US Bank Chief AI Officer Urges Focus on Scaling and Client-Centered AI

美国合众银行首席AI官呼吁聚焦规模化及以客户为中心的AI Prashant Mehrotra指出，金融机构必须告别生成式AI的花哨噱头，将可规模化的人工智能融入人工指导与数字化银行渠道中，以带来切实可衡量的客户价值。 Prashant Mehrotra分享了U.S. Bank利用AI同时强化数字化自助体验与人工银行专员能力的战略愿景。他强调，AI领导者兼具落地变革性技术与培训内部员工的双重任务，并将规模化拓展与以客户为中心的设计视为成功的关键要素。该行依托“人机协同”（Human + Digital）理念，运用自然语言与对话式AI简化手机银行App导航、支出分析及客户服务任务，同时始终坚守其深厚的以关系为驱动的银行业务根基。 AI 领导者正面临双重任务：在快速跟上技术迭代步伐的同时，既要交付具有变革性的能力，也要持续对组织和员工开展技术普及与培训。 (0:04) 作为全美第五大银行，美国合众银行（U.S. Bank）依托其领先的移动端平台，围绕客户触点、员工辅助工具和后台运营三大支柱推进 AI 规模化布局。 (0:04) 该行坚持“人机结合”（Human + Digital）战略，强调银行业务以人际关系为本，致力于推动 AI 从噱头转向为客户创造切实、可衡量的价值，而非取代人工服务。 (6:21) 对话式 AI 正在全面升级移动端自助服务体验，支持客户直接用自然语言查询余额或分摊账单，免去在繁琐应用菜单中反复跳转的操作步骤。 (6:21)

According to U.S. Bank Chief AI Officer Prashant Mehrotra, financial institutions must move past generative AI novelties to deploy scalable, client-centric capabilities grounded in a "Human + Digital" strategy that supports rather than replaces relationship banking. The Dual Mandate of AI Leadership (0:04) Prashant Mehrotra describes the current leadership mandate as a duality: executives must simultaneously implement transformative AI capabilities, educate organizational leadership and staff, and keep up with a rapidly shifting technical landscape. As an illustration of industry velocity, Mehrotra notes that OpenAI issued two major product releases within a single week. To manage this transition at U.S. Bank—the fifth-largest bank in the United States, operating under a 162-year-old charter—Mehrotra pursues a three-pronged operational structure. The institution prioritizes rolling out AI into direct customer-facing touchpoints, equipping internal support teams with complementary AI assistance, and connecting adjacent backend operations to establish a complete 360-degree view of client interactions. Human + Digital Integration in Self-Service Channels (6:21) Mehrotra emphasizes the need to pivot away from generative AI "party tricks" toward measurable client utility. Grounded in the premise that banking remains an inherently relationship-driven enterprise, U.S. Bank's "Human + Digital" framework treats AI as an augmentation layer across digital and direct channels, rather than a replacement for human bankers. Within self-service channels, this approach targets friction in mobile application navigation. Instead of requiring customers to click through layered menus or complete multistep flows for tools like Zelle, the bank integrates natural language processing. Customers can type or speak requests to check account balances, split expenses, or break down spending categories, extending prior accessibility initiatives such as early Spanish-language app support. While Mehrotra outlines the strategic intent to embed natural language automation into client workflows, the provided notes conclude before addressing specific governance, data privacy, or technical hurdles involved in scaling these models.

美国合众银行（U.S. Bank）首席人工智能官普拉尚特·梅赫罗特拉（Prashant Mehrotra）指出，银行业的AI战略必须以客户为核心，摒弃单纯的技术噱头，依托“人机结合”（Human + Digital）模式提升自助服务并赋能员工。 AI领导者的双重挑战与三支柱扩展 (0:04) 梅赫罗特拉表示，当今AI业务负责人面临着“双重任务”的挑战：一方面需要推进具有变革性的AI技术落地，另一方面必须持续对企业高层及全员进行技术通识教育，同时还要全力跟进行业的爆发式迭代——例如OpenAI在单周内便连续发布两项重大更新。根据其观点，与以往的技术演进不同，AI取得成功的核心前提是规模化应用，但这必须建立在以客户为中心的框架上。 针对拥有162年历史（特许状签署于林肯总统任期内）、全美资产规模第五大的合众银行，梅赫罗特拉提出了AI的三向推进策略：首先将AI引入面向客户的前端渠道；其次为一线支持团队配备补充性的AI辅助工具；最后将AI能力横向扩展至相邻运营环节，以此构建全方位的客户360度视图。 “人机结合”战略与超越技术噱头 (6:21) 梅赫罗特拉强调，尽管生成式AI在过去两年半中在语音、计算机视觉、内容创作及演示文稿生成等领域展现了引人注目的能力，但银行必须跳出单纯制造惊艳效果的“把戏”，转向追求能够为客户带来可衡量的实际利益。他认为银行业本质上高度依赖人际关系与信任，因此合众银行确立了“人机结合”（Human + Digital）战略。 在该战略下，AI的发展定位是补充和增强人类交互，而非取代人工服务。这一设定的目的在于确保客户在数字自助渠道以及人工直连通道之间切换时，均能获得连贯、无缝的服务体验。 优化移动自助交互与降低操作门槛 (6:21) 在自助服务（DIY）数字化渠道方面，合众银行致力于提升已连续多年排名第一的移动端应用体验。梅赫罗特拉指出，虽然移动端功能全面，但繁琐的菜单层级经常导致操作困扰，例如用户在使用Zelle进行转账时常面临重复步骤。为此，该行通过AI简化路径，让客户可以直接通过语音或文字进行自然语言提问，即时完成查账、分摊午餐账单或分析主要消费类别的操作。 梅赫罗特拉还提到，该项升级建立在银行长期的包容性实践基础之上，例如合众银行曾是全美首家在移动应用中引入西班牙语界面的银行，AI技术的融入旨在进一步打破使用障碍，改善大众化的交互体验。 梅赫罗特拉阐明了合众银行通过“人机结合”与自然语言交互重构客户体验的路径，但提供的信息主要聚焦于前端体验升级与宏观框架，未涉及模型在金融严监管环境下的风险控制机制与数据合规细节。

## Background 该视频由 Ai4 频道于 2025 年 8 月 26 日发布在 YouTube 平台上。页面简介指出，本次演讲由美国银行（U.S. Bank）首席人工智能官 Prashant Mehrotra 主讲，重点展示了美国银行如何通过“人 + 数字”（Human + Digital）战略以及智能自动化与生成式 AI 技术，来提升客户的数字体验并赋予银行员工更强的高效服务与引导能力。 ## TL;DR 演讲围绕百年老牌金融机构美国银行的 AI 落地实践展开。Prashant Mehrotra 指出，AI 的成功关键在于从客户需求出发并实现规模化应用，而非盲目炫技。在自助服务（DIY）渠道，美国银行利用 GenAI 优化后台意图分类，并结合客户数据提供实时的个性化金融服务；在人工服务渠道，银行推出了“银行家助手”，借助 RAG 技术将客服查找资料的时间从几秒钟级拉升，并逐步推广至全美分行万余名一线员工。整体战略坚持以人为本，用强大的 AI 赋能人类体验而非取代人类。 ## Atmosphere & style 整场演讲氛围专业而务实，充满了来自行业顶尖金融机构一线探索者的笃定与坦诚。演讲者语气平实亲切，既坦言技术快速迭代给从业者带来的“双重使命”与追赶压力，又怀揣对技术变革的巨大热情。叙事风格深入浅出，将百余年银行业历史沉淀与最前沿的 GenAI 架构有机结合，给观众带来一种既严谨合规又极具前瞻创新的沉浸感。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=bVkd3gG4vro&t=4s) AI 是极其具变革性的技术，但推动其落地面临着技术极速迭代（如 OpenAI 在单周内多次发布更新）与组织全员教育的双重挑战。 - [00:04](https://www.youtube.com/watch?v=bVkd3gG4vro&t=4s) AI 在企业中取得成功的关键在于实现规模化（Scaling），而这一切必须始终以客户的需求为核心出发点。 - [02:51](https://www.youtube.com/watch?v=bVkd3gG4vro&t=171s) 美国银行拥有 162 年的历史底蕴，并成功打造了全美顶尖的移动 App，其 AI 战略拒绝炫技，专注于为客户创造切实价值。 - [02:51](https://www.youtube.com/watch?v=bVkd3gG4vro&t=171s) 由于银行业极度依赖信任与客户关系，美国银行坚持“人 + 数字”（Human + Digital）战略，AI 的定位是赋能而非取代人类。 - [09:10](https://www.youtube.com/watch?v=bVkd3gG4vro&t=550s) 在自助服务（DIY）渠道中，通过引入语音与文字交互，解决传统 App 菜单层级繁琐的痛点，显著提升用户体验。 - [09:10](https://www.youtube.com/watch?v=bVkd3gG4vro&t=550s) 出于金融合规与数据精准度的考量，美国银行不直接让 GenAI 向客户生成回答，而是利用 GenAI 优化传统 NLP 无法处理的意图分类（Catching fallout）。 - [09:10](https://www.youtube.com/watch?v=bVkd3gG4vro&t=550s) 结合丰富的客户历史数据，App 能够提供一针对一的实时个性化服务，例如精准显示专属 Zelle 额度或定制按揭贷款优惠。 - [16:08](https://www.youtube.com/watch?v=bVkd3gG4vro&t=968s) 通过“银行家助手”（Banker's Assistant）将 AI 赋能给人工客服，将复杂问题的资料检索响应时间从 5-10 分钟缩短至几秒钟，并具备完整的出处追溯（Provenance）。 - [16:08](https://www.youtube.com/watch?v=bVkd3gG4vro&t=968s) 在呼叫中心验证 RAG 与知识库架构后，该 AI 能力被横向扩展至线下数千家分行（覆盖超 10,000 名员工）以及邮件自动化处理系统中。 - [19:07](https://www.youtube.com/watch?v=bVkd3gG4vro&t=1147s) AI 落地应遵循三大原则：以客户需求为起点、聚焦于当下能交付的价值，以及持续提升企业内部全员的 AI 素养（AI Fluency）。 ## Detailed walkthrough 以下是根据您提供的视频文字记录整理的详细演讲实录与分析： ### [00:04](https://www.youtube.com/watch?v=bVkd3gG4vro&t=4s) AI的变革力量与双重使命 演讲者（美国银行首席人工智能官 Prashant Mehrotra）首先向现场观众致意，并提到自己来自芝加哥。他非常感谢大家能来到现场聆听美国银行（U.S. Bank）在 AI 探索历程中的经验分享，同时也很期待能与在座的行业从业者交流互鉴。演讲者感慨道，AI 是他职业生涯中遇到的最具变革性的能力之一。 在与业内同仁的交流中，演讲者深感大家身上承担着一种“双重使命”（duality of mandate）：一方面，每个人都在积极推进或尝试落地 AI 技术，并致力于对员工、管理层及团队成员进行 AI 教育与培训；另一方面，大家又必须在技术日新月异的浪潮中拼命跟上步伐。演讲者坦言保持同步极其艰难，有时一周就会有好几次发布，例如仅在演讲前的一周内，OpenAI 一家公司就发布了两项重大更新。最后，他强调了本次演讲的核心要义：**与许多其他技术不同，AI 成功的关键在于规模化（Scaling），但一切必须从客户（Client/Customer）的需求出发。** - 演讲者：Prashant Mehrotra（美国银行 Chief AI Officer）。 - 行业现状挑战：AI 技术迭代速度极快，例如 OpenAI 在单一一周内就发布了两次重大更新，从业者既要推动落地与员工教育，又要跟上快速更新的技术栈。 ### [02:51](https://www.youtube.com/watch?v=bVkd3gG4vro&t=171s) 美国银行的百余年底蕴与“人 + 数字”战略 演讲者分享了自己在企业环境中推动 AI 发展的切身体会，对目前 AI 渗透到各个业务线、部门和应用场景的深度感到惊叹。如今几乎没有任何一个部门不在思考如何利用 AI，这让他想起了自己当年在本科撰写毕业论文时使用神经元网络（Neural Nets）的经历，深感 AI 技术的发展已经走过了漫长的道路。 随后，演讲者介绍了美国银行的背景：作为美国第五大银行，它拥有全国第二悠久的银行业执照（Charter），该执照是在亚伯拉罕·林肯（Abraham Lincoln）总统任内签署的，距今已有 162 年的历史。正是凭借这种百余年来对客户的专注，这家百年老牌银行成功打造出了全美顶尖的移动银行 App，并连续多年排名第一。演讲者强调，AI 技术的应用绝不能止步于炫技式的“派对小把戏”（party trick），而是要从“这真酷”转向让客户切实感知到收益。结合银行多年来的数字转型经验，美国银行始终坚持“人 + 数字”（Human + Digital）的融合战略，因为银行业是一个极度依赖信任与客户关系的行业，AI 不是要取代人类，而是要将数字化与人工服务完美结合。 ``` [ 客户需求 (Client) ] │ ┌───────────────┴───────────────┐ ▼ ▼ 【DIY 数字渠道】 【人工作业/服务渠道】 (移动 App / 智能语音助手) (呼叫中心 / 线下网点) │ │ └───────────────┬───────────────┘ ▼ [ AI 赋能 (Human + AI) ] ``` - **美国银行历史背景**：美国第 5 大银行，拥有 162 年历史，银行业执照由林肯总统签署。 - **数字渠道成就**：其移动端应用（Mobile App）连续多年被评为全美第一。 - **核心理念**：银行业是高度依赖关系的业务，AI 不是单选题，而是“人 + 数字”的有机结合。 ### [09:10](https://www.youtube.com/watch?v=bVkd3gG4vro&t=550s) 自助服务渠道的 AI 升级：从意图识别到实时个性化 在 DIY（自助服务）渠道方面，虽然移动 App 功能非常丰富，但传统层层翻找菜单的模式依然存在痛点。演讲者举例说，自己每次使用 Zelle 转账失败时，都必须重来并繁琐地查找菜单。如果能直接按下按钮提问并立即获得答案，体验将会大幅提升。美国银行此前已率先在 App 中引入了西班牙语支持，而随着最新 AI 技术的到来，客户可以直接在 App 中通过语音或文字提问，例如查询余额、询问如何平摊餐费、或者查询“前三大消费领域”（甚至戏称为“我的钱都去哪儿了”）。 为了保证金融数据的绝对精准与合规，美国银行在应用 GenAI 时采取了非常稳健的策略： 1. **防止直接幻觉**：出于金融安全的考量，银行并没有将 GenAI 直接面向客户生成最终回答。 2. **意图分类与准确率提升**：针对传统自然语言处理（NLP）识别率有限的问题，银行利用 GenAI 对那些传统 NLP 无法捕捉的漏网请求（fallout）进行更精准的意图分类（Intent Classification），从而不断迭代和提升自助回答的准确率。 3. **实时数据与一针对一个性化（1-on-1 Personalization）**：结合 162 年积累的丰富客户数据，App 能够提供针对性的实时洞察。例如，不再只是回答通用的 Zelle 转账额度，而是直接告诉该客户其专属的个人额度，或是为其定制专属的按揭贷款优惠（Mortgage Offer），从而在每天数千万次的数字互动中提供更快、更具可操作性的回答。 - **应用场景示例**：查询余额、分摊账单、统计消费分类（"Where is my money going"）、查询个人专属 Zelle 额度等。 - **GenAI 落地架构**：GenAI 用于后台的意图分类与优化（Catching fallout），以此间接提升前端回答的精准度，确保金融数据的合规与准确。 ### [16:08](https://www.youtube.com/watch?v=bVkd3gG4vro&t=968s) 赋能业务团队：从呼叫中心到全行万名员工的规模化扩展 除了自助渠道，当客户遇到复杂需求直接寻求人工帮助时，美国银行通过“银行家助手”（Banker's Assistant）将同样的 AI 能力赋能给一线员工。以往客户拨打 1-800 客服电话时，坐席可能需要 5 到 10 分钟去查找资料，而现在 AI 能够在几秒钟内将精准的答案推送给客服人员，并附带完整的推导依据与出处来源（Provenance）。目前，已有数千名呼叫中心代表（Contact Center Reps）在日常工作中深度使用该系统。 演讲者指出，AI 的关键在于“规模化复用”。在呼叫中心验证成功后，美国银行正在将这套基于检索增强生成（RAG）和知识库的 AI 能力扩展至线下全美数千家分行，覆盖超过 10,000 名一线员工。同时，这套技术还被用于客户邮件的自动分类与实时问题排查，在第一时间内协助解决客户诉求。 ``` ┌──────────────────────┐ │ 企业知识库 / LLMs │ └──────────┬───────────┘ │ (RAG 技术) ▼ ┌──────────────────────┐ │ 银行家助手 AI │ └──────────┬───────────┘ │ ┌─────────────────────┼─────────────────────┐ ▼ ▼ ▼ 【呼叫中心客服】 【线下分行机构】 【邮件分类与排查】 (数千名坐席秒级响应) (覆盖 10,000+ 一线员工) (实时捕捉客户问题) ``` - **呼叫中心效能提升**：客服调取复杂解答的时间从传统的 5-10 分钟缩短至几秒钟。 - **技术架构**：基于 RAG（检索增强生成）技术与公共/内部知识库结合，确保可追溯性（Provenance）。 - **规模化扩展（Scaling）**：将呼叫中心验证的能力横向推广至线下分行（覆盖 10,000+ 员工）及邮件处理系统。 ### [19:07](https://www.youtube.com/watch?v=bVkd3gG4vro&t=1147s) 总结：以人为本的 AI 普及与落地原则 在演讲的结尾，演讲者总结了核心观点：**不要为了技术而关注技术，而要关注技术如何提升人类的体验。** 在美国银行，“AI 不是为了取代人类，而是用强大的 AI 技术来赋能人类体验（Human experience powered by amazing AI）”。 最后，演讲者给出了三点建议： 1. **从客户需求出发**：永远从客户的实际需求开始，而不是从“技术能做什么”开始。 2. **注重当下交付**：立足于今天能做什么、今天能交付什么价值。 3. **提升全员 AI 素养（AI Fluency）**：员工是企业最宝贵的财富，要致力于在组织内部建立并提升 AI 素养。 - **核心名言**：“It's not replacement, it's how can human experience be powered by amazing AI.”（AI 不是取代，而是用卓越的 AI 赋能人类体验。） ## Notable quotes > [00:04](https://www.youtube.com/watch?v=bVkd3gG4vro&t=4s) AI 是他职业生涯中遇到的最具变革性的能力之一。 — Prashant Mehrotra > [00:04](https://www.youtube.com/watch?v=bVkd3gG4vro&t=4s) 与许多其他技术不同，AI 成功的关键在于规模化（Scaling），但一切必须从客户（Client/Customer）的需求出发。 — Prashant Mehrotra > [02:51](https://www.youtube.com/watch?v=bVkd3gG4vro&t=171s) AI 技术的应用绝不能止步于炫技式的“派对小把戏”（party trick），而是要从“这真酷”转向让客户切实感知到收益。 — Prashant Mehrotra > [02:51](https://www.youtube.com/watch?v=bVkd3gG4vro&t=171s) 人 + 数字 — Prashant Mehrotra > [09:10](https://www.youtube.com/watch?v=bVkd3gG4vro&t=550s) 一针对一个性化（1-on-1 Personalization） — Prashant Mehrotra > [16:08](https://www.youtube.com/watch?v=bVkd3gG4vro&t=968s) 银行家助手 — Prashant Mehrotra > [19:07](https://www.youtube.com/watch?v=bVkd3gG4vro&t=1147s) 不要为了技术而关注技术，而要关注技术如何提升人类的体验。 — Prashant Mehrotra > [19:07](https://www.youtube.com/watch?v=bVkd3gG4vro&t=1147s) AI 不是为了取代人类，而是用强大的 AI 技术来赋能人类体验（Human experience powered by amazing AI）。 — Prashant Mehrotra > [19:07](https://www.youtube.com/watch?v=bVkd3gG4vro&t=1147s) It's not replacement, it's how can human experience be powered by amazing AI. — Prashant Mehrotra ## Mentioned Prashant Mehrotra, U.S. Bank, Ai4, OpenAI, 亚伯拉罕·林肯, Zelle, GenAI, NLP, RAG, Banker's Assistant ## Video description Prashant Mehrotra, Chief AI Officer, U.S. Bank Abstracts What do smart, enhanced customer experiences look like? At U.S. Bank, our Human + Digital approach combines personalized digital experiences along with great human support and guidance. The next step in this evolution uses AI to level up both the client's digital experience and the banker’s ability to help and guide the client, answering their questions quickly and helpfully. This session will share real-world examples to show how advancements in AI improve the customer experience by providing more choices and convenience, with fast and accurate resolutions to customer needs. It will also show the ways conversational and generative AI can easily help create and enhance personalization – and detail the technology that powers it. Even a best-in-class customer experience can be improved with AI. [Source](https://www.youtube.com/watch?v=bVkd3gG4vro)

[0:04] Prashant Mehrotra: AI 是他职业生涯中遇到的最具变革性的能力之一。 [0:04] Prashant Mehrotra: 与许多其他技术不同，AI 成功的关键在于规模化（Scaling），但一切必须从客户（Client/Customer）的需求出发。 [2:51] Prashant Mehrotra: AI 技术的应用绝不能止步于炫技式的“派对小把戏”（party trick），而是要从“这真酷”转向让客户切实感知到收益。 [2:51] Prashant Mehrotra: 人 + 数字 [9:10] Prashant Mehrotra: 一针对一个性化（1-on-1 Personalization） [16:08] Prashant Mehrotra: 银行家助手 [19:07] Prashant Mehrotra: 不要为了技术而关注技术，而要关注技术如何提升人类的体验。 [19:07] Prashant Mehrotra: AI 不是为了取代人类，而是用强大的 AI 技术来赋能人类体验（Human experience powered by amazing AI）。 [19:07] Prashant Mehrotra: It's not replacement, it's how can human experience be powered by amazing AI.

[0:04] The duality of AI: Transforming Banking While Keeping Pace [0:04] Prashant Mehrotra: scaling is the absolute key to success in AI, but it must always start with the client at the center. [0:04] The fifth-largest bank in the United States holding the country's second-oldest banking charter, signed into law during Abraham Lincoln's presidency (162 years of history). [0:04] U.S. Bank holds a top-ranked mobile banking app, maintaining the number one spot for several consecutive years. [6:21] The Human + Digital Strategy and Elevating DIY Experiences [6:21] Prashant Mehrotra: move away from mere "party tricks" and "wow" moments toward measurable real-world customer benefits. [6:21] Prashant Mehrotra: Human + Digital [6:21] Banking remains a deeply relationship-driven business, so AI is designed to complement human interactions rather than replace them, ensuring seamless customer service across all digital and direct channels.

[0:04] So good morning to everyone and uh good [0:06] afternoon from where I came from. That's [0:08] Chicago home for me. Um so first of all [0:12] thank you very much uh for uh you know [0:14] coming and uh you know hearing some of [0:17] our journey uh on the AI together. Um [0:23] sorry um [0:25] I'm looking forward to you know sharing [0:27] how we are leveraging or introducing AI [0:31] um you know to improve our customer [0:34] experiences [0:35] and I'm uh you know sharing some of that [0:39] uh hopefully it's uh you know helpful to [0:42] uh many of the practitioners in the room [0:46] and also more importantly is is probably [0:49] learning you know from some of them uh [0:51] you know some of you and as I was [0:54] talking with everybody yesterday um it's [0:57] actually interesting um [1:00] AI is uh one of the most transformative [1:04] capabilities that I have come across and [1:06] I was uh talking with a a few of you in [1:09] the room and you know not in the room [1:11] yesterday and uh I was struck by the [1:14] duality of our mandate uh when it comes [1:18] to AI uh it's a transformative [1:20] capability and each one of us is either [1:24] doing or trying to you know execute on [1:28] AI uh and also uh you know educate our [1:31] workforce and and our leadership and our [1:34] team members um while um the previous [1:39] speaker as she was you know mentioning [1:41] while we are also racing to keep [1:43] ourselves up to date I mean I have been [1:46] you know trying for you know some of [1:48] these things myself and What I have come [1:50] across is it's tough to keep you know [1:54] track of [1:56] once a once a week release sometimes [1:58] three times uh you know there are [2:00] releases just last week you know we had [2:02] two major releases from open AAI and [2:04] that's just one company so um so I uh [2:10] you know I I share uh that with you and [2:13] uh I'm I'm looking forward to uh you [2:16] know talking a little bit more about [2:17] that [2:19] if there is one message I want all of [2:21] you to you know walk away with is when [2:24] it comes to AI unlike others scaling is [2:27] the key scaling is the key but it all [2:31] has to start with the client or the [2:33] customer [2:35] that's fun um [2:39] it has to have the c you know the client [2:41] at the center and I'll actually you know [2:43] talk a little bit about how you how we [2:45] are thinking about it Um [2:51] I have been you know um in in a uh [2:54] corporate environment drive driving you [2:56] know AI for a while and I am amazed at [3:00] the level of um you know proliferation [3:04] of AI across different lines of [3:06] business, different functions, different [3:08] use cases. Um and there is not an area [3:12] that in in our bank and in in my [3:16] previous role at least is who is not [3:18] thinking about how to use AI. [3:21] It's an amazing times we are living in [3:24] you know one of the most exciting times [3:25] that I have lived in and uh almost makes [3:29] me wonder how long we have come when it [3:31] comes to AI since my days of using [3:34] neural nets on my uh bachelor's thesis. [3:37] So it's uh you know we have all come a [3:39] long way and uh so [3:43] how we are focused on I'm going to you [3:45] know uh take you on a journey on how we [3:49] are introdu we are introducing AI to our [3:52] customer channels [3:54] complementing the same AI capabilities [3:56] or similar KI capabilities to our uh [4:00] workforce who are supporting those [4:02] customers and then most importantly as I [4:05] talk about scaling is how we are scaling [4:07] to other adjacencies to make sure we are [4:10] making the best use of AI and providing [4:13] a 360 view to the customer. [4:17] So if I can [4:21] so let's talk a little bit about uh let [4:23] me share you know US bank US bank uh you [4:26] know the fifth largest bank in the [4:28] country and uh it has the second oldest [4:32] charter for banking in the country some [4:35] of you might not know I did not know uh [4:37] till a you know few months before I [4:39] joined and uh the charter was signed [4:43] during the presidency of Abe income [4:48] and from one of the you know 162 year [4:50] old bank to our f through our focus on [4:54] customers that we have actually now we [4:58] have one the best mobile app in banking [5:02] and had it you know ranked top number [5:05] one for a few years now. So it is [5:10] again I cannot stress this enough it all [5:13] drives with a customer focus. [5:23] I don't need to tell the folks in this [5:25] room AI can do amazing things and genai [5:28] in the last two and a half years has [5:31] blown our minds. um we have all used in [5:35] personal and professional lives. Um it [5:38] can do you know amazing things with [5:41] voice, vision, creating content, [5:44] creating images, videos. Um I have tried [5:48] it creating you know powerpoints for [5:50] myself and my son has tried you know [5:52] writing his paper which now we watch for [5:55] or is not allowed by the school anymore. [5:58] But these are good use cases. But how do [6:01] we actually bring it to a corporate [6:03] environment and move away from a you [6:06] know just a simple you know I call it [6:09] the party trick uh to real world [6:12] experiences and uh you know benefit uh [6:15] the customers. [6:17] So [6:21] as I said before, we need to scale and [6:24] we need to bring it to real world use [6:26] case to separate it, you know, for from [6:29] hype to real world use cases and uh move [6:34] away from wow this is cool [6:37] to customers telling us that they can [6:39] actually they are seeing the benefits [6:41] from it. [6:44] So how do we do that? We [6:48] we have you know years ago we started on [6:51] our digital journey and it was one of [6:54] the transformative you know uh uh [6:56] capabilities back then and we have [6:59] always believed in [7:01] um using digital and humans together. [7:05] It's not one or the other. It's going to [7:08] be the same thing with AI. [7:11] Banking is a very very relationship [7:14] based you know uh endeavor business and [7:20] trying it's not an eitheror situation [7:23] it's how do we bring together so we [7:26] always believe in you know having humans [7:31] having the digital [7:34] and having them work together. [7:38] So no matter how you contact us, which [7:41] channel, we are there for you. [7:46] So [7:50] as I think about AI, it introduces a new [7:54] frontier in financial services. And the [7:57] way we think about it is there are times [8:00] when customers actually interact with [8:02] us, you know, through our digital [8:03] properties. And sometimes when they [8:06] can't or there's a complex need they [8:08] they actually you know get hold of us [8:09] directly. So we want to make sure that [8:12] no matter what we are providing it is [8:15] being provided in both the channels [8:18] together equally well. So [8:22] we not only pushing you know u this or [8:26] that it's making sure whether you are [8:29] doing it yourself or you are needing our [8:31] help or we are doing it together [8:34] we are you know we have those benefits. [8:39] So I'm really you know u [8:43] excited to you know move on to some of [8:45] the use cases and this one you know [8:48] before I do that and I'm going to you [8:50] know circle back is um as we think about [8:53] design as we think about new [8:55] capabilities as we think about u any new [8:59] technology we are introducing or [9:01] capabilities we are introducing the [9:04] focus is always on the client [9:08] so with that I'm going going to you know [9:10] first tell tell you a little bit about [9:12] upleveling the client experience in a in [9:14] a do do it yourself channel. [9:17] So [9:20] as I talked about it we have one of we [9:22] have the number one mobile app in [9:24] banking and we have been using this to [9:28] transform our customer experience for a [9:30] while. Um [9:36] but we have a lot of capabilities and [9:41] you know users actually use it but you [9:45] know we [9:46] I'll I'll give you one simple example [9:49] that I use is every time I'll transfer [9:52] something in zel it fails you know and I [9:55] have to go do it again but so I can [9:57] actually ask the question I can get the [9:59] answer and I can you know go through to [10:01] the menu because you know over the years [10:04] we have actually added a lot of [10:06] capabilities but what if there was a [10:09] better way what if I can actually go [10:13] press a button ask it a question and get [10:15] an answer without scrolling through or [10:17] going through many options [10:20] um [10:21] I was uh I was talking with my team it's [10:24] uh [10:26] Jetson in 2015 but it's everyday thing [10:29] today right everybody Everybody is using [10:31] it. But what if I can actually just say [10:34] that and then [10:39] now the question is we were doing it for [10:42] years. We introduced we were the first [10:44] ones in the country to introduce Spanish [10:48] in our app. So we have always been [10:51] thinking about a broader and broader [10:53] customer view in it. But what if we can [10:56] actually do more with the latest and [10:58] greatest AI? [11:01] So with that [11:03] now I can now you'll actually see on the [11:05] screen that we can talk to the app. We [11:08] can still use the app but we can talk to [11:10] it about things like what's my balance [11:13] or hey I actually you know uh had lunch [11:16] with somebody. How can I actually split [11:18] it? what are my top three you know [11:20] spending categories [11:23] or as I like to call it uh call it where [11:26] is my money going u not entirely sure if [11:30] it understands the going part but you [11:32] know it's it's really [11:34] this is where the latest and greatest [11:37] capabilities are being introduced now [11:41] we really want to make sure that when we [11:44] are providing an answer to our customers [11:46] we are highly accurate [11:48] It's money. We are dealing with your [11:52] money and my money. I'm you know I'm a [11:54] customer of the bank. So we want to make [11:57] sure that we have a set of regulation [12:00] and we are highly regulated industry but [12:02] we want to make sure anytime we are [12:04] helping you we are providing the right [12:06] answers for you. So what we have started [12:09] to do is now what now anybody who has [12:13] used the old style you know NLP in the [12:16] past they know that it only catches in [12:18] is only accurate for so so much time. [12:21] Now we are actually using it and taking [12:24] the fallout and using Genai to classify [12:28] those intents better [12:30] and in improving it further and further [12:33] to keep increasing the accuracy. So [12:37] while we are not introducing Genai [12:39] directly in the customer experience, we [12:41] are using it to still continue to [12:43] improve their direct experience [12:46] and and and we have to take the right [12:51] you know steps with right controls in [12:53] place to make sure that we are bringing [12:55] the wealth and the holistic capabilities [12:59] to bear on the experiences. [13:01] So um so we have seen a significant [13:04] improvement in what the capability is [13:08] even from a few months ago. Now this one [13:11] I actually love uh is [13:16] as I said 162 years of history. We have [13:19] a lot of data on our customers [13:22] and we have over the period used those [13:24] you know to build insights provide [13:26] better you know capabilities to our [13:29] customer. um how can we actually share [13:33] more of these insights with them? Uh I [13:36] don't know about you, but I uh on a [13:38] daily basis I get about 20 emails on you [13:41] know getting a new mortgage or you know [13:44] um a sale that I just bought from uh [13:46] from a store. Um but you would not get [13:50] that from companies that are using you [13:53] know deep insights into the customer. So [13:58] we can actually start to provide those [14:00] insights directly to those c to our [14:03] customers [14:05] and we can provide the targeted insight. [14:08] So instead of now saying what's the [14:11] general z limit or you know something [14:14] you know more generic about our c uh uh [14:17] the bank [14:20] what if I can actually provide you [14:21] what's your limit maybe you have a [14:24] different limit than you know the [14:25] standard one or you have uh you know uh [14:29] the mortgage offer would be different [14:31] for you than somebody else. [14:35] So this is where we are using real time [14:40] data personalization [14:42] and we want to get to a one-on-one [14:45] level. So I don't know if I can actually [14:48] you know go to uh you know Spain anymore [14:52] in 2500 bucks but you know overall um [14:56] you know we we are now able to not just [15:00] help you with finding things better [15:03] insights better answers and do it in a [15:06] quick and fast manner. [15:10] So um [15:14] as we have tens of millions of [15:17] interactions with our customers [15:20] through our digital app, we are able to [15:23] provide better insights, faster answers [15:27] to our customers that improves our you [15:31] know experience and that is our client [15:34] focus. So as I move uh you know as I [15:38] want to you know circle back on that one [15:40] is [15:42] in in my role you know we always are [15:44] thinking about what is the future of AI [15:46] in banking. So as I think about it is [15:50] it's not just doing you know one thing [15:51] or the other it's pro using the data and [15:54] interaction we have on our customers and [15:57] you know providing them experiences that [16:00] are more rich more informative and more [16:03] actionable. [16:06] So [16:08] with that uh I I talked about you know [16:11] do it uh you know do it yourself. [16:16] Circling back is there are times there's [16:19] a lot of times when customers want to [16:20] reach out to us [16:23] you know uh it's part of the [16:25] relationship. So how can we actually [16:26] provide them the same capabilities in [16:29] the you know when they are reaching out [16:31] whether it's in the uh phone or you know [16:34] through any other channel. So, [16:38] so we have a banker's assistant and uh [16:42] we are you know what we did was [16:48] if you have reached out to 1-800 numbers [16:50] you know how long it takes uh you know [16:53] to get an answer you know just waiting [16:55] for it so what we did was and we'll see [16:58] it here is when you call in [17:02] for US bank we are able to actually now [17:05] provide that answer to the banker within [17:08] seconds. So what used to take you know 5 [17:12] 10 minutes takes a few seconds for us to [17:15] get the answer but not only get the [17:16] answer provide them the highest [17:18] relevancy provide them you know how the [17:21] answer was arrived at and how we are you [17:24] know why it's the right one. So so we [17:27] are able to do that. So now our you know [17:30] couple thousand contact center reps are [17:32] using this on a day-to-day basis. [17:35] And and this is where you know we talked [17:38] about scaling before. [17:41] So I cannot have an AI conference [17:44] without without mentioning rack. So I'm [17:45] going to you know put that up there. But [17:47] this is based on you know uh um more uh [17:51] you know publicly available LLMs but [17:53] based on our knowledge base based on [17:55] where we are using it. So we are able to [17:59] get you know manage the database, manage [18:01] the knowledge source and provide [18:02] provenence. [18:05] I talked about scaling and I think this [18:07] is super important for us is as we think [18:10] about scaling we invested a lot in [18:12] direct experience. We invested a lot in [18:15] uh uh you know our uh contact center but [18:19] now the same capability can be applied [18:21] in our branches. So how can we actually [18:23] you know transfer this you know [18:25] knowledge and investment and [18:27] capabilities across to our different [18:31] uh you know u business uh branches and [18:35] now we are talking you know over 10,000 [18:37] people we have you know we have almost [18:40] we have a lot of branches so now we are [18:43] actually scaling this where we are you [18:45] know pro going to provide the same thing [18:47] same capability [18:49] and then using the same thing for [18:52] classifying emails [18:55] and you know hitting on uh uh you know [18:59] customer issues as they happen. So this [19:02] is where AI can help in real time in [19:04] solving customer problems. [19:07] So with that uh I'm going to you know uh [19:12] go to my key message which is don't [19:15] focus on technology, don't focus on [19:17] capabilities. How can those actually [19:20] help our human experiences? [19:23] That's you know we have we use this [19:27] phrase a lot is it's not replacement [19:30] it's how can human experience be powered [19:33] by amazing AI that's where we are so [19:38] to circle back start with the customer [19:41] needs not what technology can do what [19:44] you can do today what you can deliver [19:47] today and remember people are your you [19:51] know your your resources are your best [19:53] friends. How can you create AI fluency? [19:56] So with that, thank you very much for [19:58] your time. I'll be available for any [20:00] questions afterwards.