetic health guardian.

In late 2022, ChatGPT was officially released, and it began to sweep the globe the following year. All industries started to make friends with AI. Japan's Drugstores has been brewing this for a long time, and it all started with AI Pharmacist 1.0.

AI Pharmacist 1.0 was born in 2022. It operates very simply. It is a physical machine where customers can select questions, such as those related to colds, fatigue, and aches. After a few multiple-choice questions, the screen will display suitable medications and their shelf locations. However, the entire system was expensive. The hardware, electronic labels, and underlying software construction for one store cost as much as NT$50 million in AI investment.

"Now, a ChatGPT group plan costs only NT$500 per person," said Chen Rui-feng, the company's most enthusiastic employee in advocating for the company to embrace technology. With a background in pharmacy, he has experience in store operations, information systems, and marketing. He is now the Chairman's Special Assistant and Brand Licensing Manager. In an interview with CNA, he revealed the secrets behind Japan's Drugstores' AI decisions.

He said that in the past, whenever he wanted to do something, he had to spend a lot of effort explaining it to the boss. "When the boss saw the money and the bills, he would say, let's evaluate it again." Now, with the sharp decrease in AI model costs, an on-site AI image generation demonstration can be done during a product gift box packaging meeting. AI has already been silently embedded into the workflow.

In other words, when the results are directly presented, the boss has no choice but to nod in approval.

AI Saves Time Writing Copy; New Social Media Editors Focus on "Taste"

Chen Rui-feng stated that the company's initial reason for introducing AI was to solve new employee training problems. Currently, there are dedicated AI tools for internal corporate culture, legal and compliant product copy, and customer inquiries at stores. "Supervisors provide the weapons, and employees use these weapons."

AI also plays a crucial role in brainstorming copy for brand marketing and e-commerce departments. As long as the Japan's Drugstores style copy rewriting tool is used, no matter which social media editor is in charge, the output will not be "off the mark." However, he emphasized that personnel costs have not decreased because in the AI era, "taste" is becoming increasingly important. How to modify the "strong AI flavor" into "human language" is a test of the social media editor's skill.

He did not shy away from the fact that marketing is a high-risk profession for AI. "If you don't embrace it, you will truly be replaced." Now, with AI's assistance, the time to produce a piece of copy is only 1/20th to 1/10th of the original, which means producing 10 to 20 times more content at a faster pace. "Those who use AI will only be busier."

At the same time, numbers speak for themselves. In the past, copy generation cost the company NT$1 million per year, about NT$80,000 per month. Now, Gemini's monthly fee is only a few hundred yuan, but the output is 3 to 5 times more than before.

Furthermore, Chen Rui-feng pointed out that copy for food products cannot mention words like treatment, prevention, or symptom improvement, and cosmetics cannot claim to be anti-inflammatory or reduce scars, as these are considered exaggerated claims of efficacy. In the past, they relied on senior employees or the compliance department to review copy. Now, it can be avoided during the copy generation stage. However, when the marketing department can complete the verification themselves, it also means that "a large portion of the people doing this job are 'disappeared.'"

He listed the tools currently used by the company, including ChatGPT, Gemini, Notebook LM, and Grok, and Vids for videos. However, AI tools iterate very quickly, and large brands are caught in a multi-dimensional tug-of-war. "Therefore, supervisors must keep up closely, otherwise, if they keep using old (AI tools) foolishly, they will become weaker."

On the other hand, in the past, marketing work was directed by the headquarters' brain. He stated that by the end of the year, they plan to integrate four robots for copy ideation, image generation, compliance, and advertising into a common platform. In the future, store employees will also be able to participate in creative brainstorming. For different member profiles in various stores, such as department stores and street shops, the most precise marketing copy will be sent.

Store Staff Equipped with New AI Weapons to Find Answers with Consumers

In addition to optimizing internal work, AI is also the best partner for frontline store staff. According to industry statistics, the turnover rate for drugstore retail store staff is about 40%. However, "selling products" relies heavily on the knowledge level of the store staff, which is why "AI Pharmacist" needs to be upgraded. "We hope consumers can ask more questions."

For example, if a customer reports stomach pain, they need to ask follow-up questions like, "Does it hurt before eating, or after eating?" Because the two are different: one is stomach pain, and the other is a duodenal problem. As a pharmacist, Chen Rui-feng knows the difference, but it is difficult for ordinary store staff.

To address the knowledge gap, the underlying database of AI Pharmacist 2.0 is built on "package inserts" (drug instructions). The data includes indications, dosage, precautions, adverse reactions, etc. This "consulting advisor" includes three branches: "Symptom Consultation," "Drug Package Insert Inquiry," and "Related Sales." It is expected to be officially launched in the third quarter and deployed to store iPads. At that time, store partners can use this new weapon to interact with customers and find answers together. "If you know how to use it, you will actually become very good at selling products."

After internal stress testing, it will be gradually rolled out to Japan's Drugstores' 2 million members, moving towards a web-based approach. "You can also ask questions in the middle of the night."

AI Pharmacist 2.0 will also incorporate "human touch." Chen Rui-feng explained that consumers can share their feelings in the chat, such as "My boss scolded me today, and I didn't eat lunch." At this time, the robot might tell you that an empty stomach puts a burden on the body, causing excessive stomach acid, and prolonged fasting can lead to headaches. It will encourage the other person to eat something and take care of their body. The 2.0 version even considers incorporating content like horoscopes and MBTI, so that consumers feel understood during conversations.

Technical Cost Sweet Spot: Invest Now, Token Usage is the Next Battlefield

Japan's Drugstores' AI blueprint does not stop there. Chen Rui-feng revealed that the next step is to extend to beauty expertise and launch an "AI Beautician" that can identify fine lines and complexion, perform detailed skin condition analysis, and distinguish between bare faces, light makeup, and heavy makeup. However, the current technical costs are still high, and they need to continue observing and wait for the "sweet spot" to arrive.

He explained that the "technology" targeted by the retail industry is the so-called "third-eye beauty." The first eye refers to cutting-edge technology still in the research phase; the second eye technology is still expensive, but advanced industries are already using it; the retail industry will wait until the "third eye," meaning before the technology is officially popularized across all industries, they will start to engage with it.

Taking AI Pharmacist as an example, unlike the NT$50 million investment in 1.0, the reason 2.0 is cost-effective is precisely because of the year-on-year decrease in large model costs. The monthly fees for Gemini and GPT combined are about NT$10,000. "But is it cheap? I think the problem is further down the road." He explained that model development is now fast and relatively low-cost, but how many tokens are consumed during actual use is another matter. Therefore, they will initially set a monthly token limit of NT$100,000 and then adjust it based on the actual conversion rate.

Chen Rui-feng believes that companies are now divided into two categories: one is the pioneer who is interested in AI and discovers its mysteries through hands-on experience. "As long as you are willing to use it, you will cross over"; the other is the company that understands nothing but wants to jump on the AI train. In this case, as long as those who claim to offer related services charge exorbitant prices, companies will fall for it. "So you must try to understand."

Clearly, the AI wave has already swept across the board, changing the landscape of the workplace. Some worry about being replaced, while others are riding the wave and embracing new technologies. Chen Rui-feng, who is in charge of Japan's Drugstores' AI transformation, believes that only those who engage in lifelong learning can stand out. (Editor: Pan Yi-ching) 1150704

Choose to stand with the facts. Your every donation is a force to protect press freedom.

Download the CNA "One-Stop News" APP for real-time updates.

This website's text, images, and videos may not be reproduced, broadcast, transmitted, or used without authorization.