Arena Urges Retailers to Turn Into APIs as AI Agents Bypass Web Funnels

Your Next Website Won't Be Built for Humans: Why the Future Belongs to Conversational Commerce — Ai4

Ai4

Brands must replace static page architecture with dynamic AI endpoints and in-chat checkout because autonomous agents and LLM discovery render traditional search-and-filter funnels obsolete.

AI-assisted summary of Your Next Website Won't Be Built for Humans: Why the Future Belongs to Conversational Commerce — original source. Check the original for full context.

Source published: 2025-08-24

Detailed notes

## Background Published on YouTube by Ai4 on August 24, 2025, this presentation features Jonathan Arena, Co-Founder of New Generation, discussing how conversational commerce is rendering traditional e-commerce websites obsolete. According to the session description, the talk explores how brands must adapt to AI agents and conversational discovery by creating intelligent storefronts designed for both human shoppers and machine agents.

## TL;DR As AI interfaces like ChatGPT rapidly replace traditional search engines as the primary discovery layer, conventional linear e-commerce funnels and static product detail pages are experiencing massive bounce rates. Jonathan Arena illustrates this systemic failure with benchmarks from Samsung, demonstrating how existing sites fail basic natural language queries and trap autonomous browsing agents in UI bottlenecks. To address this friction, New Generation advocates that brands establish dedicated 'AI addresses' and re-architect their platforms as APIs capable of serving structured catalog data directly to LLMs. By adopting bidirectional context sharing and embedded in-chat checkout, merchants can capture high-intent conversational traffic while preserving direct customer relationships.

## Atmosphere & style The presentation carries the brisk, urgent authority of an industry-insider keynote, balancing stark economic critiques with visionary architectural solutions. Arena delivers a compelling wake-up call, using dry humor and empirical absurdities—such as an enterprise search engine returning washing machines when asked for camera phones—to expose how brittle traditional web design has become. The tone feels pragmatic yet transformative, engaging the audience with live benchmark comparisons before laying out a clear, agent-native blueprint for the future of online retail.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) E-commerce is undergoing its largest shift in 30 years as LLMs and AI agents replace traditional keyword search and linear website navigation as the primary discovery layer. - [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) Year-over-year referral traffic from AI platforms to commercial websites has grown by 1,900%, with over 50% of consumers planning to use AI tools for shopping. - [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) Traditional search funnels are broken because shoppers enter mid-funnel through 'side door' SEO pages, creating severe friction and high bounce rates when exploring adjacent products. - [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) Marketing spend is severely misallocated: brands invest 65% of budgets on top-of-funnel acquisition and only 4% on mid-funnel consideration, leaving average conversion rates stuck at roughly 2%. - [04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) Brands require dedicated 'AI addresses' (e.g., ai.brand.com) engineered to serve generative, dynamic UI and contextual product detail pages tailored to natural language intent. - [04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) Legacy retail search systems fail simple natural language queries, such as Samsung's keyword engine returning washing machines when searched for high-resolution camera phones. - [06:50](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=410s) Autonomous AI shopping agents face severe bottlenecks on legacy human-centric websites, taking over 5 minutes versus just 28 seconds to complete tasks on structured AI storefronts. - [06:50](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=410s) Brands must operate as APIs rather than static pages, connecting real-time catalog data, pricing, and inventory directly to AI chat clients like ChatGPT and Perplexity. - [06:50](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=410s) Embedded in-chat checkout partnerships, such as with Visa, allow customers to complete purchases directly within conversational interfaces without visiting external storefronts. - [09:05](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=545s) Bidirectional context sharing via standardized context objects ensures external chat intent seamlessly transfers to merchant websites without resetting the customer's search. - [09:05](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=545s) To prevent disintermediation by third-party chatbots, brands must reallocate spend into proprietary on-site brand intelligence to capture rich first-party intent data.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) The Discovery Layer Shift and the Broken Linear Funnel

Jonathan Arena opens the session by addressing how artificial intelligence is fundamentally restructuring the modern web. Every channel through which consumers discover products and drive traffic to merchant sites is increasingly channeled through Large Language Model (LLM) interfaces. Arena characterizes this transition as "probably the biggest shift we've seen in commerce in the past 30 years," warning merchants that their next visitor is overwhelmingly likely to be either an autonomous AI agent or an AI-referred shopper originating from platforms such as ChatGPT or Perplexity.

```mermaid flowchart TD accTitle: Traditional vs AI-Native Commerce Funnel subgraph Traditional["Traditional Linear Funnel"] T1["Top of Funnel: Keyword Search / Ads (65% Spend)"] --> T2["Side Door Entry: Static SEO / PDP"] T2 --> T3["Friction: Rigid Tree & UI (High Bounce)"] T3 --> T4["Conversion: ~2% Checkout"] end

subgraph AINative["AI-Native Commerce Architecture"] A1["Discovery: LLMs & AI Agents (ChatGPT / Perplexity)"] --> A2["AI Address / API (ai.brand.com)"] A2 --> A3["Dynamic Generative UI & Contextual PDP"] A2 --> A4["Direct In-Chat Checkout via Visa API"] end ```

As LLMs take over as the discovery layer of the internet, historic traffic mechanics are breaking down. Rather than progressing methodically through a traditional multi-step journey—landing on a homepage, browsing hierarchical menus, filtering catalogs, and adding items to a cart—consumers are entering through the "side door," landing mid-funnel directly on specific product detail pages (PDPs) or search-engine-optimized landing pages.

- Over 50% of consumers surveyed by Adobe report plans to use AI tools for shopping. - Year-over-year referral traffic from AI platforms to commercial websites has grown by 1,900%. - Traditional linear search-to-navigation pathways are breaking down as discovery moves directly into natural language chat interfaces.

---

[01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) Friction in the Midfunnel and Misallocated Marketing Spend

Brands historically constructed elaborate "nets" comprising hundreds of programmatic SEO landing pages to capture niche search queries. However, this architectural design introduces severe user friction: when a shopper arrives on an SEO page but actually requires an adjacent product category or comparative evaluation, navigating out of that rigid website tree creates massive bounce rates. Arena notes that the midfunnel—the post-discovery space encompassing consideration and cart addition—is where the vast majority of consumer interaction occurs, yet it suffers from catastrophic conversion leakage.

The economic allocation within e-commerce marketing remains severely misaligned with this reality. Brands routinely invest heavy capital into top-of-funnel acquisition while neglecting the digital shelf where consideration happens, treating mid-funnel conversion rates as an immovable plateau.

- Average e-commerce conversion rates across the web remain at an "abysmal" baseline: out of every 100 visitors who reach a website, only about two successfully complete a transaction. - 65% of overall brand marketing budgets are dedicated to top-of-funnel discovery, compared to an "anemic 4%" allocated toward consideration and midfunnel UX. - Brands have historically treated consideration and add-to-cart conversion rates as a "local maxima," failing to extract meaningful efficiency gains from static page layouts.

---

[04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) The AI Address and the Samsung Case Study

To solve the friction of static sites, New Generation is building intelligent commerce infrastructure centered around three components: dedicated AI storefronts, consumer agents, and native LLM chat integrations. A central piece of this strategy is providing brands with their own dedicated "AI address" (such as `ai.samsung.com` or `ai.nike.com`) that acts as an intelligent endpoint engineered explicitly to receive conversational, AI-driven traffic.

Arena highlights the urgency of this infrastructure through an empirical case study of Samsung—the second most visited retail website globally, commanding over 550 million monthly visits. When testing standard natural language queries on Samsung's production site, the existing keyword search infrastructure collapsed entirely:

> "What happens if you search for, you know, a very typical question, an easy natural language search? 'Phones that have the highest res cameras.' What are they going to show me? This is real. Washing machines. They show me washing machines."

In contrast, New Generation's AI storefront connects directly over the brand's product catalog to dynamically generate a branded interface in real time. It serves generative imagery, contextual descriptions, and intelligent follow-up questions to refine customer intent. When a shopper selects a device, the system renders an AI-generated product details page tailored precisely to highlight camera resolutions, sensor specifications, and features relevant to the initial prompt.

---

[06:50](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=410s) Benchmarking Agentic Shoppers and Brands as an API

The transition to conversational commerce must also accommodate autonomous AI agents browsing on behalf of consumers. To test existing readiness, New Generation ran a head-to-head benchmark using OpenAI's operator agent across both the standard `samsung.com` domain and the AI-native `ai.samsung.com` storefront using identical product queries:

- On `samsung.com`, the agent struggled for over five and a half minutes with human-oriented UI bottlenecks, getting trapped in React components, broken keyword search flows, and encountering 404 and 503 errors. - On `ai.samsung.com`, the agent parsed the structured data environment and completed the discovery task in 28 seconds flat.

```mermaid flowchart LR accTitle: Brand as an API Architecture subgraph LLMClients["AI Chat Clients"] C1["ChatGPT"] C2["Perplexity"] C3["Autonomous Agents"] end

subgraph APIBridge["New Generation Infrastructure"] B1["Programmatic Catalog API"] B2["Real-Time Price & Inventory Engine"] B3["Context Object Handler"] end

subgraph MerchantBackend["Merchant & Payment Core"] M1["Brand Product Catalog"] M2["Visa Instant Checkout"] end

LLMClients <--> APIBridge APIBridge <--> MerchantBackend ```

Addressing the widespread industry concern—"What if nobody goes to a website anymore? What if everybody stays in ChatGPT?"—Arena argues that brands must fundamentally re-architect their presence: "brands need to be an API, not just a collection of pages." By operating like financial infrastructure (drawing the comparison: "Just like Plaid connects banks to apps, we connect brands to agents"), merchants can surface live inventory and dynamic pricing directly within external chat interfaces. Through an active partnership with Visa, New Generation is introducing embedded in-chat checkout, allowing users to purchase items directly within conversational threads without needing to navigate to an external storefront.

---

[09:05](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=545s) Bidirectional Context and Winning the AI Commerce Funnel

The future of conversational commerce relies on bidirectional data exchange between LLMs and merchant platforms. Beyond serving data to chat platforms, merchants need the ability to receive contextual state when users follow links out of chat interfaces. By attaching a standardized "context object" to outbound referral links, an e-commerce site can immediately inherit the user's conversational search history and intent, dynamically rendering an interface that preserves their ongoing journey without resetting their query.

Arena emphasizes that winning brands will not allow themselves to be disintermediated or surrender direct consumer relationships to third-party chatbots. Instead, merchants must reallocate marketing capital away from basic top-of-funnel discovery and reinvest into on-site "brand intelligence." Engaging shoppers in conversational mid-funnel dialogues yields granular first-party intent data while delivering a friction-free experience optimized for both human shoppers and autonomous machine agents.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) probably the biggest shift we've seen in commerce in the past 30 years — Jonathan Arena

> The e-commerce website as you know it is obsolete. Keyword search and clicks are out; conversational commerce is here. — Jonathan Arena

> [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) abysmal — Jonathan Arena

> [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) anemic 4% — Jonathan Arena

> [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) local maxima — Jonathan Arena

> [04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) What happens if you search for, you know, a very typical question, an easy natural language search? 'Phones that have the highest res cameras.' What are they going to show me? This is real. Washing machines. They show me washing machines. — Jonathan Arena

> [06:50](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=410s) What if nobody goes to a website anymore? What if everybody stays in ChatGPT? — Jonathan Arena

> [06:50](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=410s) brands need to be an API, not just a collection of pages. — Jonathan Arena

> [06:50](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=410s) Just like Plaid connects banks to apps, we connect brands to agents — Jonathan Arena

## Mentioned Jonathan Arena, New Generation, ChatGPT, Perplexity, Adobe, Samsung, Nike, OpenAI, Plaid, Visa

## Video description Jonathan Arena, Co-Founder, New Generation

Abstracts

The e-commerce website as you know it is obsolete. Keyword search and clicks are out; conversational commerce is here.

AI agents like ChatGPT are reshaping how people discover, browse, and buy online, rapidly transforming traditional e-commerce experiences. This session explores how brands must adapt by creating conversational storefronts designed for both human and AI-driven interactions.

Learn how to future-proof your brand by dynamically rendering beautiful, instant answers, converting AI-native traffic, and building trust across human and machine audiences. Understand why your brand needs its own "AI address," discover how to integrate intelligence into your product catalog, and gain insight into seamlessly transacting with agentic shoppers.

Leading brands are already capturing more high-intent traffic through AI-native channels. If your current e-commerce strategy still relies heavily on clicks, filters, and static pages, it's time to rethink your approach. Join this session to learn exactly how to adapt and thrive in the conversational commerce era.

[Source](https://www.youtube.com/watch?v=4JbRAYp2QAo)

中文简报

AI Agent 绕过网页漏斗，Arena 敦促零售商全面 API 化随着自主 Agent 与大语言模型检索机制令传统“搜索与筛选”漏斗过时，品牌必须以动态 AI 端点和聊天内结账取代静态页面架构。随着 ChatGPT 等 AI 界面迅速取代传统搜索引擎成为主要发现层，传统的线性电商漏斗和静态商品详情页正面临极高的跳出率。Jonathan Arena 结合三星的基准测试阐明了这一系统性失效，展示了现有网站如何在基础自然语言查询面前频频受阻，并将自主浏览 Agent 困在 UI 瓶颈之中。为解决这一阻碍，New Generation 主张品牌设立专门的“AI 地址”，并将自身平台重构为能直接向大语言模型输出结构化目录数据的 API。通过采用双向上下文共享与嵌入式聊天结账，商家不仅能捕获高意向的对话流量，还能维系直接的客户关系。电子商务正迎来30年来最大变革，来自AI平台的推荐引流流量年增1900%，超50%消费者计划使用AI购物，大语言模型正取代传统搜索成为首要发现层。 (0:04) 当前品牌营销预算分配严重失衡，高达65%的预算投入漏斗顶端获客，而中端考量环节仅占4%，导致全网平均电商转化率长期停滞在约2%的低位。 (1:42) 传统关键词检索在自然语言下频频失效（如在三星官网搜索高像素相机手机却展示洗衣机），品牌亟需建立专属“AI地址”以动态生成契合意图的UI与产品详情。 (4:00) 自主AI购物代理在传统网站上完成任务需超5分钟，而在结构化AI店面上仅需28秒；品牌必须转型为API架构，结合Visa等方案实现对话内直接完成结账。 (6:50)

Reading summary

Jonathan Arena argues that e-commerce is undergoing its largest shift in 30 years as Large Language Models replace traditional search engines as the primary discovery layer, forcing brands to re-architect their websites for autonomous agents and conversational interfaces. The Breakdown of the Linear Commerce Funnel (0:04) Traditional e-commerce relied on a linear funnel where consumers navigated from homepages through hierarchical menus to product pages. Arena asserts this model is collapsing because consumers increasingly discover products inside LLM environments like ChatGPT and Perplexity, entering merchant websites mid-funnel through 'side door' programmatic SEO links or direct product detail pages (PDPs). This shift is supported by shifting consumer behavior and traffic trends. Over 50% of consumers surveyed by Adobe report intentions to use AI tools for shopping, while year-over-year referral traffic from AI platforms to commercial sites has grown by 1,900%. When natural language discovery bypasses standard search pathways, rigid website navigation creates friction and drives high bounce rates. Capital Misallocation Across the Midfunnel (1:42) E-commerce marketing economics remain misaligned with user behavior. Brands dedicate roughly 65% of marketing budgets to top-of-funnel customer acquisition, leaving an 'anemic 4%' for consideration and mid-funnel user experience. Consequently, conversion remains stuck at an average of roughly two completed transactions for every 100 website visitors. Arena observes that brands historically accepted low consideration and add-to-cart rates as an unfixable 'local maxima.' Static landing pages and programmatic SEO networks fail when shoppers require comparative evaluations or adjacent categories, leaking potential conversions at the exact stage where the majority of customer decision-making occurs. Failure of Keyword Search in Natural Language (4:00) Existing retail architectures struggle to process conversational queries. Arena points to Samsung—the second most visited retail site globally with over 550 million monthly visits—where entering a natural language query for 'phones that have the highest res cameras' failed to surface mobile devices and instead returned washing machines. To bridge this gap, New Generation advocates deploying dedicated AI storefronts at endpoints such as ai.samsung.com. Connected directly to product catalogs, these systems dynamically generate branded interfaces, contextual product descriptions, and tailored product detail pages in response to user intent rather than relying on brittle keyword indexing. Brands as APIs for Autonomous Agents (6:50) Beyond human shoppers, retail websites must accommodate autonomous AI agents acting on behalf of consumers. In a head-to-head test using OpenAI's operator agent on identical tasks, the agent struggled on samsung.com for over five and a half minutes due to React bottlenecks, search errors, and 404/503 responses, whereas it completed the task on ai.samsung.com in 28 seconds. To counter the risk of third-party platforms entirely capturing consumer traffic, Arena argues that merchants must function like APIs rather than static collections of pages, comparing the model to how Plaid connects banks to applications. By exposing dynamic inventory, structured data, and integrating partnerships like Visa's in-chat checkout, brands can complete sales directly inside conversational threads. Preserving Intent Through Bidirectional Context (9:05) Maintaining direct consumer relationships requires bidirectional data exchange between external LLMs and merchant platforms. Arena notes that when users transition from a chat client to a website, standardized 'context objects' attached to referral links can pass conversation history and intent directly to the destination site. By capturing this context, merchant platforms can render dynamic, personalized interfaces that sustain the ongoing journey without forcing users to restart their search. Arena concludes that brands must reallocate budget from top-of-funnel acquisition toward on-site conversational intelligence to capture high-value first-party intent data. Arena maintains that brands must transition from static, human-only page trees to API-driven conversational architectures to avoid disintermediation by third-party AI platforms. However, the operational complexity and merchant adoption hurdles of re-engineering legacy catalog and checkout systems into dynamic agent endpoints remain central industry challenges.

精炼阅读摘要

Jonathan Arena指出，人工智能正在重构现代互联网的流量发现机制，消费者的购物路径正从静态层级导航转向大语言模型驱动的对话界面与自主代理。商家必须从构建面向人类的传统页面转向构建“品牌即API”的对话式商务架构，以解决转化漏斗中段严重的流失问题。发现层迁移与传统线性漏斗的瓦解 (0:04) Jonathan Arena在演讲中强调，大语言模型（LLM）正日益成为互联网的基础发现层，这是过去30年电子商务领域最具深远影响的变革。越来越多的消费者不再沿着“访问首页—浏览层级分类—使用筛选器—加入购物车”的传统线性路径购物，而是通过ChatGPT、Perplexity等平台以自然语言发起探索，从而直接进入具体商品详情页（PDP）或搜索引擎优化（SEO）页面，这种“侧门接入”彻底打破了原有的站内流转机制。相关数据验证了这一流量迁移的规模：Adobe的一项调查显示，超过50%的受访消费者计划使用AI工具进行购物；同时，来自AI平台的电商引流流量年同比增长达到了1,900%。Arena认为，未来访问商户网站的“访客”极有可能是自主运行的AI代理或由AI平台推荐的消费者，现有的关键词搜索与静态页面展示已无法承接此类交互需求。营销预算错配与漏斗中段的转化瓶颈 (1:42) 长久以来，品牌方投入巨资搭建包含海量程序化SEO页面的“引流网”以捕捉长尾关键词，但这种架构存在严重的结构性摩擦。当消费者通过特定SEO页面进入网站，若需要查看相邻品类或进行横向比较时，死板的层级目录会导致极高的跳出率。Arena指出，涵盖考量与加购环节的漏斗中段（Midfunnel）承载了绝大多数用户互动，却面临着严重的用户流失。这一困境暴露出电商营销资本分配的严重错位：在整体营销预算中，高达65%的资金被倾注在漏斗顶端的流量获取上，而在决定最终转化的中段考量与用户体验上，预算分配仅占4%。由于未能突破静态页面带来的局限，全网平均电商转化率长期徘徊在约2%的极低基准线上（每100名访问者中仅约2人完成支付），许多品牌错误地将这一低转化率视为不可突破的“局部极值”，忽视了中段体验优化的潜力。 AI专属地址与动态生成界面的实践 (4:00) 针对静态网站无法响应复杂自然语言的问题，New Generation提出了包含专属AI店面、消费者代理和原生LLM集成的基础设施方案。该策略的核心是为品牌建立独立的“AI地址”（例如ai.samsung.com），作为承接对话式流量与AI驱动查询的智能交互端点。 Arena以全球访问量第二大的零售网站三星（月访问量超5.5亿次）为例说明了现有系统的缺陷。在三星官网输入常见自然语言查询“具有最高分辨率摄像头的手机”时，传统关键词搜索机制完全崩溃，向用户展示了洗衣机商品。相比之下，基于新一代架构的AI店面直接连接产品目录，实时动态生成匹配上下文的品牌界面，不仅呈现定制化的生成式图像与针对性描述，还能主动发起跟进提问以澄清购买意图，并针对用户关心的相机传感器和分辨率参数生成专属的商品详情页。品牌即API：代理基准测试与对话内结算 (6:50) 随着代客购物的自主AI代理兴起，电商基础设施必须适配机器代理的访问效率。New Generation利用OpenAI的operator代理在三星官网（samsung.com）与AI原生店面（ai.samsung.com）上执行了相同的购物任务基准测试：在传统网站上，代理受阻于复杂的React组件、断裂的搜索路径以及404与503网络错误，耗时超过5分30秒；而在结构化数据驱动的AI店面上，代理仅用时28秒即完成了商品检索。为应对“用户若停留在ChatGPT中不再访问官网”的行业担忧，Arena提出“品牌需要成为API，而非仅仅是网页集合”。正如Plaid连接银行与应用程序一样，电商系统应当作为底层基础设施向AI代理输出实时库存和动态定价。此外，New Generation通过与Visa合作，正推进在对话流中直接嵌入支付结账功能，允许消费者无需跳转即可在第三方聊天界面内一键完成交易。双向上下文传递与第一方意图数据捕获 (9:05) 对话式商务的完全落地依赖于LLM平台与商家系统之间的双向数据交换。除了单向向外部聊天平台输出商品信息，商户还必须具备承接上下文状态的能力。通过在引流跳转链接中嵌入标准化的“上下文对象”（Context Object），当用户从外部对话点击进入品牌站点时，网站能即时继承先前的对话历史与检索意图，动态渲染出延续性界面，避免重置用户的搜索进度。 Arena强调，领先品牌不应将消费者关系拱手让给第三方聊天机器人平台，而应将原本浪费在漏斗顶端的营销资本重定向到自有的“品牌智能”建设中。在漏斗中段与消费者开展多轮对话交互，不仅能大幅消除人机界面的摩擦，还能帮助品牌沉淀高价值的第一方购买意图数据，同时服务好人类顾客与自主机器代理。 Arena总结指出，未来的电商竞争取决于品牌能否将自身重构为面向AI代理友好的数据接口与自适应前端；但该愿景能否广泛普及，仍取决于外部主流LLM平台对双向上下文协议的开放度，以及去中心化对话支付生态的成熟速度。

Detailed notes — zh

## Background 该视频由行业频道 Ai4 于 2025 年 8 月 24 日发布在 YouTube 上，主讲人为 New Generation 联合创始人乔纳森·阿雷纳（Jonathan Arena）。视频简介指出，传统的关键词搜索与层级点击式电商网站正在过时，对话式商务全面来临，品牌必须构建兼顾人类与 AI 智能体交互的智能门面，以捕获高意向的 AI 原生流量并实现未来转型。 ## TL;DR 乔纳森指出大语言模型正成为互联网的核心发现层，导致传统线性电商漏斗瓦解，品牌面临严重的中漏斗转化危机与预算错配。通过三星官网将手机查询误导至洗衣机的荒谬案例，他展示了传统 UI 在自然语言意图面前的脆弱，并提出品牌需要专属的“AI 地址”（如 ai.samsung.com）。实测证明，AI 门面能让 OpenAI Operator 智能体的查询耗时从 5 分半钟锐减至 28 秒。未来品牌必须将自身重构为面向智能体的 API，通过双向上下文传递与对话内原生结算抢占新商业制高点。 ## Atmosphere & style 演讲风格敏锐犀利且节奏紧凑，主讲人以幽默风趣的现场互动开场，随后迅速切入对传统电商痛点的降维打击。当现场展示三星官网搜出洗衣机的尴尬案例时极具戏剧张力与讽刺意味，紧接着切入 OpenAI Operator 的硬核对抗实测，气氛瞬间转为严谨前瞻。整场分享不仅给传统电商从业者带来了强烈的危机警示，更以清晰的架构蓝图唤起了拥抱下一代智能商务的振奋与紧迫感。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) 大语言模型（LLM）正在成为互联网的核心发现层，彻底瓦解过去 30 年赖以生存的线性电商漏斗。 - [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) 超过 50% 的消费者计划借助 AI 辅助购物，来自 AI 渠道的推荐流量年同比增幅高达 1,900%。 - [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) 用户经由 AI 工具直接“走侧门”切入中漏斗（考量与加购环节），绕开了品牌精心设计的首页导航。 - [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) 电商预算分配极度扭曲：品牌将 65% 的预算砸向顶层获客，而决定 2% 平均转化率的中漏斗考量阶段仅获得 4% 预算。 - [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) 为穷举关键词搭建的庞大 SEO 落地页树存在严重导航摩擦，不仅用户体验脆弱，也是高跳出率的核心根源。 - [04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) 以月访问量超 5.5 亿次的三星官网为例，输入自然语言搜“高分辨率相机手机”竟返回洗衣机，印证了传统电商搜索系统的滞后。 - [04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) 品牌需要建立互联网上的“AI 地址”（如 ai.brand.com），作为承接 AI 原生流量与智能体交互的专用终端节点。 - [05:54](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=354s) AI 门面能根据用户自然语言实时动态渲染契合品牌规范的 UI 与参数详情页，并主动提出后续问题协助细化需求。 - [05:54](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=354s) OpenAI 智能体实测对比显示：专用 AI 地址凭借结构化数据仅耗时 28 秒完成检索，而传统人类 UI 耗时超 5 分钟且频遭报错阻断。 - [07:58](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=478s) 应对“用户留存在 LLM 不再访问官网”的解法是让品牌变成 API，如同 Plaid 连接银行一样将品牌连接到各大智能体。 - [07:58](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=478s) 通过与 Visa 合作推进 API 原生结算，用户无需跳出 ChatGPT 等大模型对话窗口即可直接调用实时库存完成结账闭环。 - [07:58](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=478s) 双向上下文传递允许用户跳转到品牌站时携带多轮对话历史，品牌能无缝延续用户购物心智，防范被第三方平台彻底去中介化。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) AI 正在重塑互联网：传统电商漏斗的瓦解演讲者乔纳森（Jonathan Arena）是 New Generation 的联合创始人。在会议接近尾声的轻松互动中，他直接切入核心论点：人工智能正在对整个互联网进行底层的“重新布线”（rewiring the internet）。今天人们发现产品、流量触达网站的方式，正在大规模流向大语言模型（LLM）界面。在乔纳森看来，这是过去 30 年来电子商务领域发生的最重大变革。未来访问企业网站的很可能不再是传统的人类浏览者，而是一个 AI Agent，或者是由 ChatGPT、Perplexity 等 AI 对话工具推荐引导而来的“AI 推荐型顾客”。大语言模型正在成为互联网的核心发现层（discovery layer），这一转变彻底改变了流量格局。传统电商赖以生存的线性客户漏斗——“发现网站 → 导航浏览 → 搜索筛选商品 → 结账购买”——正在走向崩溃： - **消费者行为转向**：根据 Adobe 的数据，超过半数（>50%）的消费者明确表示计划使用 AI 辅助购物。 - **推荐流量激增**：来自 AI 渠道的网站推荐流量呈现爆发式增长，年同比（YoY）增幅高达 1,900%。 - **入口路径重构**：消费者不再从品牌首页按部就班地进入，而是经由 AI 工具直接“走侧门”（side door）切入网站中段。 ```mermaid flowchart TD accTitle: 传统电商路径与 AI 驱动新路径对比 subgraph Traditional["传统线性电商漏斗（已失效）"] T1["首页访问 / 搜索发现"] --> T2["层级导航与关键词筛选"] T2 --> T3["陷入庞大 SEO 页面网"] T3 --> T4["流失率极高的商品详情页 PDP"] T4 --> T5["仅 2% 结账转化"] end subgraph AIEra["AI 时代智能商务架构"] A1["LLM 发现层 (ChatGPT / Perplexity)"] --> A2{"交互模式"} A2 -->|"在对话内直接结算"| A3["品牌 API / Visa 支付底层"] A2 -->|"带上下文跳转"| A4["品牌专属 AI 地址 (ai.brand.com)"] A4 --> A5["实时动态生成专属 UI / 规格展示"] end ``` --- ### [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) 侧门进入与中漏斗困境：极其悬殊的预算错配与转化危机用户绕开品牌精心设计的首页，直接“降落”在商品详情页（PDP）或特定的 SEO 落地页，直接进入了“中漏斗”（midfunnel）阶段。乔纳森解释，发现（Discovery）仅仅是访客找到网站并点击进入，而中漏斗的“考量（Consideration）与加入购物车（Add to Cart）”才是用户与品牌互动最深、最具商业价值的核心地带。为了捕获搜索引擎流量，各大品牌过去构建了庞大的防御网络，搭建了成百上千个杂乱的 SEO 落地页面。然而，这种依靠穷举关键词构建的页面树（website tree）极易引发灾难性的用户体验：一旦用户从某个细分搜索点进入，却发现自己需要横向对比另一品类时，网站导航就会产生极其严重的摩擦阻力，这正是全网电商高跳出率（bounce rate）的根源。 - **惨淡的行业平均转化率**：在全网电商平均水平中，每 100 名访问网站的访客中，最终仅有 2 人能够顺利完成结账购买（2% 转化率），绝大多数流失均发生在“考量与加入购物车”之间。 - **扭曲的营销预算分配**：品牌在不同环节的资本投入极度不均衡——高达 65% 的营销预算砸在漏斗顶端的“发现获客”上，而投入到决定转化生死的中漏斗“考量”阶段的预算仅有区区 4%。 - **局部最优陷阱**：品牌并不是不知道中间流失严重，而是受限于传统网页的技术范式，考量与加购环节长期处于优化停滞的“局部最优”（local maxima），无法再榨出更多体验价值。 --- ### [04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) 品牌的“AI 地址”与三星官网的“洗衣机”笑话面对这一困局，New Generation 提出了构想：如果将整个网站的内容与信息全部重组，注入一个智能化的 AI 网站，会是怎样的形态？乔纳森指出，智能商业基础设施有三大关键聚焦板块：AI 门面（AI Storefronts）、消费者智能体（Consumer Agents）以及无处不在的 LLM 对话界面。为此，他们正在与全球顶级品牌合作，建立互联网上的第一个“AI 地址”（AI Address），例如 `ai.samsung.com` 或 `ai.nike.com`。这个 AI 地址是品牌专门用来接收 AI 流量的专用终端节点，也是为新一代客户暴露 AI 原生体验的阵地。为了证明传统电商在智能化时代的迟钝，乔纳森在现场展示了一个真实的案例研究。作为全球访问量第二大的零售电商网站（月访问量超过 5.5 亿次），三星电子（Samsung.com）按理说应当拥有顶级的数字化体验。然而，当乔纳森在三星官网的搜索框中输入一个非常自然的日常查询——“拥有最高分辨率摄像头的手机”（Phones that have the highest res cameras）时，搜索结果却令人啼笑皆非： > “这是真实发生的事情。洗衣机。他们居然给我推荐了洗衣机。” 乔纳森幽默地吐槽道。这一荒谬的结果并非三星独有，几乎所有现有电商网站面对自然语言与智能体时代的搜索意图时，都显得准备不足、捉襟见肘。 --- ### [05:54](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=354s) AI 门面 vs. 传统 UI：OpenAI Operator 的实测对决 New Generation 给出的解法是“AI 门面”（AI Storefront）。它直接接入品牌的产品目录，允许消费者使用自然语言直接向品牌提问。它能实时渲染出完全契合三星品牌视觉规范（on-brand）的精美界面，实时动态生成匹配的商品图、标题及描述，甚至能像专业导购一样主动提出智能后续问题，协助用户细化需求。当用户点击某一商品时，呈现的并非千篇一律的传统商品页，而是根据用户提问上下文动态生成的 AI 详情页，精准高亮突出其最关心的规格与参数。除了服务人类消费者，更关键的考验在于：当访问网站的是 AI Agent 时，商家真的准备好了吗？乔纳森团队进行了一次真实对抗评测，调用 OpenAI 当时名为 Operator（现为 ChatGPT Agent）的智能体，使用相同的自然语言查询，同时访问传统的 `samsung.com` 与他们搭建的 `ai.samsung.com`： | 评估维度 | 传统三星官网 (`samsung.com`) | 专用 AI 门面 (`ai.samsung.com`) | |:--- |:--- |:--- | | **查询耗时** | 超过 5 分 30 秒，过程异常挣扎 | **仅需 28 秒** 快速完成检索 | | **界面机制** | 繁琐的人类 UI 组件、脆弱的关键词搜索 | 纯粹的结构化数据与程序化接口 | | **执行痛点** | 频繁进出详情页，屡次遭遇 404/503 报错中断 | 零上下文丢失，端到端流畅交互 | 乔纳森深入剖析了两者巨大差异的技术原因：AI Agent 需要的是结构化数据，而传统专为人类眼睛设计的 UI（复杂的 React 组件、陈旧的关键字搜素系统）对机器而言几乎是信息迷宫，极易导致 Agent 陷入死循环甚至请求崩溃。 --- ### [07:58](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=478s) 品牌即 API：无缝对话内结账与双向上下文未来针对行业内流传甚广的顾虑——“如果所有人未来都留在 ChatGPT 里，不再访问任何品牌网站怎么办？”，乔纳森给出了极为笃定的答案：“品牌需要变成一个 API，而不仅仅是一堆静态网页的集合。” > “就像 Plaid 将银行连接到应用程序一样，我们将品牌连接到智能体。” 这一架构具有革命性的落地场景：无论用户使用哪一款大语言模型，都能够直接在对话界面内浏览商品。这些数据不是通过落后的爬虫抓取或滞后的 SEO Feed 流生成的，而是直接通过程序化 API 直连品牌库存系统。借助实时库存状态、动态价格以及直连商户的接口，用户无需跳出对话窗口即可完成整个交易闭环。乔纳森透露，他们目前正在与 Visa 展开紧密合作，将“对话内原生结算”彻底变为现实。更深层的潜力在于双向的“上下文传递”（Bi-directional Context）：当消费者在 ChatGPT 等外部对话中产生购买意向并点击跳转到品牌站点时，他们会随身携带一个“上下文对象”（Context Object）。这意味着品牌网站能完全继承用户此前长达数轮的搜索与思考历史，无缝延续其购物心智。乔纳森在演讲最后强调，最终在 AI 时代胜出的品牌，必将全力构建全新的“AI 漏斗”： - **重视高意向流量**：全面拥抱那些偏好使用自然语言找寻商品、转化意愿极强的 AI 合格访客（AI-qualified visitors）。 - **重塑中漏斗品牌智能**：将过去盲目倾倒在顶层获客的预算，重新分配给中漏斗站内考量阶段的“品牌智能”（Brand Intelligence），坚决避免在客户关系中被第三方平台彻底去中介化。 - **打造双重极致体验**：既能为人类访客提供实时动态的个性化对话，又能为机器智能体提供敏捷高效的结构化服务，掌握 conversational commerce 时代的主动权。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) 人工智能正在对整个互联网进行底层的“重新布线”（rewiring the internet）。 — Jonathan Arena > [00:04](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=4s) 大语言模型正在成为互联网的核心发现层（discovery layer），这一转变彻底改变了流量格局。 — Jonathan Arena > [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) 每 100 名访问网站的访客中，最终仅有 2 人能够顺利完成结账购买（2% 转化率），绝大多数流失均发生在“考量与加入购物车”之间。 — Jonathan Arena > [01:42](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=102s) 高达 65% 的营销预算砸在漏斗顶端的“发现获客”上，而投入到决定转化生死的中漏斗“考量”阶段的预算仅有区区 4%。 — Jonathan Arena > [04:00](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=240s) 这是真实发生的事情。洗衣机。他们居然给我推荐了洗衣机。 — Jonathan Arena > [05:54](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=354s) 当访问网站的是 AI Agent 时，商家真的准备好了吗？ — Jonathan Arena > [05:54](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=354s) AI Agent 需要的是结构化数据，而传统专为人类眼睛设计的 UI（复杂的 React 组件、陈旧的关键字搜素系统）对机器而言几乎是信息迷宫，极易导致 Agent 陷入死循环甚至请求崩溃。 — Jonathan Arena > [07:58](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=478s) 品牌需要变成一个 API，而不仅仅是一堆静态网页的集合。 — Jonathan Arena > [07:58](https://www.youtube.com/watch?v=4JbRAYp2QAo&t=478s) 就像 Plaid 将银行连接到应用程序一样，我们将品牌连接到智能体。 — Jonathan Arena ## Mentioned Jonathan Arena, New Generation, ChatGPT, Perplexity, Adobe, Samsung, Nike, OpenAI Operator, Plaid, Visa, React ## Video description Jonathan Arena, Co-Founder, New Generation Abstracts The e-commerce website as you know it is obsolete. Keyword search and clicks are out; conversational commerce is here. AI agents like ChatGPT are reshaping how people discover, browse, and buy online, rapidly transforming traditional e-commerce experiences. This session explores how brands must adapt by creating conversational storefronts designed for both human and AI-driven interactions. Learn how to future-proof your brand by dynamically rendering beautiful, instant answers, converting AI-native traffic, and building trust across human and machine audiences. Understand why your brand needs its own "AI address," discover how to integrate intelligence into your product catalog, and gain insight into seamlessly transacting with agentic shoppers. Leading brands are already capturing more high-intent traffic through AI-native channels. If your current e-commerce strategy still relies heavily on clicks, filters, and static pages, it's time to rethink your approach. Join this session to learn exactly how to adapt and thrive in the conversational commerce era. [Source](https://www.youtube.com/watch?v=4JbRAYp2QAo)

Source quotes

[0:04] Jonathan Arena: 人工智能正在对整个互联网进行底层的“重新布线”（rewiring the internet）。 [0:04] Jonathan Arena: 大语言模型正在成为互联网的核心发现层（discovery layer），这一转变彻底改变了流量格局。 [1:42] Jonathan Arena: 每 100 名访问网站的访客中，最终仅有 2 人能够顺利完成结账购买（2% 转化率），绝大多数流失均发生在“考量与加入购物车”之间。 [1:42] Jonathan Arena: 高达 65% 的营销预算砸在漏斗顶端的“发现获客”上，而投入到决定转化生死的中漏斗“考量”阶段的预算仅有区区 4%。 [4:00] Jonathan Arena: 这是真实发生的事情。洗衣机。他们居然给我推荐了洗衣机。 [5:54] Jonathan Arena: 当访问网站的是 AI Agent 时，商家真的准备好了吗？ [5:54] Jonathan Arena: AI Agent 需要的是结构化数据，而传统专为人类眼睛设计的 UI（复杂的 React 组件、陈旧的关键字搜素系统）对机器而言几乎是信息迷宫，极易导致 Agent 陷入死循环甚至请求崩溃。 [7:58] Jonathan Arena: 品牌需要变成一个 API，而不仅仅是一堆静态网页的集合。 [7:58] Jonathan Arena: 就像 Plaid 将银行连接到应用程序一样，我们将品牌连接到智能体。

Source quotes

[0:04] Jonathan Arena: probably the biggest shift we've seen in commerce in the past 30 years Jonathan Arena: The e-commerce website as you know it is obsolete. Keyword search and clicks are out; conversational commerce is here. [1:42] Jonathan Arena: abysmal [1:42] Jonathan Arena: anemic 4% [1:42] Jonathan Arena: local maxima [4:00] Jonathan Arena: What happens if you search for, you know, a very typical question, an easy natural language search? 'Phones that have the highest res cameras.' What are they going to show me? This is real. Washing machines. They show me washing machines. [6:50] Jonathan Arena: What if nobody goes to a website anymore? What if everybody stays in ChatGPT? [6:50] Jonathan Arena: brands need to be an API, not just a collection of pages. [6:50] Jonathan Arena: Just like Plaid connects banks to apps, we connect brands to agents

Original transcript

[0:04] Hey everybody, my name is Jonathan. I'm [0:06] one of the co-founders of a company [0:08] called New Generation. Everyone seen [0:11] some great talks today. Feeling good? [0:13] >> Almost the end of the day. [0:17] AI is completely rewiring the internet. [0:20] We've hear this all the time. I'm sure [0:22] you've seen this in your conversations [0:24] today. The way people discover products, [0:27] the way traffic finds websites, [0:31] all of this is flowing through LLM [0:32] interfaces. Now, it's a big deal. It's [0:35] probably the biggest shift we've seen in [0:37] commerce in the past 30 years. [0:41] There's a great chance that the next [0:43] visitors to your website, and in this [0:45] talk, we're going to be talking [0:46] specifically about e-commerce websites, [0:50] is an AI agent or an AI referred [0:53] customer. This is somebody who's coming [0:55] to your site from uh an AI chat like [0:58] chatgbt perplexity etc. [1:03] As these LLMs become the discovery layer [1:06] of the internet [1:08] we know that this is changing traffic [1:10] patterns. Over half of customers [1:13] according to Adobe say they plan to use [1:16] AI [1:17] to shop. [1:21] Referral traffic to websites is [1:22] increasing exponentially, 1,900% [1:25] year-over-year. [1:28] That old linear customer funnel that we [1:31] all know and love, finding a website, [1:33] navigating to the website, trying to [1:35] find the product you're looking for, [1:37] maybe buying the thing on the website, [1:39] that's broken. [1:42] It's much more likely today that you [1:44] enter a website through the side door, [1:46] arriving midfunnel on a product details [1:49] page or an SEO landing page. [1:53] We call this the midfunnel because it's [1:56] post discovery. [1:58] Discovery is when somebody finds your [2:00] website and navigates there. But the [2:03] consideration and add to cart steps are [2:06] really where you're interacting with the [2:08] brand the most. [2:10] Maybe you check out and then maybe if a [2:11] brand is really lucky you return as a [2:13] customer and buy something again. [2:17] Now brands have created this you know [2:21] net to try to capture customers [2:25] and this is built you know sometimes [2:27] hundreds of SEO pages these sprawling [2:30] websites [2:32] and they try to do this to meet [2:34] customers where they are. Maybe they get [2:37] a niche search and they have a SEO page [2:38] for that and that's how you find the [2:39] brand. The problem with that is what [2:42] happens if you land here in the website [2:45] tree and you really wanted to land in a [2:48] different category or compare a [2:49] different product. This is where almost [2:51] all of the friction on websites happen [2:54] and is also the source of almost all of [2:56] the bounce rate. [2:59] Pushing customers through that linear [3:01] rigid journey is over. And look at this [3:03] customer funnel. This is average across [3:05] e-commerce websites on the internet. [3:08] It's abysmal. [3:10] We can do much better than this. Out of [3:12] every hundred customers that land on a [3:14] website, maybe only two check out and [3:16] buy something. Most of the drop off is [3:19] in between the consideration and add to [3:22] cart phase. Again, this is where all the [3:23] friction is. [3:27] And the dollars back up this story. [3:31] 65% of marketing budgets are still spent [3:35] on top of funnel discovery while an [3:38] anemic 4% are spent on consideration. [3:42] Now brands aren't dumb. They measure [3:45] everything but historically this [3:47] consideration add to cart phase has been [3:49] a local maxima for brands. They really [3:52] haven't been able to squeeze out any [3:53] more user experience or efficacy from [3:56] these steps. [4:00] But what if that could be different? [4:02] What if you could take all of the [4:05] content and information across the [4:08] entire website [4:10] and put it into an intelligent AI [4:13] website? What might that look like? [4:17] Well, that's exactly what we're building [4:18] at New Generation. [4:23] We see three critical focus areas for [4:25] intelligent commerce infrastructure. And [4:28] this is really again focused on those [4:30] customers in the midfunnel. [4:32] AI storefronts which we're going to talk [4:34] about, consumer agents of course, and [4:38] the ubiquitous LLM chat. [4:42] We're working some of the biggest brands [4:44] in the world to build their first AI [4:46] address on the internet. An AI address, [4:49] think ai.samsung.com, [4:51] ai.nike.com. This is a dedicated [4:54] endpoint to receive AI traffic and a [4:58] place for brands to expose AI [5:00] experiences for customers. [5:04] Now, I'm going to offer a quick case [5:06] study. Let's take Samsung's website. [5:09] Let's imagine you find Samsung.com [5:12] in a chat GPT question. [5:17] Fun fact, Samsung's website is the [5:19] second most visited retail store in the [5:21] entire world. Over 550 million monthly [5:25] visits. Surely they have their act [5:27] together. [5:29] What happens if you search for, you [5:31] know, a very typical question, an easy [5:33] natural language search? Phones that [5:35] have the highest res cameras. What are [5:38] they going to show me? [5:41] This is real. Washing machines. They [5:43] show me washing machines. Unfortunately, [5:45] this is not unique to Samsung. Almost [5:47] all e-commerce websites are woefully [5:49] underprepared for the agentic future [5:51] we're racing towards. [5:54] Enter the AI storefront, our solution. [5:59] This allows any customer to use natural [6:02] language and ask the brand directly what [6:05] they're looking for. And this exists [6:07] over Samsung's product catalog. We're [6:10] able to render in real time a beautiful [6:13] bespoke interface that's onbrand to [6:16] Samsung, featuring generative images, [6:19] titles, product descriptions. We're even [6:22] smart enough to ask intelligent [6:24] follow-up questions to help a customer [6:26] refine their search and journey. [6:28] Clicking on a product reveals a product, [6:32] an AI generated product details page. [6:34] And this is really exciting because we [6:36] can highlight specs, features, relevant [6:40] information, [6:42] contextual to the question the customer [6:44] asked. [6:46] We think this is the future of how [6:47] humans will shop on brand websites. [6:50] But what if it's not a human that visits [6:53] your website? What if it's an agent? [6:56] Are merchants prepared? [6:59] We put this to the test. We tasked [7:02] OpenAI's then called operator now [7:06] chatubt agent to visit samsung.com and [7:10] ai.samsung.com. [7:11] This is the same query and the same [7:13] product catalog. [7:16] Ours is on the right [7:21] and ours is finished already. In real [7:24] time, that was 28 seconds while [7:27] operators struggle for over five and a [7:29] half minutes on samsung.com, the second [7:32] most visited e-commerce website in the [7:34] world. Now, what's happening here? Well, [7:37] we know agents need structured data and [7:39] they struggle with human UI. [7:42] This is things like React components, [7:46] keyword search we've, which we've [7:48] already established doesn't work. And [7:50] they go in and out of product details [7:52] pages, hit 404 pages, 503 pages. It's a [7:55] mess. [7:59] And I get this a lot. But John, what if [8:03] nobody goes to a website anymore? What [8:05] if everybody stays in chat GPT? [8:10] Our answer to that is brands need to be [8:12] an API, not just a collection of pages. [8:14] Just like Plaid connects banks to apps, [8:16] we connect brands to agents. This is [8:19] another advantage of an AI storefront. [8:23] Pick your LLM model of choice. It [8:26] doesn't matter. [8:28] You'll be able to shop directly in the [8:30] interface and you'll be seeing products [8:33] here that are distinct. These are [8:39] rendered here via a direct programmatic [8:41] connection to the brand catalog, not via [8:44] product feeds or SEO scraping. [8:47] You'll have an AI product details page [8:49] as before. And perhaps most exciting, [8:53] with real-time inventory and real-time [8:55] price and a connection to the merchant, [8:57] you'll be able to check out with that [8:58] merchant right here without ever leaving [9:00] the chat. We're working with Visa right [9:02] now to bring this to life. [9:05] We also think this can work in both [9:07] directions. [9:08] So brands sharing data with LLMs and [9:13] customers who again link out from chat [9:15] GPT perhaps they carry with them uh a [9:19] context object. So when they land on the [9:21] website it maintains their AI user [9:23] journey with the context of the history [9:25] of their search [9:28] would be a gamecher for all merchants. [9:30] Brands that win in AI will build for the [9:33] new AI funnel. Now, this is different. [9:37] It focuses on an AI qualified visitor, a [9:40] visitor who wants to use natural [9:42] language to find what they're looking [9:43] for. [9:45] Brands will spend more time and energy [9:48] with on-site consideration, which is [9:50] what we call the brand intelligence, the [9:52] connection with a customer and the brand [9:55] directly. Most brands would not want to [9:57] be disintermediated and give up their [9:59] customer relationships. [10:01] They will reallocate spend to this brand [10:04] intelligence. [10:05] Moving midfunnel [10:09] because the ability to converse with the [10:11] brand is incredibly rich data for a [10:14] merchant. [10:16] Lastly, they will provide exceptional [10:18] experience for both agents and AI [10:21] referred customers. [10:24] All right, let's check out Jonathan. [10:27] Thank you so much.