Novant Health Uses AI Agents to Trim Healthcare Friction and Workload

Novant Health运用AI智能体减少医疗阻力与工作负荷 在临床和运营工作流中部署人工智能，既减轻了临床医生的行政负担，又扩大了患者就医途径，同时并未取代人工护理。 Vijay Sankararaman介绍了Novant Health利用人工智能消除临床护理、患者体验及运营中系统性阻力的战略[02:42]。他概述了多项重点举措，包括用于患者指导的24/7全天候AI手术智能体[06:43]、用于减轻临床医生行政负荷的环境聆听工具[14:12]，以及用于优化遥测[14:12]和结肠镜检查取消管理的预测性AI[23:26]。此外，Novant Health还利用对话式AI主动触达高风险人群[20:18]，同时强调以解决问题为先的理念，并依托严格的“人在回路”（human-in-the-loop）安全保障与治理机制[028:38], [030:36]。 Novant Health 强调医疗创新的本质在于消除系统性阻力而非盲目追逐新技术，并围绕患者体验、临床护理与运营卓越三大支柱推进 AI 转型。 (2:42) 部署 24/7 全天候 AI 对话智能体，为高风险手术患者提供术前焦虑疏导与术后康复指引，并能一站式处理跨部门的日程变更与账单咨询。 (6:43) 引入环境语音监听及 EMR 自动记录以减轻医生的行政负担与“睡衣时间”，并通过预测性 AI 模型优化住院遥测监护（Telemetry），有效缓解护士的报警疲劳。 (14:12) 在人群健康管理中，利用多渠道 AI 智能体主动触达约 20,000 名高危患者以安排早期肺癌筛查，并向基层诊疗推广糖尿病视网膜病变等 AI 便携诊断设备。 (20:18) 针对全美平均约 8.4% 的结肠镜检查当日取消率，通过预测性机器学习干预就诊缺席并优化资源调配，同时坚持人机协同（Human-in-the-Loop）以保障医疗安全。 (23:26)

Novant Health structures its artificial intelligence strategy around eliminating systemic friction across patient experience, clinician workflows, and operational logistics rather than pursuing novel consumer-facing tech. According to speaker Vijay Sankararaman, the health system pairs problem-driven deployment with human-in-the-loop governance to systematically address institutional bottlenecks. Reconceptualizing Innovation as Friction Reduction (0:04) Sankararaman compares effective healthcare innovation to the functional simplicity of an aluminum can pull tab rather than high-profile consumer gadgets like the iPhone. In healthcare, friction has historically been tolerated as the baseline state: patients struggle to navigate scheduling and billing, clinicians face exhausting administrative burdens, and organizations rely on temporary manual workarounds that accumulate into operational debt. Citing George Bernard Shaw's observation that progress depends on the "unreasonable" person who demands that the environment adapt to them, Sankararaman identifies AI as an unreasonable catalyst to eliminate friction. Novant Health anchors this philosophy across three pillars: patient experience, clinical care, and operational excellence. Continuous Patient Support and Unified Navigation (6:43) To prevent patients from relying on unverified internet searches for medical questions, Novant Health deployed a 24/7 conversational AI agent tailored for high-risk surgical procedures, including colorectal surgeries and joint replacements. The agent provides curated, contextual responses to late-night pre-operative anxiety regarding prep instructions, while clarifying post-operative recovery indicators and wound care outside clinic hours. Conversational agents also resolve fragmented administrative routing. In an illustrative scenario described by Sankararaman, an after-hours patient needing to adjust pre-surgical lab work historically had to wait for regular business hours and navigate multiple transfers for follow-up billing or physical therapy questions. Novant Health's unified agent handles rescheduling, answers billing questions, and provides coverage estimates within a single interaction without multi-department handoffs. Mitigating Clinician Burnout and Alarm Fatigue (14:12) To return time to clinicians, Novant Health adopted ambient listening tools to summarize patient visits and draft clinical charts with patient consent, alongside automated in-basket messaging and nurse handoff workflows. Sankararaman notes that specialists like orthopedic surgeons routinely spend 10 to 15 minutes reviewing fragmented electronic medical record (EMR) tabs before each visit, creating cognitive overload during appointments. In inpatient facilities, Novant Health implemented predictive AI models to analyze continuous telemetry data in real time. Continuous heart and vital-sign monitoring frequently triggers severe nurse "alarm fatigue" and causes unnecessary patient discomfort. The predictive model signals precisely when a patient can safely discontinue telemetry, which reduces alert fatigue, improves patient satisfaction, and releases monitoring equipment for incoming emergency department transfers. Predictive Outreach and Operational Capacity Management (20:18) Novant Health applies predictive analytics to consented EMR records to identify approximately 20,000 individuals across its communities at high risk for early-stage lung cancer. Because manual outreach was operationally unfeasible, the health system deployed a conversational AI agent that conducts multi-channel outreach, answers patient questions, and schedules diagnostic appointments directly. The system plans to expand this outreach framework to breast and colorectal cancers, while deploying point-of-care AI diagnostic devices to allow primary care doctors to perform instant diabetic retinopathy screenings in underserved areas. Operationally, predictive machine learning addresses significant procedural bottlenecks like same-day colonoscopy cancellations, which average approximately 8.4% nationally (about one in twelve appointments). Because late cancellations freeze dedicated four-hour blocks, specialized gastrointestinal staff, and pre-prepped surgical supplies, Novant Health utilizes predictive modeling to evaluate cancellation drivers, intervene proactively, and dynamically reassign clinical resources. Problem-First Prioritization and Safety Governance Novant Health adheres to a problem-first strategy that filters out technological "shiny objects" to prioritize the "vital few" operational and clinical bottlenecks. Rather than asking what new AI tools can do, the organization focuses on where patients experience access delays and where clinicians encounter administrative friction. Because of healthcare's high-stakes nature, the health system implements an AI Trust and Safety Governance protocol to oversee both proprietary models and vendor solutions. This framework enforces strict clinical safety and quality guardrails while ensuring all implementations remain human-in-the-loop, positioning AI to enhance human collaboration and decision-making rather than replace clinical staff. Novant Health emphasizes that its technology plans and AI strategies involve forward-looking projections subject to industry-wide shifts and operational adjustments. The organization maintains that the long-term utility of healthcare AI will depend on rigorous trust and safety governance and keeping human clinicians central to critical decision-making.

Novant Health 数字化高管 Vijay Sankararaman 指出，医疗人工智能的真正价值不在于追求耀眼的新技术，而在于像易拉罐拉环一样以极低感知消除系统性摩擦；该机构将 AI 作为“非理性”的变革驱动力，围绕患者体验、临床诊疗和运营效率重构医疗服务生态。 医疗创新的本质：以“非理性”技术破除系统阻力 (0:04) 在医疗健康领域，人们往往将创新联想为诸如智能手机般引人注目的产品。Sankararaman 提出了不同视角：真正的创新可能像铝制易拉罐上的拉环一样低调而深远，其核心目的在于消除底层流程中的系统性阻力。长期以来，医疗生态系统默认接受了繁复的阻力——患者需要面对复杂的就医门槛与账单流程，医护人员则深陷繁琐的行政事务。各类临时解决方案的堆叠，最终演变成了庞大的组织与运营债务。 引用剧作家 George Bernard Shaw 关于“非理性人推动世界改变”的格言，Sankararaman 认为行业若想突破现状，就不能一味迁就现有流程，而必须引入拒绝妥协的变革力量。Novant Health 将人工智能确立为这种“非理性”变革工具，并围绕三大核心支柱展开全面布局：重构直观且可及的患者就医体验、减轻医护人员行政负荷使其专注于临床本质，以及优化医疗机构的资源分配与运营效率。 重塑患者体验：全天候 AI 智能体跨越流程壁垒 (6:43) 许多患者在出现症状时，往往会首先转向网络搜索引擎或通用大模型求助，而非直接联系医疗机构。为了弥合这一信任与信息缺口，Novant Health 将基于循证医学的上下文感知 AI 深度嵌入自身系统，提供自然语言预约排期、分诊导航、虚拟查房、跌倒监测以及账单咨询服务。针对结直肠手术或关节置换等高风险择期手术，该系统推出了“人在回路”的 24/7 AI 助手，专门解决如 60 至 62 岁髋关节置换患者在术前数周出现的准备焦虑，以及术后关于伤口包扎和恢复指征的即时疑问。 在传统就医场景中，患者深夜致电调整检查时间常常受限于营业时间，并需在排期、账单和康复等多个部门之间反复转接与等待。Sankararaman 以虚拟患者 Margaret 为例说明，统一的智能对话代理不仅能在非工作时间即时确认术前化验改期，还能在单次交互中无缝解答跨部门问题，包括手术自付费用预估及术后物理治疗的医保覆盖细则，彻底消除了传统流程中的部门壁垒与等待延迟。 赋能一线临床：环境感知与智能遥测缓解执业疲劳 (14:12) 医护人员是医疗机构最宝贵的核心资产，但往往承受着巨大的记录压力。以骨科医生 Dr. Valentine 的日常工作为例，仅在午餐前接诊 5 至 6 位患者，每次问诊前就需耗费 10 到 15 分钟在碎片化的电子病历（EMR）各标签页中翻找病史、用药与化验记录，诊疗过程中亦不得不分心应对病历书写。为此，Novant Health 引入了环境感知监听工具，经患者知情同意后自动总结问诊并生成病历草稿，同时辅助处理患者收件箱信息并优化护士排班与交接流程，有效减少了医生的“睡衣时间”（即下班后的文书工作）。 在住院监护环节，过度持续的遥测监护容易引发护士的“报警疲劳”，并降低病患舒适度。Novant Health 部署了预测性 AI 模型，对患者体征数据进行实时分析，精准识别患者何时不再需要遥测监测并提示撤机。这一举措不仅减轻了护理人员的报警负担，提高了患者满意度，还及时释放了紧缺的遥测监测设备，为急诊科（ER）需要转入病房的危重患者腾出了关键资源。 主动健康与精益运营：早期疾病干预与资产动态调度 (20:18) Novant Health 的 AI 应用进一步拓展至院外的高风险人群主动管理。依托合规授权的 EMR 数据与预测分析模型，该机构在服务社区中识别出约 20,000 名早期肺癌高危患者。面对人工电话无法及时覆盖的运营难题，该系统上线了 24 小时语音与文本对话 AI 代理，主动触达患者、解答疑虑并直接协助预约筛查排期，该模式正逐步推广至乳腺癌和结直肠癌领域。针对专家资源匮乏的偏远与低医疗资源地区，机构还推广了基于 AI 的床旁物理诊断设备，使全科医生在常规诊室即可完成糖尿病视网膜病变快速筛查并实现即时转诊。 在运营优化层面，Sankararaman 强调了预防性结肠镜检查的资源调度挑战：全美范围内该检查当日取消率平均高达约 8.4%（约十二分之一），导致包含消化科医师、麻醉师、专业护士及专用物资的 4 小时手术时段陷入闲置。Novant Health 利用机器学习预测模型深入分析预约取消的诱因，提前主动介入以改善患者履约率，并在出现变动时动态调配医疗资产和耗材，这一精细化管理正延伸至供应链和收入周期等核心环节。 治理与人本原则：坚持以问题为导向的 AI 信任架构 Novant Health 在推进数字化变革中恪守“问题优先”策略，避免陷入技术狂热，而是严格聚焦于患者体验中的断点与医护人员的行政负担，坚持解决“关键少数”问题。由于医疗决策直接关乎生命健康，机构建立了严格的“AI 信任与安全治理”（AI Trust and Safety Governance）体系，与内外利益相关方紧密协作，对自研算法和外部第三方采购方案进行全周期评估，将临床质量、安全合规与机构信任置于纯技术指标之上。 Sankararaman 总结强调，横跨患者服务、临床照护与后台运营的全部 AI 实践，核心基调始终是坚持“人在回路”（Human-in-the-Loop）。人工智能在 Novant Health 的定位绝非取代专业医护人员，而是作为消除流程摩擦的赋能工具，旨在将医患各方更紧密地凝聚在一起，辅助人类专家做出更精准、更具温度的医疗决策。 Novant Health 展示了将 AI 深度融入医疗各环节以系统性降本增效的实践路径。不过，演讲开篇的安全港声明同样指出，文中所涉及的技术愿景、预测分析及战略规划均受制于不可控的行业内外部变量，AI 在临床合规与大规模落地中的实际表现仍需长期审慎检验。

## Background 该视频由专业峰会组织 Ai4 于 2026 年 8 月 20 日发布在 YouTube 平台。视频由 Novant Health 首席 AI 官 Vijay Sankararaman 主讲，围绕医疗健康领域的 AI 规模化转型展开，分享了医院如何从零散试点走向企业级应用，在创造可衡量业务价值的同时改善患者体验并切实减轻临床医护负担。 ## TL;DR 演讲将医疗 AI 的本质比作易拉罐上的拉环，强调其核心使命不是追求炫目噱头，而是消除医疗系统长期存在的系统性摩擦。Novant Health 确立了患者体验、临床护理与卓越运营三大落地支柱：在患者端通过 24/7 全天候对话式 AI 代理提供贯穿围手术期指导、夜间预约改期与账单咨询的一站式支持；在临床端利用环境智能监听和收件箱自动化减少医生的“睡衣时间”，并通过预测模型优化遥测设备周转与高危疾病主动筛查；在运营端重点化解结肠镜检查临时取消等资源瓶颈，始终秉持以痛点为导向和“人在回路”的 AI 信任与安全治理体系。 ## Atmosphere & style 整场演讲呈现出务实、睿智且极具人本关怀的基调。开篇通过询问观众对创新的定义并以“瞬移”和“易拉罐拉环”对比破题，迅速拉近与受众的距离，巧妙地把技术崇拜转化为对流程痛点的反思。讲者语调沉稳而坚定，不仅真切道出医护人员被堆积文书与“睡衣时间”绑架的职业困境，更结合 Margaret 等典型患者案例与硬核运营数据进行层层拆解，使整场分享兼具企业高管的战略前瞻性与一线临床落地的深切共情。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=4s) 创新的本质不只是像 iPhone 那样耀眼夺目，许多朴实无华的日常创新（如易拉罐拉环）核心目的在于消除深层系统摩擦与流程阻力。 - [02:42](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=162s) 医疗系统长期将低效摩擦视作常态，Novant Health 围绕患者体验、临床护理与卓越运营三大支柱，以 AI 推动底层转型以打破流程惯性。 - [06:43](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=403s) 针对早期疾病搜索脱离医疗系统的问题，通过自然语言 AI 提供临床验证的专业问答，无缝衔接挂号预约、导医与药房服务。 - [06:43](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=403s) 针对高风险复杂手术推出全天候 24/7 围手术期 AI 智能体，配合“人在回路”审核，实时解答患者在深夜出现的照护与准备疑虑。 - [15:44](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=944s) 通过知情同意下的环境智能监听（Ambient Listening）与智能收件箱管理，减少医生下班加班录入病历的“睡衣时间”，让医护回归“顶格执业”。 - [15:44](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=944s) 引入 AI 预测模型智能评估住院患者遥测监测（Telemetry）需求并提示停用，既能缓解医护警报疲劳，又能加速急诊转病房设备的流转周转率。 - [20:18](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1218s) 结合 EMR 与预测模型锁定约 2 万名潜在肺癌高危人群，借助 24/7 对话式 AI 代理通过语音与短信主动触达并完成筛查预约。 - [20:18](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1218s) 将搭载 AI 算法的便携筛查设备下沉至基层全科门诊，打破地理屏障解决偏远地区糖尿病视网膜病变的早筛难题。 - [23:26](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1406s) 利用 AI 深入分析结肠镜等检查的高取消率痛点，实现全流程资源优化，并通过 24/7 对话式 AI 代理一站式解决预约改期、账单与理疗报销咨询。 - [28:38](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1718s) AI 落地坚持从真实痛点与摩擦切入，建立严格的 AI 信任与安全治理体系，并始终确保“人在回路”，旨在赋能协作而非替代医护人员。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=4s) 开场与创新的本质：从 iPhone 到易拉罐拉环 在演讲伊始，Novant Health 首席 AI 官 Vijay Sankararaman 首先提出了前瞻性陈述（Safe Harbor Statement）免责声明，指出医疗健康行业的规划、优先级以及技术与 AI 相关的探索具有前瞻性，受多种不可控因素影响可能发生变化。随后，他向现场观众抛出一个核心问题：“当你想到‘创新’时，脑海中会浮现什么？”观众席陆续给出了“新产品”、“改进”、“用不同的方式做事”，甚至还有一位观众提到“瞬移 / 远距离传送（Tele-transportation）”，引发现场一阵轻笑。Sankararaman 指出，大众最常联想到的创新往往像 iPhone 那样——外观光鲜流畅（sleek）、直观（intuitive）、拿在手中极具吸引力，属于显而易见的技术奇迹。 然而，他引导大家思考另一种虽然不显眼但同样具有变革意义的日常创新：易拉罐上的“拉环（pull tab）”。拉环的诞生彻底颠覆了消费者饮用罐装饮料的方式，重塑了工业包装、集装箱化运输乃至整个供应链体系。Sankararaman 强调，创新并不总是像 iPhone 那样耀眼夺目，很多伟大的创新虽然朴实无华，但它们的核心目的极其明确——消除摩擦（removes friction）。 ```mermaid flowchart LR accTitle: 创新的两种形态 subgraph S1["显性创新 (High-Visibility)"] A["iPhone / 消费电子"] --> B["前沿体验与视觉颠覆"] end subgraph S2["隐性消除摩擦 (Friction-Removal)"] C["易拉罐拉环 (Pull Tab)"] --> D["重构包装与供应链流程"] E["医疗场景中的 AI 应用"] --> F["消除行政与就医阻力"] end ``` --- ### [02:42](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=162s) 医疗生态中的系统性摩擦与“不合理的人” Sankararaman 直言，医疗行业多年来已经逐渐习惯将“摩擦”视作常态，整个生态系统中的各方都在被动适应这种低效： - **患者端**：最大的痛苦往往不在于临床治疗本身，而在于就医周边的一切——寻找和获取医疗资源（access）、繁琐的挂号流程以及复杂的支付账单周期。 - **医护端**：医生和护士每天被堆积如山的行政杂务所淹没，深陷流程惯性（process inertia），无法专注于自身最擅长的专业诊疗。 - **机构与生态端**：医疗机构、保险支付方（payers）、技术提供商等各方利益相关者，早已默认了这套充满惯性的庞大体系。 面对这些痛点，系统中的人往往会寻找各种“临时变通方案（workarounds）”和打补丁式的快速修复（quick fixes）。但这些变通方案日积月累，最终演变成沉重的系统债务。Sankararaman 引用了萧伯纳（George Bernard Shaw）的名言，指出推动变革的永远不是那些默默适应世界的“合理的人”，而是那些拒绝妥协、试图改变现状的“不合理的人（the unreasonable man）”。在当今医疗变革的语境下，AI 正是打破这一系统性惯性、推动底层转型的关键力量。在 Novant Health，AI 的战略布局紧紧围绕患者、临床医生和团队成员，聚焦于三大核心支柱： 1. **患者体验（Patient Experience）** 2. **临床护理（Clinical Care）** 3. **卓越运营（Operational Excellence）** --- ### [06:43](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=403s) 重塑患者体验：从门诊导医到全天候高风险手术导航 在患者体验这一支柱上，Novant Health 的目标是提供无摩擦、可及、负担得起且能够“主动迎合患者需求”的服务。Sankararaman 提到，过去当人们感到身体不适时，第一反应往往是打开手机充当“Google 医生”或“ChatGPT 先生/女士”，导致疾病搜索与咨询发生在正规医疗系统之外，医疗机构根本没有早期介入的机会。为此，Novant Health 正在通过基于自然语言的 AI 交互弥合这一鸿沟，将经过临床验证的专业知识置于丰富的医疗上下文中，安全地解答患者疑问，并无缝衔接至预约挂号、医疗导航和药房服务。 此外，Novant Health 将 AI 深度融入患者全流程照护体验中： - **住院与就医透明度**：利用虚拟查房（virtual rounding）及智能通知，以直观的方式告知家属医生何时查房，减少焦虑与无效等待。 - **环境智能安全监控**：借助超越人类感官的 AI 视觉与感知技术，实现跌倒、滑倒的提前预警与防范。 - **财务与账单支持**：利用 AI 解答复杂的医疗账单、支付选项、保险覆盖范围及财务援助问题。 针对结直肠手术或髋膝关节置换等高风险、长照护周期的复杂手术，患者在术前术后往往面临巨大的信息焦虑。Sankararaman 举例说明：一名 62 岁的女性患者在接受自选膝关节置换手术前 15 天，会对饮食禁忌、身体准备、术后伤口包扎与护理注意事项等产生无数疑问。这些问题往往出现在深夜，而此时临床团队无法实时在线。为此，Novant Health 推出了具备上下文专业知识的 **24/7 全天候 AI 智能体（AI Agent）**，并设立了“人在回路（human-in-the-loop）”机制以确保回复内容的专业与准确，让患者在围手术期全天候感受到医疗团队的陪伴与支持。 ```mermaid flowchart TD accTitle: AI 全天候患者全流程交互与导航架构 subgraph Intake["前期介入 & 入口"] A["患者出现症状 / 搜索疑问"] --> B["安全、上下文丰富的自然语言 AI 接口"] B --> C["门诊预约 & 财务/医保账单答疑"] end subgraph Perioperative["围手术期 24/7 智能导航"] D["高风险手术规划 (如髋膝关节置换)"] --> E["24/7 AI 术前准备/饮食/术后护理智能体"] E --> F{"人在回路 (Human-in-the-Loop) 审核"} F --> G["患者获得即时、准确的连续照护支持"] end Intake --> Perioperative ``` --- ### [14:12](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=852s) 解放临床生产力：让医护回归“顶格执业” 在阐述完患者体验后，Sankararaman 将话题转向第二大核心支柱——**临床护理（Clinical Care）**。医生与护士是医疗系统服务社区最宝贵的财富，但现实中他们的日常工作被各种重复繁琐的行政事务层层包围，严重阻碍了他们进行“顶格执业（practicing at the top of their license）”，即充分发挥其最高专业执业资质的价值。 为了改变这一现状，Novant Health 致力于借助人工智能和数字化技术，与临床医生和患者紧密协作，重塑临床医生的执业形态与业务流程。Sankararaman 随后以该院骨科医生 Dr. Valentine 的实际工作故事为例，开始深入阐述 AI 如何在真实的临床诊疗场景中协助医生减轻文书与行政负担…… ### [15:44](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=944s) 减轻医护行政负担与优化遥测监测流程 医生每天在午休前通常就要接待 5 到 6 位预约患者。在走进每一间诊室之前，医生都需要花费至少 10 到 15 分钟来准备病历、阅读洞察信息、了解患者近期的病情变化、药房与药物选择以及化验结果等。这些分散在电子病历系统（EMR）各处的数据，需要医生逐一收集、整合与处理。当医生站在患者面前时，他们全身心倾注于倾听与关怀，但脑海中也难免会焦虑后续如何写完病历、补全图表并赶往下一位患者。为了缓解这种行政重负，Novant Health 引入了系列 AI 辅助产品：在获得患者知情同意的前提下使用环境智能监听（Ambient Listening）自动生成病历摘要，并利用智能收件箱管理（Inbasket Management）自动拟定回复草稿，从而切实减少医生的“睡衣时间”（Pajama Time，即下班回家后加班录入病历的时间）。 这种 AI 赋能不仅限于医生，更延伸至护理和药剂团队。团队正在探索如何改进护士交接班、减轻护士长排班负担，并打通临床医生与药剂师之间的端到端闭环。以住院部的心电遥测（Telemetry）为例，该设备虽能实时监测患者的心跳与生命体征，但也常常引发“警报疲劳”（Alarm Fatigue），且患者也不愿长时间身上贴满传感器与导线。过去只能由医护人员被动观察并在确认无需使用时拆除；如今通过接入 AI 预测模型，系统能精准预测何时不再需要遥测监测，及时提示停用。这不仅减轻了医护的警报疲劳、提升了患者体验，还让急诊科（ER）准备转入普通病房的新患者能够更快获取并使用这些宝贵的遥测设备。 ```mermaid flowchart TD accTitle: 临床工作流中的 AI 赋能与资源优化 subgraph S1["临床工作流减负"] A["多源离散数据整合<br/>(EMR/检验/用药)"] --> B["AI 环境智能监听与摘要"] B --> C["AI 自动拟定收件箱草稿"] C --> D["减少睡衣时间 (Pajama Time)"] end subgraph S2["遥测监测智能管理"] E["遥测设备实时生命体征"] --> F["AI 预测模型评估"] F --> G["智能提示及时停用"] G --> H["缓解警报疲劳 & 释放设备给新患者"] end ``` - **医生准备负荷**：每位患者就诊前需 10-15 分钟阅读分散在 EMR 中的化验、用药及病历数据。 - **核心落地产品**：知情同意下的环境智能监听生成摘要、智能收件箱自动草拟回复。 - **赋能角色延展**：从临床医生扩展至护士交接排班与药剂师协同。 - **遥测优化价值**：降低医护警报疲劳，改善住院患者舒适度，加速急诊转病房设备的周转率。 --- ### [20:18](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1218s) 高危疾病早期筛查与基层 AI 赋能 面对患者尚未自觉的高危疾病，早期主动干预至关重要。以肺癌为例，通过分析患者授权的 EMR 数据并结合预测模型，系统在社区中精准筛查出约 20,000 名潜在高危患者。若单靠人工拨打电话通知这 2 万人预约筛查，人力完全无法在黄金时间内覆盖。为此，Novant Health 部署了 24/7 全天候多模态对话式 AI 代理（Conversational AI Agent），通过语音与短信主动触达患者，在医生授权指导下提供科普说明、解答疑问并完成筛查预约，确保患者在病情恶化前获得干预。 团队进一步将该模式拓展至结直肠癌、乳腺癌等与时间赛跑的高危病种筛查。而在糖尿病视网膜病变（Diabetic Retinopathy）等领域，偏远农村地区的患者往往缺乏眼科专科检查条件。Novant Health 通过与合作伙伴协同，将搭载 AI 算法的便携物理设备直接交付到基层全科医生（PCP）手中，使患者在初级诊疗环节即可完成眼底筛查，并在发现异常时快速转诊至专科医生，有效打破了医疗资源的地理屏障。 - **肺癌高危筛查规模**：利用 EMR 与预测模型锁定约 20,000 名潜在高危人群。 - **全天候触达机制**：通过对话式 AI 提供 24/7 语音与短信触达、答疑与预约。 - **病种拓展规划**：复用至结直肠癌、乳腺癌等早筛项目。 - **基层设备赋能**：将集成 AI 的便携筛查设备下沉至全科门诊，解决偏远地区眼底早筛难题。 --- ### [23:26](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1406s) 运营卓越：肠镜预约优化与端到端全天候智能服务 在医疗运营卓越（Operations Excellence）层面，Novant Health 重点解决导致资源浪费与流程摩擦的核心痛点。以结肠镜检查（Colonoscopy）为例，该项预防性筛查对患者而言需要繁琐的前期准备，对院方而言则涉及长达 4 小时的手术时段锁定，涵盖麻醉师、主刀医生、护士团队及专业器械耗材。全美同日取消率平均高达 8.4%（约 1/12 的检查在当天临时取消），一旦发生临阵取消，该时段便无法释放给其他等待患者，导致昂贵的医护与物资资源闲置。团队运用 AI 深入分析取消原因、实施精准介入引导，并智能调配医护和物资资产，显著提升了资源利用率。 为了展示 AI 如何彻底改变端到端患者体验，讲者举了一个典型场景：52 岁的患者 Margaret 预定数天后进行骨科髋关节置换手术，但在术前准备阶段前一天晚上 8 点突发状况，急需调整术前的化验检查时间。过去这种夜间来电接线员无法直接修改系统，只能让患者第二天一早再次致电门诊协调；而现在，24/7 对话式 AI 能够直接识别诉求并实时办理化验改期。不仅如此，当 Margaret 顺便咨询手术账单问题，或询问主治医生提及的术后物理治疗（理疗选项、费用及医保报销）时，同一个 AI 代理都能连贯解答，无需像以往那样经历多次转接与漫长等待。 ```mermaid flowchart LR accTitle: 患者夜间复杂服务诉求 AI 一站式解决流 A["患者晚8点致电<br/>(髋关节术前诉求)"] --> B["24/7 对话式 AI 代理"] subgraph S["一站式即时业务处理"] B --> C["1. 术前化验预约即时改期"] B --> D["2. 手术账单明细在线解答"] B --> E["3. 术后理疗方案及医保咨询"] end S --> F["无需人工二次回拨/多次转接<br/>患者获得无缝体验"] ``` - **结肠镜运营痛点**：全美当天取消率约 8.4%（1/12），导致 4 小时手术块及麻醉、护理团队资源空置。 - **全流程触点覆盖**：将 AI 赋能拓展至供应链、联络中心、收入周期管理（Revenue Cycle）及医院整体运营。 - **夜间一站式服务范例**：AI 代理同时跨界解决“预约改期”、“账单查询”与“术后理疗/医保覆盖”，彻底消除转接等待。 --- ### [28:38](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1718s) 以问题为导向的 AI 哲学与信任安全治理 Novant Health 在推进 AI 变革时始终坚持一套核心原则：绝不从“AI 能做什么”出发，而是从“问题出在哪里”、“患者在哪里受阻”、“流程摩擦在哪”、“是什么给医护造成了沉重的行政负担”切入。团队坚持“聚焦于至关重要的少数，而非微不足道的多数”（Vital few over trivial many），将解决问题的执念完全放在患者体验上，这构成了运营卓越的本质。 在医疗健康这一特殊领域，AI 的安全与可信治理至关重要。团队建立了严格的“AI 信任与安全治理体系”（AI Trust and Safety Governance），与内部各利益相关方协同，并与外部监管及行业机构保持紧密联动。无论是自主研发的模型、采购的商业模型还是合作方的产品，都必须通过严格的安全、质量与信任审查。贯穿患者体验、临床护理与医院运营这三大支柱的共同主线，是利用智能变革消除摩擦、辅助优质决策并始终保持“人类在环”（Human-in-the-loop）。AI 从来不是为了取代人类，而是重构协作模式、凝聚各方智慧并做出更优决策的机制。 - **立项决策逻辑**：以业务摩擦与医患痛点为唯一起点，拒绝追逐噱头。 - **AI 信任与安全治理**：对自研、采购及第三方模型实行全流程安全与质量把控。 - **核心发展理念**：坚持“人类在环”，以 AI 促进协作与减负，而非替代专业人员。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=4s) 当你想到‘创新’时，脑海中会浮现什么？ — Vijay Sankararaman > [00:04](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=4s) 很多伟大的创新虽然朴实无华，但它们的核心目的极其明确——消除摩擦。 — Vijay Sankararaman > [02:42](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=162s) 在医疗健康生态系统中，各方早已习惯将摩擦视作常态，整个系统都在被动适应这种低效。 — Vijay Sankararaman > [02:42](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=162s) 推动变革的永远不是那些默默适应世界的‘合理的人’，而是那些拒绝妥协、试图改变现状的‘不合理的人’。 — Vijay Sankararaman > [06:43](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=403s) 过去当人们感到身体不适时，第一反应往往是打开手机充当‘Google 医生’或‘ChatGPT 先生/女士’，导致咨询发生在正规医疗系统之外。 — Vijay Sankararaman > [14:12](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=852s) 医生与护士是医疗系统最宝贵的财富，但日常重复繁琐的行政事务严重阻碍了他们进行‘顶格执业’。 — Vijay Sankararaman > [15:44](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=944s) 利用 AI 环境智能监听和智能收件箱管理，从而切实减少医生的‘睡衣时间’。 — Vijay Sankararaman > [20:18](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1218s) 在肺癌早筛中，系统精准锁定约 20,000 名潜在高危患者，单靠人工根本无法在黄金时间内覆盖，而 24/7 对话式 AI 代理打破了这一瓶颈。 — Vijay Sankararaman > [23:26](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1406s) 结肠镜全美同日取消率高达 8.4%，一旦临阵取消便会导致长达 4 小时的昂贵医护与手术资源闲置。 — Vijay Sankararaman > [28:38](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1718s) 绝不从‘AI 能做什么’出发，而是从‘问题出在哪里’、‘患者在哪里受阻’切入，聚焦于至关重要的少数，而非微不足道的多数。 — Vijay Sankararaman > [28:38](https://www.youtube.com/watch?v=Bw2N1YjFTK8&t=1718s) AI 从来不是为了取代人类，而是重构协作模式、凝聚各方智慧并做出更优决策的机制。 — Vijay Sankararaman ## Mentioned Vijay Sankararaman, Novant Health, Ai4, iPhone, George Bernard Shaw, Dr. Valentine, Google, ChatGPT, EMR (电子病历系统), Telemetry (心电遥测), Ambient Listening (环境智能监听), Inbasket Management (智能收件箱管理) ## Video description Vijay Sankararaman - Chief AI Officer | Board Advisor | Product & Technology, Novant Health AI isn't replacing the human touch in patient care - it is making room for more of it. Sharing our approach to deliver remarkable patient experience while reducing clinician burden with AI. How we are moving from AI pilots to enterprise-scale transformation across patient experience and operational excellence. How that may apply to peer industry verticals. How our transformation focus is anchored in measurable business value than shiny object syndrome. [Source](https://www.youtube.com/watch?v=Bw2N1YjFTK8)

[0:04] Vijay Sankararaman: 当你想到‘创新’时，脑海中会浮现什么？ [0:04] Vijay Sankararaman: 很多伟大的创新虽然朴实无华，但它们的核心目的极其明确——消除摩擦。 [2:42] Vijay Sankararaman: 在医疗健康生态系统中，各方早已习惯将摩擦视作常态，整个系统都在被动适应这种低效。 [2:42] Vijay Sankararaman: 推动变革的永远不是那些默默适应世界的‘合理的人’，而是那些拒绝妥协、试图改变现状的‘不合理的人’。 [6:43] Vijay Sankararaman: 过去当人们感到身体不适时，第一反应往往是打开手机充当‘Google 医生’或‘ChatGPT 先生/女士’，导致咨询发生在正规医疗系统之外。 [14:12] Vijay Sankararaman: 医生与护士是医疗系统最宝贵的财富，但日常重复繁琐的行政事务严重阻碍了他们进行‘顶格执业’。 [15:44] Vijay Sankararaman: 利用 AI 环境智能监听和智能收件箱管理，从而切实减少医生的‘睡衣时间’。 [20:18] Vijay Sankararaman: 在肺癌早筛中，系统精准锁定约 20,000 名潜在高危患者，单靠人工根本无法在黄金时间内覆盖，而 24/7 对话式 AI 代理打破了这一瓶颈。 [23:26] Vijay Sankararaman: 结肠镜全美同日取消率高达 8.4%，一旦临阵取消便会导致长达 4 小时的昂贵医护与手术资源闲置。 [28:38] Vijay Sankararaman: 绝不从‘AI 能做什么’出发，而是从‘问题出在哪里’、‘患者在哪里受阻’切入，聚焦于至关重要的少数，而非微不足道的多数。 [28:38] Vijay Sankararaman: AI 从来不是为了取代人类，而是重构协作模式、凝聚各方智慧并做出更优决策的机制。

[6:43] Vijay Sankararaman: Mr. or Miss Google [6:43] Vijay Sankararaman: ChatGPT [14:12] Vijay Sankararaman: pajama time [28:38] Vijay Sankararaman: vital few over the trivial many [30:36] Vijay Sankararaman: AI for us is not about replacing humans. AI for us is a mechanism that reimagining pulling people together, collaborate together, make better decisions.

[0:04] Some of what I'm going to be talking um from a  Noant health perspective includes information [0:10] about you know or plans or priorities or work  with technology and AI forward-looking in nature [0:19] as with any other healthcare organization you  guys know of these forward-looking statements [0:27] do contain information that is materially subject  to change because of some of the uncontrollable [0:34] factors you know that comes around that's what's  reflected in the safe harbor statement you [0:39] just want to put it out there when you think of  innovation what comes to your mind thank you new [0:46] products new products new things what things what  uh what image comes to your mind improvement doing [0:54] things differently is that a particular product  comes to your mind when you think of innovation [0:59] Tele transportation. Okay, that was interesting.  I've done this many times. I never heard that one. [1:03] But that's that's awesome. Like you know when we  think of innovation, here [snorts] is one that [1:09] a lot of us talk about, right? Sleek, intuitive,  compelling to hold. No question this is something [1:18] that is a marvel of its age. An iPhone visible  right now. When we think of some innovations that [1:27] is transformative in nature but not as visible  say like an iPhone thoughts anything that you [1:35] can think of things that are not that visible  but every day you see them you think man that's [1:42] really innovative what about a pull tab right like  you know years ago when this was first introduced [1:55] It's absolute magic, right? It not only changed  the way how consumers used an aluminum can, [2:03] how the industry packed things in it,  containerized it, moved it through supply [2:09] chain revolution. You can keep adding on and  on. All that with a simple almost invisible [2:19] pull tab. Innovations do not always have to be  in yourrface gliding iPhone. Not nothing to do [2:28] with iPhone. I love iPhone by the way. But  it could also be as simple, less glamorous. [2:33] But the purpose like the pull tab removes  friction. Let's use the word friction here. [2:42] So when we think of health care, health care over  the years has come to accept friction as the norm. [2:55] Everybody's just used to it or getting used to it.  You think about a patient, it's not the clinical [3:02] care where they find the most problem. It's  everything around it. getting to the care, talking [3:07] about access to the care, talk about the the whole  payment cycle around the care. You talk about the [3:14] clinician side of it, it is the administrative  pile that buries them day in and day out. [3:24] It's the ability for them to stay on top  of what they can do or they should do, [3:30] but the process inertia that's  wrapped around them causes that. [3:35] You think about organizations everything around  it the providers the payers the stakeholders the [3:42] technology who have you have come to accept  the fact that it is built with inertia. [3:55] Now see every once in a while when we live  in live in the world we come up with inertia [4:01] we come up with the pain points I talked  about people find workarounds healthcare [4:07] there's lots of workarounds we all can talk  about it clinicians have some patients have [4:12] some administrators have some the workarounds  is the way of life and we find quick fixes [4:20] along the way and these quick fixes pile  up and over time That's a debt by itself. [4:29] But every once in a while, regardless of what  the system as a friction, there is always [4:34] something. There's always someone challenges the  unreasonability around it challenges the why there [4:42] should be. You know, there's a reason I like  this quote from George Bernard. The essence of [4:51] this quote, it's the context of my topic here. It  is not what a reasonable person accepts and moves [5:00] around is going to be the beacon of change. It is  always the unreasonable. The one who's looking to [5:07] dramatically shift the game is the one or is the  thing that's going to help create transformation. [5:17] That my audience is the power of AI, artificial  intelligence. You got you have heard a lot in [5:25] the last day or two and you will hear more but  buried in all of that is for us at no one health [5:34] when we look at how we want to transform the way  we see health care it's all centered around the [5:43] patient the clinicians and the team members  who work tirelessly to make things happen. [5:50] When we think about how we can leverage the  power of artificial intelligence, we are [5:58] centered around the fact on three things. Patient  experience, clinical care and the operational [6:06] excellence. And what we mean by each of these  will expound that in the next few minutes. But [6:12] the fact of the matter lies is that when we  think about friction, an inertia that exists [6:20] around the ecosystem that we operate. Our goal and  our cause remains to bring remarkable experience [6:27] and remarkable care to the communities we serve  to ensure when we deal with the variety of factors [6:35] around us, how do we transform it, how do  we remove the friction and go from there. [6:43] Now when we think about patient  experience, we think as a patient [6:50] what matters the most. Now when we think about  how a care can be delivered friction-f free, [6:59] accessible, affordable, and one that can  meet where the patient needs it to be met, [7:10] not the other way around. Now, there's a  lot of factors that goes in looking at it. [7:15] We think about beyond just scheduling, beyond  just having to have a way for them to enter [7:22] into the system. It is a lot about how  that can be rendered in a way that is [7:29] easy for them to intuitively get there. And  there are a number of things we look at to [7:36] give you a sense of some of these things.  Look, you know, one of the things we we [7:42] look at that is putting yourself in the  shoes of someone who's looking for care. [7:51] One of the first things they want to know is  that is the care available for them? Is the care [7:56] they're looking for is available? Do I have the  clinician that I'm looking for is available? Do [8:01] I have a provider facility I'm looking for as part  of this? So how do we get to that friction-free? [8:10] How do we make it easy and seamless in the speak  of artificial intelligence natural language? How [8:18] do we get that in a way that is easy for every  age group in the community that we serve to get [8:24] there? Now beyond just having that access to the  door, we're looking beyond in terms of think about [8:35] somebody having a high-risk surgery. And I'll  tell a little bit more about some of the work [8:40] we're doing there. But how does that navigation or  care navigation through the scenarios go through? [8:49] And once somebody's going through a care  opportunities like virtual rounding, [8:57] informing the patient's family in the  most intuitive way when a clinician [9:03] is about to come and can they put a weight on it. [9:09] Having intelligence beyond what a  human can provide, overseeing it, [9:15] looking at a fall detection, a trip, a  fall, a slip that allows us to ensure [9:23] care is provided before the trip happens.  The list goes on. But another important, [9:33] but all of this is the financial assistance.  A lot that goes this day, this age. [9:40] Ensuring that the patients are having the  right way to help answer their questions [9:46] about a billing, a payment, financial  assistance, coverage, what have you. [9:56] Now, you know, the the other piece we look at  that is a lot of us perhaps can attest to this [10:05] When a particular condition arises, [10:11] one of the first things we do  normally is pull up a phone, [10:15] become Mr. or Miss Google or Miss Chad  GBD, Mr. Chad GBD. You get the idea. [10:21] We want to know what's going on. We  want to know what else is out there. [10:27] The point I'm trying to make is that a condition  a surge begins outside the system to begin with [10:35] and then not even giving a fat chance for  a healthcare to become a part of it early [10:40] on. So what we are doing is bringing that gap,  bridging that gap, bringing that evidence inside, [10:47] allowing a contextrich opportunity for our  patients to be able to safely ask away the [10:57] questions they have, which also leads into beyond  scheduling, navigation, pharmacy, what have [11:07] you. things that is holistic in nature giving  them an answer to get to a care they deserve. [11:18] I'll give you a little bit more  context around the surgical experience. [11:26] There are certain kinds of surgeries that many  of us in the room are aware of which are high [11:30] risk and expect to have a large number of you know  emphasis around care. Think of colorectal, think [11:38] of hippony. One of the recent work we have we  have done. What we came to find out prior to that [11:47] is when a particular patient goes through one of  these high-risisk surgeries, they have questions [11:56] days before the surgery is going to be performed.  Think of think of a 60 62 year old female having [12:08] to you know been uh elected by by themselves to  have a hip and knee surgery. Let's say the surgery [12:15] is about 15 days away. A lot of questions come  about. They want to know how they prepare couple [12:25] weeks ahead their conditions. How is that going  to suit in in readiness for that? What things [12:34] they can eat? What things they cannot? How do they  prepare their body in in in to get ready for the [12:39] surgery? A lot of questions. Now traditionally  these questions were posed to the clinician [12:52] they are part of the clinician team nurse team  they're part of which is perfectly fine which [12:57] is which is really where the care should come but  we all know there are a lot of questions before [13:04] the surgery the day of the days leading after  the surgery how should I prepare my my healing [13:14] What should I be wrapping it? What should I  not? What should I look for? What should I not [13:18] look for? When do I call? For every one of these  questions, and a lot of times these questions come [13:24] at late hours in the day when the clinician and  the clinician team may not be readily available. [13:35] We prepared and and offered up a 24/7 AI agent  that now has an equal knowledge contextually [13:46] speaking. Now obviously there is a human in the  loop. We are very wary of it that allows us to [13:50] make sure that the response is curated. But on the  same note, for the patient community, this is an [13:59] incredible opportunity for them to feel they're  always connected with the clinician team 24/7. [14:12] We talked about the patient experience. [14:15] I'll switch gears to the second pillar  which is all about our clinician care. [14:24] For the communities we serve, our clinicians  are the most priced positions we have to [14:31] serve our community. But when you look at  virtually several clinicians out there, [14:41] nurses out there, their lives over  the years has come to be inundated [14:48] with tasks that keep piling on and on,  putting them behind otherwise what they [14:58] believe they need to work on top of the  license. That's with any health system. [15:05] So what we have been you know tirelessly working  is to come up with innovative and transformative [15:14] ways leveraging artificial intelligence leveraging  technology leveraging more importantly reimagining [15:22] the business of a clinician the profession  of a clinician working alongside with them [15:29] working side along with our patient voices to  help to see what changes things. For example, [15:37] let me let me walk you through a story where one  of our orthopedic surgeon, Dr. Valentine, he walks [15:44] in on a given day, five to six appointments  during the day before even the lunch hits. [15:52] Every patient he's going to walk into, he needs at  least 10, 15 minutes sometimes to prepare notes, [16:00] to have a reading into the insights, what  they're going into, recent medical changes, [16:08] choices of pharmacy picked, medication picked,  lab results, you get the idea. Those take time [16:14] for for the clinician to digest. These are  discrete pieces of information. Even within [16:23] the EMR discrete piece of information floating  everywhere and all has to be gathered, assembled, [16:28] processed and then he walks into the door.  Obviously, when he's at the door with the patient, [16:37] the clinician puts every effort to ensure they are  at most listening and caring for the for the for [16:45] the patient. But on the same note, they're also  thinking how am I going to now complete my notes, [16:53] complete the charts, move on to the next. The good  news, what we did was listening through all of [17:01] that, we have a variety of AIEL products that are  now at play to ease the administrative burden of [17:12] the clinician. that includes ambient listening  with the permission of the patient automated [17:18] chart summary. We also have inbasket management  for our clinicians where draft automated draft [17:28] um of their inboxes take care so that they can  get as we cliched words of pajama time. Now we're [17:38] also looking not just at the clinicians  we're looking at our nurse population. [17:42] How do we improve the handoff? How do we  reduce the burden that the nurses have to [17:47] schedule their time, the nurse managers?  How do we look at things that are beyond [17:54] just a clinician? And now we look at  pharmacy. How do we help to connect [17:59] the dots between the end-to-end holistic  cycle of a a clinician and a pharmacist? [18:09] A use case I want to run by is all about  telemetry. As some of us know here, telemetry is [18:21] one of those things which helps us to as a  clinician to manage to understand the heartbeat [18:27] and all the vitals that a patient has stuck around  your body. It it is also a bit of a an hassle for [18:38] both the patient and the clinician. As important  vital signs, the telemetry emits it creates or [18:47] has the ability to create alarm fatigue. At some  point the nurses and the and the and the rounds [18:54] clinicians do face that. So how do we minimize  that? But from a patient end they don't want to be [19:00] stuck with this telemetry for more than what they  should be on it. The only way it used to be that [19:07] was monitored is that the the impatient services  the community was able to watch for it and when [19:15] the needs were in there were taken out. Not too  long ago, we worked on a predictive set of models [19:24] through artificial intelligence that connected  with the telemetry, studied and gave us a much [19:29] better way to inform when the telemetry services  do not need to be used on the patient anymore. [19:38] And that patient loved it. The clinicians loved  it. No less alarm fatigue. Equally important was [19:49] when new patients perhaps somebody in an ER  ready to move up to the inpatient ward they [19:56] now have more access to the telemetry devices  than before that's a simple less fancy less [20:05] glamorous practical way of how we continue  to leverage AI in the business of healthcare [20:18] Let me move on to the high-risisk  disease. So we talked about diseases, [20:24] conditions when patients walk in there knowingly.  But then there is quite a few high-risk diseases [20:37] that the patients themselves don't know of  which makes it harder. One such thing is lung [20:44] cancer. Based on the information that we have  from our patient consented EMR or predictive [20:51] models we put in place, we were able to gather  up and and size up the population of roughly [20:56] about 20,000 population in our communities  who potentially have the risk of having [21:04] u a stage of lung cancer. Which means the goal  is to outreach these population informing them [21:15] to schedule a visit to have it checked out and  should care be needed they are able to access [21:23] the care before it's too late. 20,000 patients  there's only so many phone calls we could make [21:31] in time. Fortunately, we equally worked  on the side with a conversational AI. Our [21:38] conversational AI agent has been informing  the patients 24/7 voice and text outreach, [21:47] giving them a much needed information through  the consent of the clinicians, an opportunity for [21:53] them to schedule and have any other questions that  can be answered. We didn't stop with lung cancer. [22:00] We are going after additional other high-risisk  conditions. Think of colon, think of breast [22:06] cancer. Think of other conditions that most  important is time. And how do we leverage [22:13] AI to cut the time and the administrative  handling that otherwise would be needed. [22:23] I'm going to flip through the next couple of  slides uh in interest of time. But this particular [22:28] one similar to the high-risisk disease diabetic  retinopathy for certain conditions that somebody [22:38] some know of when you have a diabetes having your  eyes checked for patients in that condition is not [22:47] always an option. particularly think about rural  areas, particularly think of locations where you [22:53] may not have that facility. So, one of the things  we did was we partnered and we're able to bring a [23:02] simple solution AIEL physical device that can and  has been led in even at the hands of their primary [23:11] care physician who able to have this conducted  and refer them to a specialist. if needed. [23:26] Let me let me flip to the uh the third  important pillar, operations excellence. [23:34] So I'll give you a scenario, right? [23:39] Colonoscopy screening. Virtually every clinic  in the country encourages the population to [23:49] be screened after you attain a certain  age and regularly screen depending on [23:54] your situation. As you can imagine, it is  one of those preventative screening which [24:06] requires a certain amount of preparation on  the patient side, on the population side. But [24:12] when you flip the narrative to a provider,  to a hospital, it's a 4-hour block. You need [24:20] an anesthesiologist. You need the preparing  clinician. You need the nurse clinician. You [24:26] need the equipment. You need the supplies. You  got to be able to block the time to get there. [24:35] When we looked at it, the national average  roughly about 8.4 in 4% same day cancellations. [24:46] 1 in 12 cancels same day. And think about that  cancellation. What happens here is once that [24:55] cancellation happens now the block cannot be  removed. You're not able to easily use that [25:03] slot for another another person who otherwise was  wanting to get the time. the anesthesiologist, [25:12] the nurse and the supplies are now have to be  you know waited. So what we have done is now we [25:19] are leveraged AI to understand and and and see  why these cancellations happen influence help [25:26] when necessary also allocate valuable resources  the assets that includes the clinicians and the [25:35] supplies for the right place that is absolutely  helps our community and ourselves to be able to [25:42] to serve the community better. That's operation  excellence for us. Now, we didn't stop there. [25:50] We're looking across we're going across supply  chain. How do we better influence our contact [25:55] centers? How do we better influence our revenue  cycle supply chain and the hospital operations? [26:04] Now, I'll take it home with the narrative.  Imagine you have a 50 52 year old patient who [26:14] has scheduled for her orthopedic hip replacement  in few days. But on the but just literally the day [26:26] before the scheduled procedure or preparation  procedure 8:00 post dinner something happened [26:36] some situation she now wants to change  the the the lab work before the surgery. [26:45] In the old days when the when the patient calls  there's obviously somebody to take the receive the [26:52] call but they're not at a position to go make any  modifications. They have to they're asked to call [26:59] the following morning so that the attending  clinician office can help reschedu the lab [27:06] and obviously reschedule rest of the procedures.  What we did, we basically now have an intelligent [27:15] conversational agent 24/7 is able to attend  the call and also help Margaret in this case, [27:26] handle any changes because it's intuitive enough  to know all of that. Now, it doesn't stop there. [27:33] The same Margaret while having to reschedu her lab  work ahead of the surgery has questions about the [27:42] billing. The agent in the prior world, the human  agent would have to need it a different person, [27:50] make a different number, two more rings, eight  more minutes to get the call. No more. Right now, [27:57] the conversation agent is able to answer the  billing question over and above the scheduling. [28:02] It doesn't stop there either. The same patient  in some scenarios, illustratively speaking, also [28:10] has a question about what her clinician mentioned  that look post surgery down the line. My clinician [28:18] wanted to know about getting a physiootherapy. Can  you tell me more about the opportunities there, [28:24] the options there, the cost, my coverage, all of  that? Again, yesterday year it would have needed [28:31] a different person to help answer those questions.  Now the agent can answer all. The point I'm trying [28:38] to make home here is when we look at how we can  leverage AI, we do not start with the question [28:48] of what AI can do. We start with the question of  where are the problems? What where do people get [28:55] stuck? Where do our patients get stuck? What is  the process in Asia? What's causing our clinicians [29:01] to not know get stuck and administratively  burdened? Those are the questions allow us [29:08] to go choose the work that we can put. We focus  on vital few that trivial many. Our choices of [29:17] sol our choices of problem solving are obsessed  over patient to give the remarkable experience. [29:24] That is operational excellence for us. giving us  a chance to be able to serve the community better. [29:34] Now, for the informed us in the room here, [29:37] artificial intelligence is extremely important  in terms of the guard rails it works around. [29:45] We have a strong AI trust and safety governance  that works with the right stakeholders internally, [29:53] keeping close tabs with all the right external  constituents, ensuring the models we build, [30:00] the models we procure, the partners we work with,  all of the all of the above has the right safety, [30:09] quality and trusted means. And that's not just  because we are caring about how these models [30:19] operate. More importantly, the health care  space we operate, we want to make sure that [30:27] the trust and the quality lies fair and square  in the interest of patients and the clinicians. [30:36] Now in the all three things I talked about [30:41] patient experience, clinical  care and the operations. [30:48] The common thread for all of  that is how do we leverage [30:55] intelligence and transformation to simply  remove friction to create better decision [31:02] making bring human in the loop. AI for us is not  about replacing humans. AI fors is a mechanism [31:12] that reimagines pulling people together,  collaborate together, make better decisions. [31:18] That thought process continues to fuel us  to be able to solve important problems. [31:25] Thank you. With that, I can take  uh I guess couple of questions.