#INTRO

As we move into the age of agentic commerce, the touchpoint where we meet customers is no longer a single storefront. Customers now ask external AI first, consult with AI within our own channels, and sometimes carry a concern that began in our store out to external products or services to see it resolved.

From our brands' and channels' perspective, this flow breaks down into three types of AI agents. At its AI & Data Innovation Day event, AWS(Amazon Web Services) described these three agents using the terms Incoming Agent, Onsite Agent, and Outgoing Agent.

First, the Incoming Agent is the external AI that customers encounter before they ever reach our channels. This is when an external AI such as ChatGPT, Gemini, Claude, or Perplexity answers a customer’s question and, in doing so, recommends our brand, products, or content and links back to our brand-owned online store. Customers no longer just type keywords into a search bar. They ask AI first, with questions like “Recommend a cream for sensitive skin,” “Tell me a good product for starting retinol,” or “What cushion is good for dark spots?” and discover a brand by following the answer. This is why so many people today are paying attention to Generative Engine Optimization (GEO).

The On-site Agent is our own agent, the one customers meet within our brand-owned online store or brand channels. AMORE CHAT, the AI beauty counselor, is exactly this. When customers ask, within our channels, “Will this suit my skin?” “Which of these two is better?” “Can I use this together with that product?” or “Why isn’t my coupon applying?” The On-site Agent helps them explore their options and decide right then and there. The On-site Agent is also where we can most clearly observe customer questions and context. It is the place where valuable data accumulates, allowing us to hear firsthand what customers are actually worried about and where they hesitate within our own channels.

Finally, the Outgoing Agent is the agent that connects customers to outside options when a need they bring to our brand-owned online store cannot be resolved with our own products or within our own channels. For example, if a customer asks about external products or services, such as other companies’ products that can be used alongside ours, dermatology procedures and skin care methods, or fashion items that pair well with a look, our agent can act as an extended touchpoint, providing outside information and options to meet the customer's broader needs.

Today, among these three agents, I’d like to introduce the real voices of customers captured in ‘AMORE CHAT, the AI beauty counselor,’ the On-site Agent on AMORE MALL.

A Look Inside AMORE CHAT’s Customer Conversations

AMORE CHAT, the AI beauty counselor, is a beauty agent that talks directly with customers. It currently accompanies customers on their beauty journey from the AMORE MALL home screen, My Page, and Customer Center, providing services such as skin diagnosis, personalized recommendations, product consultation, and customer service inquiries.

To run this service well, understanding the questions customers ask matters more than anything else. That’s what lets us provide better answers and accompany customers naturally on their beauty journey. Customer questions were sometimes very vague, and sometimes very specific. Looking closely at the specific ones, it becomes clear which products a customer is hesitating over, what skin concerns they have, and which brands and products they are comparing. From here, let’s look at the results of analyzing the actual questions asked by the 15,000 customers who used AMORE CHAT over the past three months.

Before We Begin: Who Uses AMORE CHAT

Before looking at the questions themselves, we first looked at who our customers are. Given that a conversational chatbot is the interface, we expected a younger skew, but the actual age and gender distributions were nearly identical to those of AMORE MALL’s overall customer base. This suggests that even if the AMORE CHAT service were expanded to other D2C(Direct to Customer) sites or brand channels, it would likely not create an entirely new, independent user base; rather, the existing core customers of that channel would carry over to using AMORE CHAT.

Looking at Customers’ Main Questions: ① Key Categories

Looking at the questions that came into AMORE CHAT by intent, the largest share involved product suitability based on skin type and concerns. Customers wanted to confirm whether a given product would suit their skin, based on their own skin condition. The second-largest category was product comparison questions and requests for recommendations, showing that customers were not simply checking product information but asking for help choosing the product that best fit their own situation.

Commerce-related questions such as order/payment, delivery/pickup, and price/discount/benefits followed; however, by share, skin type/suitability and comparison/recommendation questions stood out far more. Rather than expecting simple FAQ answers from AMORE CHAT, customers expect it to reinterpret products for them based on their own skin condition and purchase context. Below is a table organizing customers’ actual question messages by intent.

Looking at Customers’ Main Questions: ② Skin Concerns

Looking at the top skin-concern keywords customers mentioned, concerns tied to overall skin condition, such as dryness and dehydration, made up a large share. At the same time, concerns about anti-aging care, such as melasma/dark spots/pigmentation, dullness/tone, and wrinkles/firmness, also ranked near the top. Given that AMORE CHAT has a high concentration of female users in their 40s, this distribution aligns well with our user base demographics.

Concerns about sensitivity and irritation, breakouts and acne, oiliness, and pores also appeared. Rather than searching for products around a single skin concern, customers were asking questions premised on combined conditions, such as being dry and sensitive at once, or worrying about firmness while also being concerned about dark spots. This means we need to understand the combinations of concerns customers actually have and be able to suggest products and routines suited to those combinations.

Looking at Customers’ Main Questions: ③ Inquiries by Product Category

Looking at inquiries by product category, creams/lotions and sunscreen accounted for the largest share, followed by essences/serums, toners, cushions/base makeup, and cleansers. This shows that customer questions are concentrated on the product categories closely tied to the basic skincare routine.

Creams/lotions, sunscreen, and essences/serums, in particular, are products where the right choice depends on skin concerns and how they're used. Customers were not simply checking product categories; they were asking which product would suit them best based on their own skin condition, the season, how it felt on their skin, and their existing routine. Cushion/base makeup inquiries also accounted for a meaningful share, confirming that comparison questions, based on shade, feel, and comparison with existing products, continue not only in skincare but in makeup as well.

Looking at Customers’ Main Questions: ④ Ingredient Inquiries

Looking at the top ingredient keywords, retinol accounted for the largest share, followed by vitamin C, centella/cica, collagen, ceramide, and hyaluronic acid. This shows that customers are not exploring products by brand or category alone; they are checking efficacy and suitability based on specific ingredients.

Notably, retinol and vitamin C are ingredients tied to key skin concerns such as wrinkles/firmness, melasma/dark spots/pigmentation, and dullness/tone. At the same time, they are ingredients that raise questions customers find hard to judge on their own, such as initial irritation, order of application, and whether they can be used together with other ingredients. Mentions of centella/cica, ceramide, and hyaluronic acid show that concerns such as sensitivity/irritation, barrier care, and hydration are also carrying over into ingredient questions.

Looking at Customers’ Main Questions: ⑤ Brand Inquiries

Looking at question types by brand, product recommendation and skin-concern questions accounted for a high share across most brands. This shows that even within a given brand, customers are asking which product to choose based on their own skin condition or purchase purpose.

That said, there were differences in which question types stood out by brand. HERA and espoir showed relatively high shares of shade/color questions, reflecting a strong base-makeup selection context, while IOPE stood out for ingredient- and routine-related questions. AESTURA showed skin-concern, ingredient, and routine questions together, reflecting a strong sensitive-skin and barrier-care context. innisfree, HANYUL, and LANEIGE showed relatively high shares of recommendation and skin-concern questions, confirming a pattern of customers looking within a brand for the product that fits their particular concern.

How Can We Put These Results to Use?

① A Starting Point for Brand Content, Including GEO

The customer-question analysis we’ve walked through so far shows that AMORE CHAT is not simply an On-site Agent for handling customer inquiries. Real conversation data is accumulating in AMORE CHAT about customers’ skin concerns, which product categories they struggle to choose between, which ingredients and usage methods they’re curious about, and what kinds of questions tend to come up by brand. This data lets us interpret customer needs retrospectively and also provides clues about what content we should prepare going forward.

This can also become an important starting point from a GEO perspective. When preparing for GEO, we often end up imagining what questions customers might ask an external AI. But by referencing and analyzing the questions left with our On-site Agent, we can prepare based on real data instead of imagination. Questions customers ask within our own channels are likely to be posed in a similar form to those posed to external AIs such as ChatGPT or Gemini. I believe, then, that AMORE CHAT’s question data can become the raw material for GEO content. By identifying in advance what content is needed for which brands and products, and what comparisons, recommendations, usage instructions, and ingredient explanations are required, we can help external AI understand and recommend our brands and products more accurately.

② Strengthening Personalized CRM

Another possibility for expansion is hyper-personalized CRM(Customer Relationship Management). Where CRM up to now has largely started from 'what did they do' (purchase history, cart activity, click behavior), AMORE CHAT lets us see ‘what are they wondering about and asking right now.’ From a customer’s questions alone, we can read signals such as whether they’re comparing products, worried about skin suitability, struggling to choose a shade number or color, or postponing a decision because of a coupon or delivery condition. This means we can go beyond age, gender, and purchase category to classify and understand customers in much finer detail, based on their current intent and stage of concern.

From this perspective, AMORE CHAT conversations can change the starting point of CRM messaging itself. Even among customers interested in the same product, one may be worried about ingredients and irritation, while another may be checking price and benefits — and yet another may want to know how it differs from a product they already use. I believe, then, that AMORE CHAT-based CRM can be upgraded from a uniform approach, sending the same promotional message to every customer, to a hyper-personalized approach that connects customers to more relevant information, content, benefits, and purchase support, based on the questions and context they’ve actually left behind.

#OUTRO

In the age of agentic commerce, customer touchpoints can come together into a single cycle. Customers discover our brand through external AI, then, within our channels, talk with AMORE CHAT and leave behind their skin concerns and purchase hesitations. Those conversations, in turn, become raw material for improving brand content, product detail pages, and customer consultation, and well-organized content can become the foundation for our brand being discovered more accurately by external AI. And when a customer’s need cannot be fully resolved within our own store, new possibilities for expansion open up, ones that connect to external products, services, and partner experiences.

Today, I’ve shared the analysis of AMORE CHAT, the touchpoint at the center of this cycle where we hear customer questions directly. This analysis is less a rigorous research study of customer questions than a look at the main trends in the conversations that have accumulated in AMORE CHAT so far. Even so, I hope it has been an interesting and useful reference, even in a small way, for understanding how customers actually phrase their questions, what skin concerns they have, what product characteristics they ask about, and which kinds of inquiries recur by brand.