Eve AI Quintuples Law Firm Caseloads to Unlock Small-Dollar Claims

Eve AI助律所案量激增五倍，激活小额索赔市场 通过将从案源接收到和解的全流程诉讼自动化，消除将小额索赔排除在法院之外的人力瓶颈，使风险代理律所能够扩大接案量，并在竞争中胜过按小时计费的辩护律师。 Jay Madheswaran与Alex Rampell探讨了为何AI在法律领域的主要价值在于为按胜诉提成收费的原告律师拓展收入与承案能力，而非仅仅为按小时计费的律所削减成本。由于高昂的行政管理间接成本，律师此前普遍受限于最低5,000美元的律师费门槛，被迫拒绝超过90%的潜在客户，致使普通索赔人缺乏法律代理。通过将复杂的程序化流程自动化——包括客户案源接收、调取医疗记录、隔夜文件审查以及起诉要求书起草——Eve等平台让单名律师承办的活跃案件量从100起扩展至500起。这一转变借助与结果挂钩的定价模式将律所转型为类软件化运作，消除了大型企业保险软件长期维持的数据不对称，并将律师的精力重新引向法庭辩护。 法律AI平台Eve正在实现原告从案源接收到和解诉讼工作流的自动化，使洛杉矶一家律所每位律师的办案能力提升五倍，从100起案件增至500起。 (0:00) 高昂的行政管理成本目前迫使原告律所拒绝估计高达96%的劳工与就业案源，因其无法达到5,000美元的基准律师费门槛。 (7:15) 法律科技领域的软件定价正脱离传统的按席位付费SaaS模式，转向与风险代理回款保持一致的按结果和算力计费模式。 (40:00) 受制于计费工时的辩护律所正落后于经AI赋能的原告律师，后者利用自动化代理可在数小时内反制程序性拖延策略和证据突袭。 (50:00)

Alex Rampell and Jay Madheswaran argue that specialized AI software can transform small contingency-fee plaintiff law firms into high-revenue businesses by eliminating severe labor bottlenecks rather than merely cutting operating costs. The Economics of Legal Access and Rejection (4:47) Unlike corporate defense firms billing clients upwards of $2,000 an hour, plaintiff practices operate on contingency fees, forcing attorneys to fund administrative labor upfront. Jay Madheswaran explains that law firms typically reject cases expected to generate less than $5,000 in fees because the underlying labor overhead guarantees a financial loss. In practice, this economic threshold prices out ordinary individuals with small-dollar claims, such as minor vehicle accidents below insurance deductibles or complex workplace harassment inquiries, leading firms to reject up to 96% of labor and employment leads after spending hours on uncompensated preliminary research. Scaling Caseloads Through Workflow Automation (13:46) Because contingency practices face near-limitless consumer demand, their revenues are throttled strictly by back-office capacity. Madheswaran notes that deploying end-to-end legal AI allows small practices to expand capacity drastically; for example, one Los Angeles firm grew from managing 100 active cases per attorney to 500 without expanding headcount. Platform agents take on repetitive operational tasks—including client intake calls, preliminary legal research, medical record retrieval via outbound voice agents, and overnight file auditing to flag missed evidentiary details—freeing attorneys to focus on high-value legal insights and trial preparation. Application Moats and Insurance Data Asymmetry (20:00) Madheswaran and Rampell assert that general foundation models alone cannot solve plaintiff litigation because less than 1% of filed cases reach public trial verdicts, leaving the critical pre-demand settlement benchmarks absent from web scrapers. While institutional insurers use nationwide valuation software like Colossus to drive payouts to statistical minimums, localized plaintiff firms have historically operated blind. By encoding collective, non-confidential practice strategies and analyzing private settlement patterns without training models directly on privileged case files, specialized legal platforms balance this informational asymmetry against corporate carriers. Disrupting Defense Tactics and Legacy Pricing Models (50:00) The structural contrast between contingency-fee plaintiffs and hourly defense counsel is reshaping litigation tactics. Defense firms, disincentivized from working quickly by their hourly billing model, frequently employ delay tactics, such as dumping discovery requests immediately before major holidays. Plaintiff firms running automated intake agents have neutralized these strategies by returning comprehensive discovery responses the same day. As repetitive administrative interactions shift toward agent-to-agent negotiation, software vendors are abandoning traditional per-seat SaaS tiers in favor of outcome-based pricing that mirrors contingency fee recoveries. While AI can execute the overwhelming majority of procedural and drafting steps, Rampell and Madheswaran acknowledge that human judgment remains essential for nuanced courtroom advocacy. Furthermore, as the operational friction of filing claims drops toward zero, institutional constraints will likely shift to downstream choke points like court docket backlogs, potentially forcing future disputes into automated dispute resolution systems.

Alex Rampell 与 Jay Madheswaran 认为，法律人工智能在原告律所的核心价值并非削减员工以降低成本，而是通过自动化端到端诉讼流程打破人力瓶颈，使律所能够成倍扩张案件承接量并显著增加营收。 原告法律的经济瓶颈与案件筛选门槛 (4:47) 传统大型企业律所采用按小时计费模式，而原告律所普遍采用风险代理机制，即律所前期垫付所有办案人力成本，仅在胜诉或达成和解时按比例抽成。在此模式下，办案质量与投入的法律劳动量高度挂钩，但普通个人客户无力开具无上限的付费支票。律所运营面临严峻的经济约束：处理一件普通车祸或劳动纠纷，若潜在诉讼标的仅有数千美元，扣除法务助理、初级律师等高昂人工成本后，律所往往无法获利甚至面临亏损。Madheswaran 指出，当前如果一起案件预期无法为律所带来至少 5,000 美元的律师费收入，律所基本不会接案。Alex Rampell 也提到，由于无法以极低成本雇佣办案人员，标的额仅 1,000 美元的索赔者往往完全无法获得代理服务。例如在劳动与雇佣法领域，律师通常需投入 10 小时进行法律调研、事实核查与访谈，高昂的前置筛选成本导致高达 96% 的劳动纠纷咨询在初审阶段即被直接拒绝，大量普通民众的合法权益因此被排斥在民事司法体系之外。 端到端自动化与营收扩张逻辑 (10:00) 许多观察者误以为 AI 在法律行业的应用主要用于裁员以压降运营支出，但 Rampell 与 Madheswaran 指出，原告律所面临的并非市场需求不足，而是极其严重的人力产能受限。原告律所常通过公路广告吸引大量案源，却因后勤与处理能力有限而 routine 拒掉约 90% 的线索。Eve 等垂直 AI 平台通过将初步案件筛选、事实调查、起诉状撰写乃至调解简报生成整合为自动化工作流，有效将软件转化为庞大的“影子劳动力”，让律所不再受制于传统聘用规模。在实际案例中，洛杉矶一家原告律所在没有大规模扩招的情况下，通过接入 Eve 使每位律师管理的活跃客户容量从大约 100 人大幅攀升至 500 人，实现了 5 倍的案件吞吐量增长。软件从根本上成为律所扩大客户规模、增加整体营收的杠杆，而非单纯的节流工具。 数据壁垒与对抗保险公司的信息不对称 (20:00) 在人身伤害和事故索赔中，诉讼各方存在极大的信息不对称。保险巨头（如 Geico、State Farm）通过集中全国海量理赔数据，并利用 Colossus 等算法系统精确计算最低赔付额度，采取拖延与打压策略；相反，原告律所高度分散于各州各县，缺乏跨区域的判例与和解数据。由于 99% 以上的案件在庭审前达成私下和解，最具参考价值的真实赔付基准完全无法从公开网络获取。Madheswaran 强调，单纯的基础通用大语言模型无法解决该问题，犹如将爱因斯坦直接空降至律所也无法立刻产出效益。Eve 在严格遵守律师-客户保密特权、不利用客户私密数据训练核心模型的前提下，整合了行业办案策略与继续法律教育知识体系，并开发了 Auditor 工具在夜间自动审查全部卷宗以发现事实矛盾与遗漏。通过量化案件预期价值，原告律师得以精准判断何时接受保险公司提出的和解要约，何时坚决推进诉讼。 代理人对抗与长流程智能体演进 常规诉讼通常包含长达数月乃至数年、涉及数千至上万次交互的复杂流程。Madheswaran 观察到，Eve 正在通过自动化语音外呼智能体自主与医疗机构、保险公司通话以调取病历，甚至出现了 AI 智能体与对方自动化语音系统互相博弈与协商的场景。在程序对抗层面，被告辩护律所因长期依赖按小时计费模式，缺乏采用效率工具的主动动力，甚至常利用假期前夕集中提交证据开示请求以压垮原告团队。而原告方借助自动化智能体，能够在数小时内甚至当天完成全面答复，彻底瓦解了传统的程序拖延策略。Madheswaran 预测，未来大量常规的索赔要约与谈判拉锯将由“AI 代理人对抗 AI 代理人”自主完成，而人类律师的工作将更加聚焦于深度的本地化情境洞察与真正的法庭辩护，AI 驱动的替代性争议解决机制也将承接大量高频争议，缓解法院系统的积压难题。 律所软件化与成果计费模式重塑 (35:00) 随着垂直大模型技术的渗透，传统原告律所正在经历类似软件公司的模式转型。在传统的偶发胜诉分层下，律所获得赔偿额约三分之一作为分成，但其中约 50% 被法务助理及行政人力开支消耗，另有约 30% 用于市场获客，留存利润极薄。当 AI 取代了大量的机械性行政劳动后，律所得以释放利润空间，并将其重新投资于算力与 Token 消费以处理更大规模的业务。与此同时，法律科技的商业模式也正在脱离传统的按席位订阅软件模式，向直接挂钩案件处理结果的成果定价转变。Madheswaran 预测，尽管复杂巨额诉讼仍可能保留按小时计费，但标准化法律事务将逐步转向基于算力成本的确定性结果定价；大型原告律所也在借助标准化的 AI 工作流基础设施，更顺畅地推进对中小型律所的整合与滚打包购。 对话指出，原告法律服务的核心痛点在于诉讼流程极度繁复，单个案件涉及多达上万次操作交互，这为长流程智能体提供了巨大落地空间。不过，Eve 的技术演进仍受限于繁重的复杂工作流逆向工程研发，尚未完全实现全流程 100% 无人干预；同时，随着索赔门槛与立案成本断崖式下跌，案件洪流也将给法院审判排期与监管机构带来下半程的结构性拥堵瓶颈。

## Background 本期内容由 Alex Rampell 与 Jay Madheswaran 于 2026 年 10 月 30 日通过 RSS / web 播客平台发布。节目简介指出，Eve 首席执行官 Jay Madheswaran 与 a16z 普通合伙人 Alex Rampell 共同探讨了 AI 如何通过大幅降低诉讼成本，将原本在经济上不可行的小额案件转化为有价值的业务，并详细解析了 Eve 如何实现原告法律业务从接案到和解的全流程自动化，进而推动中小型律所跨越式发展。 ## TL;DR 节目深入剖析了传统原告律所面临的人力供给瓶颈与高达 96% 的案件抛弃率，指出 AI 的核心价值在于扩大接案规模而非单纯裁员。通过具体案例展示，Eve 等长远景智能体（long-horizon agents）将单名律师可承接的案件负荷从 100 起提升至 500 起，并推动法律行业从传统的按工时/席位计费向基于成果的定价模式转型。对话进一步探讨了智能体之间自主攻防、破解大型辩护律所消耗战术的行业新图景，展望了软件化律所利用垂直数据壁垒成长为十亿美元级企业的可能，以及司法纠纷向自动化中立调解（ADR）分流的未来走向。 ## Atmosphere & style 对话洋溢着硅谷风险投资人与前沿创业者之间思维碰撞的敏锐与兴奋感，既有扎实的微观经济学逻辑推演，又不乏生动幽默的实战轶事。两位对谈者语调轻快而笃定，从咖啡馆偶遇诉讼律师聊到万次沟通触点中的“AI 对轰”奇景，充满了以技术颠覆传统垄断的“大卫对抗歌利亚”式乐观叙事。整场讨论兼具宏大行业视野与细微执行洞察，让听众清晰感受到法律技术变革前夜那种既务实又令人振奋的真实张力。 ## Bulletin - `05:00` 传统原告律所受限于人力成本，通常需要每起案件至少有 5000 美元的预期分成才愿接手，导致劳动法等领域高达 96% 的当事人诉求在初筛阶段被拒。 - `00:00` 洛杉矶某律所借助法律 AI 平台 Eve，将单名律师平均管理的案件数量从约 100 起大幅提升至 500 起，效率实现 5 倍飞跃。 - `20:00` 超过 99% 的原告案件在立案或索赔前私下和解，缺乏公开判例数据，通用基础大模型若脱离垂类业务语境和隐蔽数据将难以胜任实战诉讼。 - `40:00` 法律软件与律所商业模式正从按席位/按小时计费向成果导向（Outcome-oriented）演进，律所将节省的人力成本转化为算力采购以实现收入扩张。 ## Key takeaways - `10:00` AI 在原告律所的核心价值是驱动收入扩张（接纳原本无法盈利的海量小额案件），而非单纯裁员降本。 - `05:00` 微观经济供给瓶颈导致法律市场失灵：打赢仍亏损的小额争议被拒之门外，劳动法领域拒绝率高达 96%。 - `10:00` 原告律所本质是采用风险代理的小微企业，制约其发展的瓶颈不是广告获客，而是内部劳动力的供给不足。 - `15:00` 软件市场实际上是庞大的影子劳动力市场，AI 正在直接充当虚拟劳动力以释放数千亿美元的潜在经济产出。 - `20:00` 行业真实博弈数据高度隐蔽：不到 1% 的案件形成公开裁判文书，绝大多数诉前和解细节无法被通用网络爬虫获取。 - `22:38` 保险公司长期利用 Colossus 等算法压低赔付标准，而 Eve 正在通过多维度数据辅助原告在每一处关键节点作出最优博弈决策。 - `25:00` 产品 Auditor 可在夜间逐页自动审查长达数千页的诉讼案卷，消除传统人工办案中普遍存在的事实遗漏隐患。 - `27:15` 长程智能体正在接管长达 6 至 12 个月、涉及上万次沟通的琐碎流程，人类律师得以回归真正的法庭辩护与人文关怀。 - `46:17` 常规交涉正演变为智能体与智能体（Agent-to-Agent）的对战，例如语音 AI 外呼智能体互相试探并绕过对方的规则套路。 - `32:00` 司法资源有限将促使纠纷向非诉讼机制分流，具备中立视角的 AI 调解智能体有望大幅加速事实厘清与公正裁决。 - `40:00` 传统大所依赖的按工时计费与按席位订购正走向退潮，按确定性算力交付法律结果的成果定价模式正在确立。 - `50:00` 商业模式错配导致攻守易势：被告方按工时计费的拖延与“假期突袭”战术，被原告方 24 小时全自动响应的 AI Agent 瞬间瓦解。 ## Detailed walkthrough ### `00:00` 原告法律业务的痛点与 Eve 的定位 传统观点常常认为 AI 的主要作用就是裁员和削减开支，但对于律所而言，借助 AI 实际上能够大幅拓展创收空间。历史上，原告律师事务所（plaintiff law firms）的运转几乎只需要一部手机和一个电子邮箱，市面上从未有真正的软件能够解决它们的核心痛点——劳动力瓶颈（labor constraint）。普通人不可能像大企业面对白鞋所（white-shoe law firms）那样，以每小时 500 到 600 美元乃至 2000 美元的价格开出无上限的空白支票。因此，原告律所普遍采用风险代理（contingency fee）模式，即只有帮当事人胜诉并拿到赔偿后才收取分成。 Eve 正在通过 AI 彻底改造这一流程，目标是实现从客户首次来电（intake）到最终达成和解（settlement）的端到端自动化。这种提效已经在现实中显现成效：洛杉矶的一家律所利用 Eve，使得单名律师能够管理的案件数量从约 100 起跃升到了 500 起。 谈到为什么通用的通用人工智能无法直接解决律所问题时，Jay Madheswaran 打了一个形象的比方：即便把爱因斯坦放到律所里，他短时间内也无法产生实际价值，因为法律业务高度依赖特定的专业流程、事实挖掘与律所经验。正如 Jay Madheswaran 所指出的： > “What Eve is aiming to do is really handle that whole thing end to end of how do you represent a client all the way from the very first call to getting them a settlement purely with AI and give that to the lawyers to then amplify the outcomes that can actually provide. And think about giving David a slingshot to take down Goliath. That's what we're doing.” - 传统原告律所的核心制约是律师人力工时，而非缺乏技术工具。 - 风险代理模式为普通大众与强势机构抗衡提供了最初的平权工具，但律师的高成本仍让门槛居高不下。 - 洛杉矶律所借助 Eve 将单律师管理负荷从 100 起提升至 500 起，效率提升高达 5 倍。 --- ### `05:00` 经济学供给曲线失灵与 96% 的案件抛弃率 从微观经济学视角来看，当赔偿金额较低（例如只有 1000 美元的损失）时，市场就会出现严重的“市场失灵”（market failure）。即使是风险代理模式，由于律师必须支付视案情而定的平行工作量、辅助人员工资以及固定研究成本，律所如果无法从单起案件中获得至少 5000 美元的律师费抽成，往往根本不会接案。哪怕是损失了 3000 到 4000 美元免赔额的车祸摩擦，对普通打工者而言可能意味着整整一个月的生计，但在美国没有任何律师愿意介入，因为接案本身不仅不赚钱，甚至“即便打赢了也会亏本”。 在劳动与雇佣法（labor and employment）领域，这种困境尤为尖锐。美国大多数州实行随意雇佣制（at-will employment），当一名普通雇员（例如餐饮住宿服务业员工）在职场遭遇骚扰致电律所时，律师需要花 30 到 45 分钟接听电话筛选信息，随后还要花费大约 10 个小时去检索判例法、核实事实背景，看能否将问题归入法定保护条款并匹配诉由（causes of action）。一番周折下来，律师往往发现预期的提成根本达不到 5000 美元的盈亏平衡线。正因如此，劳动法领域有高达 96% 的当事人诉求在咨询阶段就被律所直接拒之门外。 Jay Madheswaran 深刻指出了原告律师在经济账面前的无奈现实： > “Today, if a lawyer doesn't make at least $5,000 fees off a case, they do not take it on. That is the market dynamics today. And even at the $5,000, they're losing money on a lot of cases. They're really banking on a few cases paying out very large.” Eve 的介入直接打破了这一成本瓶颈： 1. **自动接待与初步评估**：来电打入时由 Eve 接听，结合律所历史洞察与专业判断标准，即时识别案件是否具备法律诉由与受理价值。 2. **免去 10 小时人工预研**：原本需要律师耗费 10 小时的初步法律检索与问讯，变成即刻完成。 3. **快速启动索赔**：系统可直接自动起草索赔函（demand letter）或完成前置立案准备，使经济可行性的门槛大幅下移。 4. **长周期行动闭环**：一起典型案件通常持续 6 个月到 1 年，其间涉及上千次来回互动的细碎步骤，随着 AI 准确率不断提升，Eve 正在逐步接管这一长周期链条中的复杂工作。 ```mermaid flowchart TD accTitle: 传统原告律所案件筛选与 Eve AI 自动化流程对比 accDescr: 展示了传统人工筛选导致96%案件被拒与Eve实现即时评估并推进长流程的对比 subgraph Traditional["传统人工筛选路径（劳动密集）"] A["客户来电求助"] --> B["律师接听咨询（耗时 30-45 分钟）"] B --> C["判例检索与事实确认（耗时约 10 小时）"] C --> D{"预期律师费是否达到 5,000 美元？"} D -- "否 (占 96%)" --> E["拒案（普通人丧失维权机会）"] D -- "是 (占 4%)" --> F["正式接案，起草索赔文书"] end subgraph EveFlow["Eve 驱动的 AI 路径"] A2["客户来电求助"] --> B2["Eve 实时接入并理解律所偏好"] B2 --> C2["即刻匹配诉由并完成初步研判"] C2 --> D2["自动起草索赔函 / 前置准备"] D2 --> E2["承接 6-12 个月内上千步端到端流程"] end ``` ### `10:00` 咖啡馆的偶遇与原告律所的真实商业模式 大部分人对 AI 的刻板印象往往是“降本增效”、“裁员节约开支”。Alex Rampell 分享了一个发生在大约九个月前的真实趣事：他在加州洛斯阿尔托斯（Los Altos）的一家 Blue Bottle 咖啡馆向人解释 Eve 的逻辑——使用 Eve 并不是为了裁员削减成本，而是让律所能够接下更多案件，实现“收入扩张”（Revenue Expansion）。正聊着，一个牵着大狗、神色看起来有些严肃甚至吓人的男人朝他走来。Alex 本身有点害怕大狗，心里正犯嘀咕对方是不是被激怒了，结果对方开口问：“你们刚才聊的那家公司叫什么？我是个出庭诉讼律师（Trial Attorney），我想试用一下。”Alex 笑着表示，自己当场就给 Jay Madheswaran 带来了一位新客户。 无论是加州的 101 高速还是贯穿美国东海岸（从佛罗里达一直延伸到缅因州）的 I-95 州际公路上，沿途随处可见各类原告律所投放的巨幅广告牌，疯狂招揽车祸和人身伤害案件。表面上看，律所愿意砸重金做广告吸引潜在案源；但现实中，由于人力成本高昂，律所不得不拒绝掉大约 90% 打进来的咨询电话。Jay 解释道，原告律所本质上属于小型企业（Small Businesses），创始人和律师读的是法学院而非商学院，其宗旨是为普通人提供代理服务。律所的赚钱逻辑极度直接：“通常来说，你代理的人越多，你赚的钱就越多。”然而，接手一个案子的服务成本极高，限制律所发展的根本瓶颈不是案源需求，而是律所内部的劳动力供给。 ```mermaid flowchart TD accTitle: 原告律所业务瓶颈演进与 Eve 的整合 accDescr: 展示律所从单人作坊式运转到建立呼叫中心遭遇瓶颈，最终由 Eve 平台整合接管的过程 subgraph Phase1["初级阶段：单人律所"] A["户外广告引入案源"] --> B["个案经理或前台记录电话"] B --> C["律师业余时间疲于回访，杂务缠身"] end subgraph Phase2["规模化阶段：传统呼叫中心"] D["外包呼叫中心 / 独立接案团队"] --> E["广告获客爆发，团队迅速过载"] E --> F["受制于人力瓶颈，仍被迫拒掉海量案源"] end subgraph Phase3["Eve 智能化整合"] G["Eve 自动化接管：接听、初筛、流转"] --> H["抬高案件处理底线 (Raise the Floor)"] H --> I["律师仅在核心专业判断与客户交付时介入"] end Phase1 -.->|"业务扩张"| Phase2 Phase2 -.->|"技术升级"| Phase3 ``` Jay 指出，律所的瓶颈在不断漂移。微型律所通常只有一名律师加一个案件经理或办公室主任，电话接进来后草草记下，律师再在忙乱的空档期回复；当律所尝试规范化、在某地搭建呼叫中心（Intake Team）后，一旦营销广告效果好、电话被打爆，呼叫中心又迅速沦为新的瓶颈，律所依然受限于人力规模。针对这个痛点，Jay 总结道：“我们基本上将所有这些流程整合进了 Eve 平台，从而让他们只在真正需要的时候才被调动进来。”Eve 不断抬高业务的基线（Raise the floor），让律师把精力聚焦在真正能改变客户诉讼结果的高价值专业判断上。 - 普遍认知误区认为 AI 主要是裁员工具，但对原告律所而言，AI 核心价值在于扩大接案规模，拉动收入增长。 - 原告律所无法接纳所有案源的根本原因在于劳动力市场失灵，律所不可能以极低廉的成本雇佣到足以处理每通电话的副手（Paralegal）。 - 很多律所团队读的是法学院而非商学院，缺乏复杂的运营系统，导致业务瓶颈在接待、初筛和跟进环节反复出现。 ### `15:00` 从 100 到 500：AI 驱动的劳动市场替代与软件重构 引入 AI 代理后，律所的人效迎来了飞跃。Jay 透露，洛杉矶（LA）的一家合作律所通过最大限度地利用 Eve，成功将“每位律师管理的客户数量从大约 100 人提升至 500 人”。该律所将自动化铺满全流程：从最初接听电话、资格初筛，到彻底消除不同律师和案件经理之间反复交接沟通的“信息接力问题”（Relay problem），所有环节均交由 Eve 的 AI 智能体打通。甚至在走向调解之前，Eve 能全自动生成调解摘要（Briefs），明确指出当事人应当关注的核心重点、基于以往类似案由推断出的预期和解结果，以及过往成功案件所采用的具体谈判策略。正如 Jay 所言：“我认为这就是法律的未来，身为一名律师，你未来的工作形态将更像是一名数据科学家和工程师，不断对系统进行工程化调优，让它变得越来越出色。” Alex 进而分析了法律软件市场长久以来的反常现象。以前风险投资机构（VC）普遍看衰法律软件市场规模，认为行业不过几十亿美元的软件支出——大所（如 Wilson Sonsini）的核心软件是按小时计费的工时追踪系统（Time Tracker），而风险代理律师的核心工具几乎就是 Microsoft Word。投资人误以为这个赛道天花板极低，却忽视了一个庞大的宏观事实：虽然软件采购开支看似少得可怜，但法律行业撬动交付的最终经济产出高达数千亿美元。律所此前无法交付更多产出，纯粹是受困于人工供给，而现实中并不存在“12 美分一小时”的廉价劳动力。 > “事实证明，软件市场实际上就像是一个庞大的影子劳动力市场，它的规模远超任何人的预期。” —— Alex Rampell Jay 进一步剖析了为何原告法务（Plaintiff Law）长期被主流企业级软件遗忘。在 Eve 出现之前，原告律师开门营业几乎只需要一部手机和一个邮箱地址，甚至不少独立执业律师连 Microsoft Word 都不愿购买。因为原告律师从事的是强对抗性业务（Trust boundary situation），双方立场完全敌对，根本不存在像审查商业合同那样的双向协同和批注修订（Redlining）。长久以来，Harvey 等绝大多数法律软件公司都扎堆服务大型辩护律所（Defense side），因为辩方客户预算充裕、付费爽快。相比之下，原告律所全是自负盈亏的小企业，每个月都在紧盯收支账本；如果一套软件不能从第一天起就为他们带来实打实的案件结果，“他们就不会续费，甚至会指着鼻子骂你，想方设法找你麻烦”。但也正因如此，原告律师的核心诉求极其纯粹——他们渴望代表更多客户，他们渴望打赢官司，而现代 AI 软件正是打破其劳动力供给瓶颈的关键武器。 - **人效跃迁**：洛杉矶标杆律所借助 Eve 将单名律师平均跟进的客户容量从 100 人直接推高至 500 人。 - **全链路代理渗透**：AI 覆盖了从最初线索筛选、内部案件流转、调解摘要起草到和解方案推演的全部工作流。 - **软件即劳动**：AI 软件不再仅仅是记录数据的工具，而是直接充当虚拟劳动力，将原告法务长期受制的劳动力供给短缺彻底释放。 - **原告与辩方软件差异**：辩护端软件以协同、合规和企业预算为驱动；原告端软件则完全受投资回报率（ROI）、成案率和和解回报驱动，容错率极低。 ### `20:00` 垂类壁垒与隐秘数据：为什么空降爱因斯坦也做不好诉讼 过往律所试图通过运营人员（People Operations）来标准化案件流程，但往往以失败告终，因为每个诉讼案件的细节都千差万别，业务始终受制于高素质人才的招聘瓶颈。如今借助 AI 对非结构化数据的处理和转化能力，这些曾经高度依赖人工经验的任务终于能够被沉淀并固化为标准化流程，让单个优秀律师有能力轻松管理 500 起案情各异的复杂案件。针对当前应用层初创公司普遍担忧会被底层大模型实验室（如 OpenAI、Anthropic）降维打击的生存危机，Alex Rampell 指出，纯粹的被动防御毫无意义，真正的护城河在于网络效应以及外界无法触达的独家数据资产。 谈及 Eve 相比通用大模型的原生优势时，Jay Madheswaran 形象地打了个比方：即便通用模型聪明如爱因斯坦，直接扔进律所也是寸步难行。Jay Madheswaran 直言：“The problem is if you drop Albert Einstein into a law firm today, he will not be productive for quite some time.”（问题在于，就算你今天把阿尔伯特·爱因斯坦直接扔进一家律所，他在相当长一段时间内也无法产生任何生产力。）造成这一困境的核心根源，在于原告法律业务（Plaintiff Law）的绝大多数核心数据在公开网络上完全无法被爬取：真实世界中只有不到 1% 的案件会最终走向法院判决并形成公开判例汇编，绝大多数案件在发出索赔函（Pre-demand）甚至正式立案诉讼（Pre-litigation）之前就已经达成了和解。 ```mermaid flowchart TD accTitle: 原告法律业务数据流与落地痛点 accDescr: 展示绝大多数案件在诉前和解导致数据无法公开爬取，以及通用智能模型在专业语境中水土不服的机制。 subgraph PrivateData["未公开诉讼过程（>99% 案件）"] A["发生事故 / 权益受损"] --> B["诉前协商 / 否认阶段"] B --> C["发出索赔函 (Demand Letter)"] C --> D["庭前达成私下和解"] end subgraph PublicData["公开数据源（<1% 案件）"] D -.->|"极少数未解决"| E["正式开庭审判"] E --> F["公开法院判决书"] end G["通用大模型 / 实验室"] -.->|"无法爬取私下和解与谈判细节"| PrivateData H["Eve 垂类系统"] -->|"沉淀私下博弈与本地化办案经验"| PrivateData ``` - 原告法律业务的数据冰山：不到 1% 的案件最终会进入法院判决并产生公开案例法，99% 以上的关键理赔博弈数据深埋在私下谈判与和解协议中。 - 通用智能的局限：仅依靠公开网页训练的基础大模型，缺乏处理诉前和解、保密谈判所需的关键行业语境与战术经验。 --- ### `22:38` 保险公司的算盘与信息不对称：从 Colossus 到本地化索赔信 在常规人身伤害或车祸事故中，受害者与保险公司之间的博弈极度不平衡。以车祸理赔为例，无责任方的受害者即便受了重伤、车辆全损，对方保险公司（如 Geico）的第一反应往往也是无休止的“否认、否认、再否认”，试图以 1 美元的荒谬价格打发受害者。保险公司之所以占据上风，是因为它们在全网集中处理海量案件，甚至早在多年前就引入了名为 Colossus 这类听起来带有威胁意味的算法软件，专门根据诸如“大拇指骨折”等具体伤情计算最低给付价格。此类软件的本质并非为了满足当事人的实际康复需求，而是计算保险公司在何种底线金额下可以合法地逃避赔付责任。 法律体系存在的根本目的就是纠正这种不公，但现实中原告律所面临的阻碍极其繁复。即便在 Google 上搜索“如何针对软组织损伤撰写一份完美的索赔信”，也根本找不到可用的实战模板，连新入行的律师助理都必须耗费数年摸索。更棘手的是，司法实践具有极强的地域碎片化特征：不同的地方法院系统、地方规程以及特定的对方辩护律师，其行事风格和应对策略完全不同。Jay Madheswaran 指出，Eve 正致力于将这些高度离散、多属性的隐性经验转化为即时决策辅助：“So what Eve is doing is it's turning this historically incredibly difficult to utilize multi-attribute, multidimensional information that is basically what is the best thing I can do at this particular moment to improve the outcome of the case.”（Eve 正在做的事情，就是将这种历史上极难利用的多属性、多维度信息进行转化，在每一个特定节点回答那个核心问题：我此刻究竟该采取什么最优行动来改善案件结果？）这种决策不仅存在于案件立案（Intake）阶段帮助律师筛选最值得接手的客户，而且在每个案件的生命周期中都会重复上演成千上万次。 --- ### `25:00` 数据隔离与集体智慧：Eve Auditor 如何消除案件疏漏 面对美国各州分散的中小型律所格局（例如新泽西州的 4 人律所与亚利桑那州的 7 人律所），Alex Rampell 提出，既然它们面对的共同对手都是 State Farm 或 Geico 这样的全国性保险巨头，律所之间是否存在共享博弈情报的可能？Jay Madheswaran 阐述了 Eve 在数据隔离与合规边界上的清晰立场：在涉及客户隐私和法律强制的“律师-委托人特权”（Attorney-Client Privilege）方面，平台有着极其严格的物理区隔，绝不使用律所的保密案件数据去训练通用大模型。 但与此同时，Eve 敏锐地抓取了律师行业既有的“集体智慧交流”习惯。现实中，律师们通常会定期出差参加继续法律教育（CLE）研讨会，在酒会上互通有无，分享针对特定保险公司辩护套路的破解之策。Eve 将这种线下经验交流机制通过软件进行了系统化放大，不仅内部聘请专业律师追踪全国诉讼胜率，还推出了核心产品“Auditor”。该产品能够在夜间自动对长达数千页的诉讼案卷材料进行逐页深度扫描，敏锐捕捉可能被疏漏的事实细节。Jay Madheswaran 透露：“And we have this product called Auditor, which actually analyzes every single case through every single page, overnight, and looks for situations that a perfect attorney would catch every time.”（我们开发了名为 Auditor 的产品，它能在夜间通读每一个案件的每一页卷宗，筛查出一个完美律师每一次都能抓获的关键隐患与细节。）实践证明，以往几乎每一个未借助 AI 的传统案件中都会存在被人工遗漏的重要信息，而现在这种系统性人为疏忽正在被软件彻底终结。 --- ### `27:15` 对抗性系统的未来：Agent 军备竞赛与人类律师的终极角色 随着原告律师借助 AI 获得了强大的对抗能力，整套法律生态也必然迎来反制。Alex Rampell 展望，面对原告方发起的高效诉求，保险公司未来也必将升级技术手段，双方将演变为高度智能化的双向博弈——从类似电话客服遭遇超低成本网络攻击（DDoS）的拉锯战，演变到双方自动化 Agent 的直接交锋。以往索赔额仅有两千多美元的微小案件，受害者因维权成本过高往往只能打落牙齿和血吞，保险公司得以肆无忌惮地克扣赔偿；而现在，受害者化身为手握 AI 武器的大卫，拥有了挑战巨人的反击力量。 谈及在长程智能体（Long-horizon Agents）主导诉讼、律师甚至逐渐转向类似“法律数据科学家”的未来图景时，双方探讨了人类与机器的职责分工。在文字起草、事实检索、理赔计算等繁重环节几乎全由算法承接的背景下，Jay Madheswaran 认为，人类当事人在陷入危机时永远需要真实的人性代表与情感依托。Jay Madheswaran 总结道：“Whether that representation, like who does the work on the representation, you can assume it's 99% AI, maybe 99.9% AI, there's always going to be that little bit that is the human touch, right? And we do believe that people are going to make a positive difference to the outcomes of the case.”（无论具体代表工作是由谁完成的——哪怕你可以设想 99% 甚至 99.9% 都是 AI 做掉的——总会有那一点点不可替代的人性温度。我们坚信，人类终究会对案件的最终走向产生决定性的积极影响。）最终，AI 负责铺平所有繁杂的技术与信息鸿沟，而真正的律师则专注于提供高阶的人文关怀、战略把控与正义引导。 ### `30:00` 机构博弈的打破与 AI 驱动的庭审辩护回归 在传统的诉讼博弈中，辩护方（即保险公司或大型企业机构）始终占据绝对优势。机构坐拥海量的资金与资源，更掌握着“时间”这一极具杀伤力的武器，通过无休止的流程拖延让背负债务、难以承担诉讼费用的个体原告筋疲力竭。机构长期依靠集中化与标准化的操作压制原告。然而，AI 正在颠覆这种不平衡的对抗关系。随着技术让原告律师敢于真刀真枪地推进诉讼，案件开始具备真正的杀伤力。未来大量日常的诉求往来与拉锯，将转变为智能体对智能体的自主交锋。Jay Madheswaran 描绘了这种未来格局： > “And all the routine back and forth, I think, is going to happen through agents talking to agents.” 在这种智能体自动处理海量案卷与常规交涉的背景下，人类律师的核心价值反而更加突出：深入理解委托人独特的案件细节、把握事实全貌，并洞察所在辖区司法环境以及地方对手律所的特点，以此实现有针对性的法律博弈。过去，原告律所极力避免出庭，因为“开庭审理会压垮整家律所”，所有人必须全天候投入庭审准备、陪审团遴选等繁重事务，律所本质上只是一个“和解协议推动机器”（Settlement motion）。而现在，Eve 让律所能够从事务性泥潭中解脱出来，重新回归真正的法庭辩护（Real advocacy）。 - 机构方凭借资金和时间储备拖垮个体原告的传统手段正在失效，AI 赋予了原告案件更强的抗衡底气（Teeth）。 - 律师的角色从繁琐文书处理转回高价值的实际辩护，关注如何为处境艰难的委托人争取最大利益。 - 借助 AI 自动化，小团队也能承担起与大机构对薄公堂的精力和成本。 ### `32:00` 法律数据科学化与司法体系的外部化演进 诉讼领域长期存在着类似诉讼投资（Litigation Finance）的决策难题。在风险代理模式下，律师承担了房租、员工薪资等全部运营开支，面对保险公司抛出的和解要约（例如 Geico 提出的 12,000 美元），律师究竟该争取到 15,000 美元、直接搏取 100,000 美元，还是立即和解？由于历史上缺乏全行业通用的和解数据公开库，这类决策往往依赖拍脑袋。Eve 正通过沉淀前所未有的海量和解与判例数据，赋予律所如“数据科学家”般的决策能力，精确定位每起案件的最优解，帮助律师在最恰当的时机选择和解或诉至法庭。 ```mermaid flowchart TD accTitle: AI 时代法律争议解决演化路径 accDescr: 展示 AI 如何通过数据科学辅助原告决策，并将案件分流至非诉讼解决机制 subgraph Decision["案件评估与决策 (Eve 数据科学赋能)"] A["案件事实与历史赔偿数据"] --> B{"和解 vs. 诉讼收益模型"} B -->|"评估最优结果"| C["数据驱动的对抗/和解策略"] end subgraph Resolution["司法分流体系演变"] C --> D["法院诉讼瓶颈 (法庭资源有限)"] C --> E["ADR / 自动化争议解决"] E --> F["AI 中立调解智能体"] F --> G["消除偏见的客观事实裁决"] end ``` 针对未来司法系统的承载力，Alex Rampell 举了自己的例子：在与富国银行（Wells Fargo）发生 4 小时的纠纷无果后，他使用 Codex/ChatGPT 在 4 小时内起草并完成了起诉程序、甚至找到了法律文书送达员（Process server），最终排期却要等到来年二月；专利申请领域也面临着生成门槛降低但美国专利商标局（USPTO）成为瓶颈的困境。Jay Madheswaran 指出，未来司法体系必然会像现有的 EEOC（平等就业机会委员会）、退伍军人事务委员会或劳资仲裁一样，走向法院外部的分流机制。其中最关键的演化是自动化争议解决（ADR）： > “...where AI-enabled mediators or central parties that have a neutral view of both sides are going to use AI to really speed up how they come up with judgment.” 中立调解方将利用 AI 梳理历史脉络、厘清事实，在消除人为偏见的前提下极速输出裁决。 ### `35:00` 律所商业模式重构：软件化、超垂直细分与规模化并购 当每个律师的管理上限从 100 起案件跃升至 500 起甚至数千起，传统小型律所的地理边界被彻底打破。过去，律所被视作一种“生活方式型业务”（Lifestyle business）而非风险投资回报标的，其根本瓶颈在于经济模型：以一起 25,000 美元的和解案为例，律所抽取约三分之一（约 8,000 美元）的律师费，其中约 50% 消耗在人力成本上，30% 用于获客营销，剩下的微薄利润很难再投资于提高服务水平；以往提升产能的唯一途径只有高薪招聘更多资深律师。 然而，AI 正促使律所向软件公司的商业逻辑靠拢。律所开始沉淀出专属于自己的独特数据资产，并在特定细分赛道建立壁垒： - **从大类目向超垂直演进**：30 年前是包办一切的街坊律师（Neighborhood lawyer），随后分化出知识产权、劳动法等专长；如今正在进一步细分至诸如“特定事实模式下的软组织损伤”或特定类型的医疗缺陷诉讼。 - **跨区域精准获客**：超垂直律所针对特定病症或事故模式投放广告的获客成本，远比 Morgan & Morgan 这种巨头在各地铺天盖地打户外广告牌更具性价比。一旦在某一细分领域建立极高胜率，律所即可将单点专长迅速推向全美。 - **巨头的并购整合狂潮**：体量媲美 Morgan & Morgan 的头部大所正在利用 Eve 加速行业并购。过去收购其他律所面临极为痛苦的组织文化、流程重组和人员换血等变革管理，如今通过统一的 AI 基础设施，律所能在保留案件处理逻辑的同时直接交付成果，大幅缩短并购的投资回收周期。 Jay 概括了这一代律所转型的本质动力： > “I think the transformation that's going to happen is a lot of these law firms are going to look more like software companies. I think they're going to finally participate in the software revolution that's been happening for the last 30 years they've been left out of.” 最后，Alex Rampell 指出整个软件行业正在告别传统的按人头按月付费（Per-seat/month）模式。过往各家 SaaS 官网千篇一律的“星巴克大杯、超大杯（Tall, Grande, Venti）”阶梯定价（如 79、99、129 美元）正在被结果导向定价（Outcome-oriented pricing）所取代，法律行业与软件定价模式的交汇正在迎来全新的拐点。 ### `40:00` 从按人头付费到按结果付费：法律定价模式的范式转移 软件行业目前正经历一场剧烈的定价模式震荡。传统的 SaaS 公司大多采用按席位计费（seat-based pricing）模式，但随着 AI 的普及，许多律所甚至不再需要那么多工位或账号，导致买家质疑席位费的合理性，迫使整个 AI 软件生态向“交付成果并按成果计费”（charge for outcomes）转型。Alex Rampell 指出，原告法律业务（plaintiff law）从一开始就是纯粹的结果导向机制（胜诉分成模式），但其他诸如移民法、遗产信托法等非结果导向的领域，客户同样极其渴望按结果买单。Alex 以自己在 McDermott Will & Emery 的遗产信托律师为例调侃道：那位律师极其优秀但也非常昂贵，“我想对我的案子来说，最终的结果就是我的死亡，所以我真不知道自己想不想买这个‘结果’。”现实中律师之所以坚持按小时计费（billable hours），并非单纯想敲诈客户，而是因为委托人脑中的诉求与实际表达之间存在严重的数据不对称（data asymmetry），且项目的范围往往会随着时间不断漂移。 ```mermaid flowchart LR accTitle: 法律定价模式的演变逻辑 accDescr: 展示从传统的按席位与按小时收费向未来基于算力与数据预测的成果定价模式的演进。 subgraph Traditional["传统收费模式"] A["SaaS 席位计费 (Per-Seat)"] B["计时收费 (Billable Hours)"] end subgraph PainPoints["核心痛点与不对称"] C["数据不对称与范围漂移"] D["无法预估算力/工时与预期价值"] end subgraph OutcomeBased["AI 驱动的成果定价"] E["聚合数据 + 算力消耗量化"] F["端到端交付结果并分润"] end Traditional --> PainPoints PainPoints --> OutcomeBased ``` Jay Madheswaran 对这一判断表示完全赞同。无论在原告方还是被告方，最大的难点都在于：在真正接手案件之前，没有人知道它到底需要耗费 10 个工时还是 100 个工时。为了攻克这一难题，Eve 在过去一年多里深入各个具体实践领域，专注开发在今天被称为“长远景智能体”（long-horizon agents）的技术。Jay 强调：“最终我认为，实现法律民主化所必需的，是从一组事实出发，你必须知道需要多少计算和资源才能得出结果。”（“And ultimately, I think, what is needed for democratizing of law is from a set of facts, you need to know how much compute, resources you need to arrive at an outcome.”）当前这种在大型律所和企业法务部门司空见惯的现象——比如企业在 C 轮融资时，因为无法预估工作背后的预期价值（expected value），只能签下一张 10 万美元的“空白支票”——本质上都是由于信息缺陷所致。而随着 AI 能够把案件工作拆解并分摊为确切的算力成本，整个法律行业正迎来类似“大卫对决歌利亚”（David vs. Goliath）式的结构性颠覆。 ### `43:24` Eve 定价演进与法律工作底层价值链的重构 谈到 Eve 自身的商业模式，Jay Madheswaran 回忆称，公司刚成立时团队什么都不懂，直接套用了星巴克式的“中杯、大杯、超大杯”（Tall, Grande, Venti）三档订阅模式。然而深入客户之后，他们迅速发现律师们真正在乎的根本不是软件功能清单，而是“我如何能够服务更多客户”以及最终的案件结果。这一现实迫使 Eve 在一年半前——在 AI 行业宛如“狗年乘上一百倍”的飞速周期中——果断切换为基于成果的定价模式。 Alex Rampell 重新梳理了原告律所的价值分配模型：律所通常抽取赔偿金的三分之一，而这三分之一的收益中有一半要作为直接人力成本消耗掉。如果 AI 能将这 50% 的劳动成本免除，律所就能把省下的开销反向投入到更多算力、更多 Token 和由 Eve 执行的工作中，从而赚取数倍的利润。Jay 认为按小时计费虽然不会彻底消失，但它的命运必然会走向与按席位计费同样的退潮之路；除了少数历经多年才能见分晓的特殊案件外，大部分常规法律业务的成果都是明确且可预期的。Eve 致力于将交互简化到极致：律师只需把招揽进来的客户资料扔给 Eve，系统便会自动推进全部流程，并在关键节点才把律师召回介入。 - 原告律所典型财务模型：案件判赔总额的三分之一作为律师费，其中约 50% 被律师及法务助理的人工成本抵扣。 - 法律服务行业新闭环：律所将节约下来的人力开支直接转化为 Token 采购和算力支持，实现律所利润与服务体量的倍增。 - 产品终极体验：“律师只需说：‘这是我刚接进来的客户，交给 Eve 吧。’Eve 会在需要你的时候再拉你进来，然后你就可以高枕无忧了。” ### `46:17` 智能体攻坚与端到端自动化中的“AI 对轰” 在剖析公司发展的瓶颈时，Jay 指出，Anthropic 和 OpenAI 等合作伙伴公开展示的法律评测基准大多聚焦于孤立的单一任务（single outcome），测试模型在某个任务上是否勉强合格；但现实世界的法律实践完全是复杂的多步骤智能体运作（agentic work）。目前阻碍 Eve 扩张步伐的最大速率限制器并非市场需求，而是顶尖工程人才的紧缺。面对庞大而尚未被满足的法律需求，他们正把每一个具体案由（case type）像科学实验一样解构，试图为每一种案件打造一个能够调动 Claude 等各类底层大模型工具的“微型律师大脑”（mini attorney brain）。光是机动车交通事故（moving vehicle accident）这一个细分子品类的端到端完全自动化，就蕴藏着数千亿美元的市场价值。 当 Alex 询问是否已经实现“从电话接听、客户建档到直接拿到保险公司 Geico 赔偿支票”的完全自主闭环时，Jay 透露他们尚未完全跑通全自动化，但已经无限接近。一个典型的案件往往涉及多达上万次细碎的沟通往返。以人身伤害（PI）案件为例，调取医疗记录往往需要打两周的电话向保险机构和医疗机构索取。Eve 现已开发出由语音 AI 驱动的外呼智能体（voice AI outbound agents）来承接这项工作。有趣的是，在这一环节中已经开始上演“智能体与智能体互相通话、各自摸索破解对方套路”（agents talking to other agents and learning how to get past each other）的奇特攻防景象。尽管目前依然需要少量人工介入来优化最终成果，但端到端的全自动处理已经在一部分客户的生产环境中进行实质性内测。 > “在 Eve 的外呼智能体中，我们实际上使用了语音 AI 来拨打外呼电话，有趣的是，在那个领域我们已经看到智能体在与另一个智能体交谈，并且它们在互相学习如何绕过对方，这太搞笑了。”——Jay Madheswaran（“And with Eve Outbound Agents, where we actually have voice AI that can do the outbound call, and funnily enough, that's an area where we're already seeing agents talking to other agents and learning how to get past each other, which is hilarious.”） ### `50:00` 攻守易势：AI 如何瓦解被告律师的“假期突袭”战术与商业模式错配 在谈及自动化与效率提升时，Alex Rampell 提出疑问：当原告律所的办案成本大幅下降时，其对手方（如保险巨头或被告辩护律师）的反应与博弈策略是否也随之发生了改变？Alex Rampell 指出，像 Geico 这样每年处理数万起理赔的保险公司通常有着极其僵硬的理赔算法和谈判规则（“如果你索赔这个数额，我的第一次和解出价通常就是那个数”），而不是随意开价。但如果原告这边的边际成本彻底归零，博弈的平衡是否会被打破？ Jay Madheswaran 回应指出，目前被告辩护方依然严重依赖传统的按小时计费模式（billable hour），因此行动速度并未跟上原告律所的变革步伐，这也造成了双方极大的商业模式与定价不对称性。Jay Madheswaran 评价道：“I would say largely still, the defense side very much still works on hourly performance business model. So they haven't quite caught up to the same pace as our attorney. So like there's a lot of, I guess, pricing asymmetry here.” 过去，被告律师非常擅长利用心理战和流程折磨原告律师。Jay Madheswaran 分享了一家合作律所遇到的真实案例：被告律师经常故意将证据开示（discovery request）的送达日期拖延到感恩节或重大节假日前一天。他们明知原告律师即将去度假，企图借此毁掉对方的假期，迫使对方在重压之下仓促应付，交出质量低劣的答复。 然而，面对这种反复出现的恶意拖延策略，该律所直接利用 Eve 构建了一个专门的 AI Agent。下一次当被告律师再次在节假日前夕发起“证据开示突袭”时，Eve 的 Agent 当天便全自动生成了完整、严谨的答复并直接回传。Jay Madheswaran 描述当时的场景称，被告团队彻底傻眼了：“So next time this came through, like they responded immediately with the responses and the defense were like, what is going on? The tactics don't... And like we're seeing there's so many variants of that that is happening now, which is causing defense to realize, wait, this is what happens when the other side has resources. Oh, crap. And that's what's fun to see.” 当原告方凭借 AI 拥有了近乎无限的算力和即时响应资源时，被告方沿用了几十年的传统消耗战术瞬间失效。 ```mermaid flowchart TD accTitle: 被告传统证据开示消耗战术与 Eve AI Agent 应对流程对比 accDescr: 展示被告律师在节假日前发起证据开示突袭，传统原告面临假期被毁或仓促应付，而引入 Eve AI Agent 后实现当日秒级答复并破解战术。 subgraph DefenseTactic["被告辩护方策略（按小时计费模式）"] A["拖延证据开示节点"] --> B["重大节假日前一天突然送达 Request"] end subgraph OldWay["传统原告应对困境"] B --> C["律师被迫假期加班"] C --> D["仓促草率答复 / 丧失谈判主动权"] end subgraph EveAgentWay["Eve AI Agent 应对方式"] B --> E["Eve 证据开示 Agent 自动解析"] E --> F["当天即时生成严谨合规的答复"] F --> G["秒级回传被告 / 彻底瓦解恶意心理战"] end ``` - **按小时收费与风险代理的错配**：被告律师按工时获利，倾向于将流程拉长并制造繁琐文件，而采用风险代理的原告律所借助 AI 实现了速度与吞吐量的指数级爆发。 - **节假日开示突袭（Holiday Dump）失效**：以往利用休假空档逼迫原告方犯错的心理消耗战，在拥有 24/7 不间断处理能力的 AI Agent 面前被彻底化解。 - **“拥有资源的对立面”**：过去只有大型企业辩护团队才能凭借庞大助理团队实施“文件轰炸”，现在小型原告律所凭借 AI Agent 具备了对等的反制与瞬间交付能力。 ### `51:40` 总结展望：将 Eve 带向万家律所与法律普惠的未来 在播客的结尾，Alex Rampell 对整场探讨进行了复盘，惊叹于一个典型的伤害诉讼案件中竟然存在多达 10,000 次沟通与触点（touches），并对 a16z 能够作为 Eve 的投资与战略伙伴、助力其扩展至接下来的 10,000 家律所感到无比振奋。 Jay Madheswaran 表示，Eve 目前的征程才刚刚起步，尽管团队每天已经协助处理数十万名委托人的案件需求，但他们的终极目标是触及并改善数亿普通人的生活。最后，Jay Madheswaran 借此机会向对该领域感兴趣的优秀技术与业务人才发出邀请，欢迎加入 Eve 共同重塑法律行业的运转体系。播客在 a16z 惯常的投资免责与合规声明中圆满结束。 - **办案复杂度认知**：Alex Rampell 坦言，此前未曾意识到一桩案件全生命周期需要经历上万次交互触点，AI 的介入让这些原本耗尽人力的琐碎节点得以高效串联。 - **业务覆盖面与愿景**：Eve 目前每天协助数百家律所处理数十万名委托人的业务交互，长期愿景是通过降低维权门槛，让数亿潜在受害者都能以经济可承受的成本获得顶级法律保护。 ## Notable quotes > `15:00` 事实证明，软件市场实际上就像是一个庞大的影子劳动力市场，它的规模远超任何人的预期。 — Alex Rampell ## Mentioned Alex Rampell, Jay Madheswaran, Eve, a16z, Albert Einstein, Blue Bottle, Wilson Sonsini, Microsoft Word, Harvey, OpenAI, Anthropic, Claude, Colossus, Geico, State Farm, Auditor, EEOC, USPTO, Wells Fargo, Codex, ChatGPT, Morgan & Morgan, McDermott Will & Emery ## Video description What happens when AI makes legal work cheap enough to take cases lawyers couldn't afford to touch before? Eve CEO Jay Madheswaran argues that AI won't just cut costs for law firms. It could expand access to legal representation by making an entirely new class of cases economically viable. Jay joins a16z General Partner Alex Rampell to explain how Eve is automating plaintiff law from the first client call through settlement, and why some firms are already using it to increase the number of clients each attorney can manage from roughly 100 to 500. They also discuss how AI could shift lawyers toward higher-value advocacy, why agents may eventually negotiate with other agents, and what happens to the legal system when bringing a case becomes dramatically cheaper. [Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/de88968f-40b2-409c-befe-1d337a8ca6a9/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=de88968f-40b2-409c-befe-1d337a8ca6a9&feed=JGE3yC0V)

[15:00] Alex Rampell: 事实证明，软件市场实际上就像是一个庞大的影子劳动力市场，它的规模远超任何人的预期。

[0:00] Jay Madheswaran: If you drop Albert Einstein into a law firm today, he will not be productive for quite some time. [35:00] Jay Madheswaran: I think the transformation that's going to happen is a lot of these law firms are going to look more like software companies. [4:47] Jay Madheswaran: Today, if a lawyer doesn't make at least $5,000 fees off a case, they do not take it on. [10:00] Jay Madheswaran: The math on the law firm side is very simple. The more people you represent, typically speaking, the more money you make.

[0:00] Historically, plaintiff law really didn't need more than a cell phone email address. That was kind of it. No software really solved their core problem, which is labor constraint. Most people think of AI as like, oh, you're going to fire everybody and cut costs. And it's like, no, actually, like if I use Eve, I make a lot more money for my firm. People can't pay $500, $600 an hour on a blank check for however long the case might take. What Eve is aiming to do is really handle that whole thing end to end. One firm in LA went from managing about 100 clients per attorney to 500. Wow. That is a type of thing that is possible. Why is Eve better than just kind of like raw intelligence? If you drop Albert Einstein into a law firm today, he will not be productive for quite some time. That is a general problem that Eve is solving. And why won't that person be productive? Well, it's due to a number of things. One is millions of people have legal can't afford to take on. AI could change the economics. In this episode, A16Z general partner Awe Trampel sits down with EVE founder and CEO Jay Madeswaran to discuss how AI could expand access to legal representation. Jay explains how EVE handles work from client intake through settlement, allowing some attorneys to manage five times as many cases without simply adding more people. They also discuss what happens when AI agents start handling thousands of steps across why lawyers could spend more time on actual advocacy, and what the legal system might look like when agents increasingly negotiate with other agents. big supposedly white shoe, I guess is the term law firm is where you go to Harvard Law School and you bill $2,000 an hour. But there's like another side of law, which actually is bigger, more distributed. And I guess we would call that plaintiff law. So like every time you drive on 101, you see all these billboards and maybe kind of talk about what is EVE, what form of law and legal AI do you focus on and maybe a little bit about the industry in general, because I feel like most Americans are familiar with it, but they don't necessarily realize. Maybe explain EVE in your own words. Absolutely. One, thank you for having me here, Alex. Always fun to talk about EVE. What EVE is, is legal AI for real people, right? Like people like you, me, not institutions, not in-house legal organizations. not the people that already have all the resources. But how do we actually provide legal services to people who, that otherwise can't afford it. And this is how the plaintiff law industry really came about. Because what ended up happening was, you talked about how Wilson Sincini might charge $2,000 an hour, and that's really the business model. Lawyers' times are expensive, and you're trying to point lawyers towards more and more expensive time. The same is true for most lawyers. They have a lot of bills to pay, and today, legal work, the more work you put behind it, you can represent your facts better. You can get a better outcome. That is the state. of world we live in today. To account for this, Lawyers understand people can't pay $500, $600 an hour on a blank check for however long the case might take. So the plaintiff industry came about on working on contingency law. So what this means is they only get paid when they win the case for the client. And this is what's allowed them to even have any sort of leveling playing field with people versus institutions. And what Eve is aiming to do is really handle that whole thing end to end. of how do you represent a client all the way from the very first call to getting them a settlement How do you do that end-to-end thing? Purely with AI. and give that to the lawyers to then amplify the outcomes that can actually provide. And that's what Eve is doing and think about giving David a slingshot to take down Goliath. That's what we're doing. So what I find fascinating about this is you really, I kind of think about econ 101, where you have a supply curve that goes up into the right. And what does that mean? Because remember, there's price and quantity. So here's quantity and here's price. So as the price goes up. then the quantity supplied goes up. So if I can make $100,000 by doing something, wow, I'm totally in. But the demand curve is the opposite direction. - That's right. - So price and quantity go this direction, so the demand, the number of units that I want, actually I should say this, [5:00] This is quantity, this is price here. So price goes down, quantity demanded goes up considerably, and if I think about applying this to law, like what if I have $1,000 of damages? I can call a lawyer and, even for contingency, right? They might not want to take my case because if they take a third of the outcome, it's like they want a bigger case and it's going to require a certain amount of work. And AI really, it kind of liberates these market failures to a certain extent. Because you mentioned kind of giving David a bigger slingshot. A lot of times David just, well, number one, if I have $1,000 of damages, I certainly can't pay $1,200 an hour for a lawyer. So like I'm out for that. The problem is that the settlement might not be big enough, so the lawyer might want to serve me, but he or she has paralegals that they have to pay and so on and so forth. But if the software can do the work, now the market opportunity is so much bigger. You can actually deliver justice for a lot of people that otherwise, they weren't priced out of it. Because how are you priced out of contingency? But it's not really in the interest of a lawyer. They want to serve a thousand customers that they get, but they can't. They have to focus on the biggest ones to the detriment of the smaller ones. Because again, it's kind of the right of that equilibrium point. I don't know if you agree with that or how you think about that. You're a person. You deserve a great lawyer. to represent your case. And usually you only need a lawyer when your back is against the wall, right? The best of us, the worst of us comes out this time and we don't have the skill sets to represent ourselves. That's why this whole profession exists. That's why the legal system is about a third of our government. It's for this reason. And today, What's happening is because Lawyers still need to get paid. It's still bottlenecked on legal work. Today, if a lawyer doesn't make at least 5,000 fees... Offer case. they do not take it on. That is the market dynamics today. And even at the 5,000, they're losing money on a lot of cases. They're really banking on a few cases paying out very large. They're not even guaranteed to win, period. So they already take that. That's the contingency aspect. But it's kind of like, even if I win, I lose because my costs of carrying the case are going to be higher than the actual potential outcome even if I win. That's exactly right. I Most people here have been in a car accident, fender bender of some sort. If it's below the deductible, No lawyer in America will touch that case. It's just not economically worth it, even if the deductible is $3,000, $4,000. To an individual, that is a difference between a paycheck being gone versus not. And there's no representation they can get for that. And the reason is actually from their side, We have to understand from their side too, right? So let's say you've been getting harassed at work. and your employer could stop it. Like in hospitality industry, let's say you could tell your customers, "Hey, don't harass my employees," and change processes to make that happen. Today, That doesn't happen. Because when that person gets calls... The lawyer on the other side has to pick up a call, Field that call. That's about 30 to 45 minutes of work. And then after that, go do legal research because in employment in America, pretty much everything's at will. And you have to tie... a problem to like a protected clause of some kind. And it's called causes of action. So you have to tie it to a cause of action. And that work, requires them to go spend 10 hours to go research your case law, research the facts, go back and forth and ask them more follow-up questions, see if they qualify. And while they spent all that time, they might still realize... hey, this doesn't even make me $5,000. Right. So... That is why 96% of cases on the labor and employment side, for example, gets rejected. So you can never find anybody to represent you unless you come out of pocket. In many cases, people don't have money to come out of pocket. That's exactly right. So you just kind of take it unless something really egregious happens. And that is the current state. And it's all bottom-neck ton legal labor. And this is what EVA's meant to do, is to, how do you now take that 10 hours is now, immediate with Eve. A call comes in, Eve actually answers the call. and, automatically uses his insights on the law firm, to know whether or not this is a case the lawyer has to take on or not, and then hit the ground running, with a demand or the initial preliminary work done. And that completely changes everything. The economics of that type of case... And we're still at the very beginning stages. Like what's happening is... Eva's getting more and more capable of handling outcomes. end to end over a very long horizon. These cases take six months, a year to launch, and through them, probably a thousand actions. back and forwards that are happening. And as accuracy of Eve is getting higher and higher, this long horizon becomes possible. So the more this becomes possible, even more complicated work [10:00] So more ties down to a constant case. So now all these people without representation, get representation. And if you really ask yourself, like, How many times in your life? if, Asking for legal help wasn't a friction point. Would you have asked it? That number is way higher than we think. Yeah. So it's a funny story. This is a true story. I was in Blue Bottle, a little coffee shop in Los Altos, and I was explaining EVE to somebody probably nine months ago. And I was explaining it because most people think of AI as like, oh, you're going to fire everybody and cut costs. And it's all about cost savings. And it's like, no, actually, like if I use EVE, I make a lot more money for my firm. And it's not because I'm cutting costs. It's like I can take on more cases. It's actually revenue expansion. and at the end of this I mean we were talking about many topics but Eve came up and this guy comes up to me looks a little angry he has a big dog and I'm kind of afraid of big dogs and he's like what was that company you were talking about? It's like, oh, is he going to get mad? He's like, I'm a trial attorney. I want to try it. Oh, great. So I got you a new customer there. Thank you. But the whole point is that you expand the revenue opportunity, but maybe kind of walk through like what do the dynamics, because everybody is driven on 101 or I-95. I-95 is the corridor that goes from Florida all the way up to Maine or something, you know, up and down the Eastern Seaboard. And it's like I've driven almost all of that. And it's like I feel like half of the ads are for trial attorneys, plaintiff law. So they're obviously advertising there because they want to get called. It's like, hey, I have this accident or I have this issue. But what do the economics look like? So this is why, again, I think Eve is so fascinating because I have to turn down 90% of the leads that I'm getting, I assume. lawyers. You're a revenue expansion play for them. That's from the legal side. But what do the economics look like for a law firm? That's exactly right. When most people drive by, they see the face and They forget about the people behind them and the amount of work every single call actually takes. That's what happens today. That's what Eve is there to help with. And the math on the law firm side is very simple. the more people you represent, Typically speaking, the more money you make. But because you can take on more people and you... honestly are denying and not taking on many cases because they're too expensive to take on, the demand from their perspective is not really constrained. It's really the supply of labor. Like, do I have... enough people, inside to actually service that problem. And to make things more complicated, Different firms here are in different realms of problems. where the bottleneck keeps shifting. So a small firm might just have one attorney that's has a case manager or some sort of office manager that feels initial call, and just notes it down. And then the lawyer then response back in their free time. They're kind of running chaotically, running around. And then there's a next level, which is they find a way to systematize it. They have an intake team of call center somewhere that intakes it. And now that becomes a new bottleneck. So now when you increase your marketing spend, let's say you looked particularly attractive in that billboard and a lot of people are calling in, suddenly now that team is overwhelmed. And you're limited, again, by how many people you can represent. And this is the daily challenge these small businesses are dealing with. They're small businesses. They're not massive enterprises. They want to represent individuals. And everyone here went to law school, not business school. They want to be there for the people. So we're basically consolidating all of that into the EVE platform. such that, they are getting pulled in only for what they need to. like their lawyer insights, their particular insights on how do they take This case, Eve keeps raising the floor. How do I take it from this floor to the next level? And that is where I believe really all lawyers spend time talking to their people and helping get client outcomes. Well, and this is kind of the market failure point to a certain extent, which is I could serve all 5,000 leads that came in if only I could hire a paralegal for 12 cents per hour, right, or whatever the token cost of it. But it's like, I can't. So I would just, there's no labor market that could actually function. But with AI, now I can. Hereafter, I have a nice head of hair. But what's that for your best customers that you work with? I imagine the revenue transformation is pretty significant. Yeah, one firm in LA went from managing about 100 employees clients per attorney. to 500, wow, right? [15:00] And that is... That is a type of thing that is possible. And they went... To their credit, they worked really hard to achieve this. They utilized EVE up to the max of all the way from intake, fielding the call, doing qualification, doing it. automatically automatically removing all the relay problem that they have of passing things from one attorney to another, to a case manager, all that is now handled within agents in EVE. And all the way to settlement where it's actually... briefs before they go to a mediator discussion that actually tells you, here's what you need to focus on. Here's what the settlement outcome should look like. for cases like this that we worked on. And here are the tactics used before that got you here. and That is what the future looks like. It's truly... moving the case forward continuously. And you're helping making it do a better job than it could do by itself. That is, I think, the future of law. where you as a lawyer are really going to look more like a... almost a data scientist and engineer, engineering something to be better and better. And that is a shift. the top law firms on EVE mentally kind of go through? Yeah. Well, I kind of think of if you look at your average law firm where they don't really know AI, and actually, like, this is kind of an interesting kind of take on the world of AI-enabled applications in general. Like, what was the software stack for law? before, well, if you're a big firm like Wilson Cincini, probably the most important thing is like your time tracker, right? Because you're billing by the hour. If you're in contingency law, it's like, well, your time matters, but you're not billing your customers on that. You probably need to have like a, you know, each case, here's how much money I'm spending on the case. But like, I imagine your primary tool was Microsoft Word. And there wasn't, and this is how like VCs would have like gotten this industry wrong before, which is like, oh, well, the software market for if law is small. They all buy Microsoft Office. Yeah, Microsoft Office is kind of expensive, so therefore the software spend is in the tens of billions of dollars a year, or something seemingly comically small. But wait a minute. the outcomes that they deliver is in the hundreds of billions of dollars a year. They can't deliver more outcomes because they're really constrained on labor. and they can't hire labor for 12 cents an hour because it doesn't exist. So it turns out the software market is kind of like this shadow labor market, and it's like massively bigger than anybody would have expected. And like it's always just a weird thing because it's like you can't say the software market is small because you're a legal software company, but people know Harvey or they know like corporate. or e-discovery tools that have been around. But I imagine, what was the main software stack before Eve entered? Was it just Microsoft Word? Or what was, in order for me to go into business, I graduated from law school, I want to be in this particular subsector of the field, what was my software stack? You know, this is the unfortunate thing about plaintiff law and institutions, is historically, you know, it's a lot of people who are in the field. Plaintiff Law really didn't need more than a cell phone Email address. Honestly, even Microsoft Word, some people in the solos would not buy it for cost reasons, other reasons. No, it's because if you think about it, you actually don't do that much redlining. It's a trust boundary situation. So your one side is... kind of opposition to the other side. So you're not really collaborating. Yes. In the same way that you are when you're doing contract review or something along those lines. So that was kind of it. And no software... really solved their core problem. which is exactly what you mentioned. They're labor-constrained. That is the core of how they do anything at all. And because they have labor and they're good at managing their labor, they found a way to solve every problem to them look like solid labor. And that's how they've they've grown up. And as a result of that, What ended up happening was a lot of resources on software went towards... defense institutions. Because that's where the budgets work. For Harvey, you know, really any company, honestly, legal software companies today, vast majority of them are on the defense side. Because that's where the budgets are. And that's what's easy to build for, easy to get money for. And on the plaintiff's side, These are small businesses that are watching their pocketbooks every month. If you're not driving outcomes for them, From day one? You're out. You're not getting renewed. they're, They're going to curse you out. You know, anything they can do, they're going to come after you. Right. And that's what's made things so difficult. But what we realized is, well, Their core problem is we know what their incentives are. They want to represent more people. They want to win. And we can do that now. You know, there's one... The beauty of software today is [20:00] Once you do a task manually, even tasks that include crunching new piece of data and transforming it, That's kind of what AI is really good at. You can now encode it as a process. That is a fundamentally new shift. That has never happened before. That's what they were trying to do with people operations. but never worked because every single case still different enough that you can't do it. It's still bottlenecked on hiring smart enough people to be able to handle it. And now... one smart person can actually handle like we talked about 500 different cases of very different facts, very different situations, And how do you even keep that in your brain? It's hard. Right. But now you don't have to. And I guess one of the things that a lot of founders are thinking about right now at the application layer is like, oh, maybe I'm roadkill from the big guys, right? Like the labs that produce the intelligence, maybe they're going to go tackle MySpace next. And I'm toast. And there are many ways of defending against that. But defending kind of sells it short because if you're able like. if you have a network effect or if you have data that actually nobody else has, you can deliver better outcomes. So, you know, who cares if everybody uses ChatGPT or Claude behind the scenes or on their own? But maybe talk about like, I'm sure you've thought about this a lot. You know, why is Eve better than just kind of like raw intelligence? And how does it get better? Because it does. The more practitioners that use the product. Yeah, we're betting on these things. models getting better and better. And our value is exactly that. If you have a thousand Albert Einsteins to work with to do litigation or the best attorney ever to do every single task, we would take it. The problem is if you drop Albert Einstein into a law firm today, he will not be productive for quite some time. And that is a general problem that Eve is solving. And why won't that person be productive? Well, it's due to a number of things. I think one is in plaintiff law, the vast majority of data is actually non-scrapable. about less than a percent of cases actually settle and get turned into case opinions. Vast majority settle pre-demand or pre-litigation. And some even before that, right, where it's actually just a negotiation of, hey, this is how I was wronged. Please fix it. And to make that concrete, it's like, I'm in a car accident. I'm not at fault. You know, the adversary, if you will, like the other side's insurance company is like, oh, you weren't that hurt. We'll give you a dollar. Right. They always try to like deny, deny, deny. deny rule number two is deny. It just goes on. Those are all the rules of insurance carriers. So now I hire a lawyer to actually get what I'm owed because I actually am injured and my car is actually totaled and they just don't want to pay me. You get a settlement. It doesn't go to case. It doesn't go to actual trial. But you know that Geico actually did pay $29,000. And that's very, very useful information, I assume. That's right. Because insurance is the one that's dealing with community. They know all the cases, but because it's sharded across all the different local firms, they don't know exactly Geico's every single move. That's exactly right. And they even have ominous sounding software like Colossus, which tracks, they try to basically say, a broken thumb, this is how much we think it's worth. Which is... a hard thing to do. In their defense, it's a hard thing to do, but they are doing that, but also saying, this is how much we can get away with giving them away for. and irrespective of that individual's actual need. of their particular situations. And that's what the legal system is meant to fix. That's what it's there for. And what now is happening is, and this is the interesting bit and the challenge, is if you go and Google, how do you write, an amazing demand letter, Again, State Farm for soft tissue injury cases like that today exists nowhere. even for a new... paralegal that is joining a plaintiff firm, This is something they have to learn. And on top of that, it gets more complicated because when you're in local jurisdiction, you're not going to be able to do that. The local... environment like the court systems, your opposing counsel, they all behave a little differently. And that context matters, too. So what Eve is doing is it's turning this historically incredibly difficult to utilize multi-attribute, multidimensional information that is basically what is the best thing I can do at this particular moment. to improve the outcome of the case. And that same question is repeated thousands of times during the case. And a Human brain is just not capable of kind of keeping all of that in mind. but AI is... And that is the type of data we help build for law firms and they're seeing the immediate value from it. And the first place where they see value is, of course, intake. Who do I even take on? Who am I good at representing and maybe [25:00] You start there. and keep getting better. Right. And I imagine, um, Because these aren't giant firms, Like if I'm a lawyer in New Jersey with a four-person, four-attorney practice, and over there there's another lawyer in Arizona with a seven-attorney practice, but their counterparties tend to be State Farm and Geico and whatnot at the limit, it's actually fine to share information? It's kind of like, you know, I work with a dental AI company and it's like, well, the dentist in Iowa can't serve patients in Florida, but they want to learn from the collective actions of each other. But I guess, how do you think about, you know, just data sequestration and whatnot? Does that ever become an issue or how do you think about using it? We've largely followed the rules of engagement that already exist in this industry, but we've amplified it with software. That's kind of our stance, right? So the rules of engagement is for their own cases, they want to be very private. Like that is important by law. It's attorney client privilege. And we don't train on that data. ever. However, what attorneys also do is to share this knowledge, they actually travel, they attend CLE, and they share tactics with each other on how they went up against State Farm's X and Y defense. Because State Farm repeats that defense across the board, right? And not picking on State Farm, anyone is fair game here. But that is what they're already doing. So we hire our own attorneys and track success rates of cases for our platform. And we want to help, if you're using EVE, you should have more success. That's what we track. And we have this product called Auditor, which actually... analyzes every single case. through every single page, overnight, and looks for... situations that a perfect attorney would catch every time. and, Across the board, unfortunately, vast majority cases have, almost every single case has missed information. Right. And now that's no longer a thing. Yeah. And maybe talk about the future of where things go. Because on the one hand, I can imagine a world where it's like, one of these days Geico is going to get a little bit smarter. And they're probably going to like, you see their deny, deny, deny plan. That's how they got smart before, right? Because they realize it is an adversarial system. Like, you know, I lost my house in a fire and I'm going to claim I had a Picasso in there. And then they're going to think that I'm claiming that I have a Picasso. They're just going to deny the claim, deny, deny. Like, so each party is kind of fighting each other, but they're going to get superpowers. I mean, you see this playing out a little bit with Muse and It's impossible to get. And they have a human answering the phone. I can DDoS them now that I have to pay one penny per hour or something. And then eventually they're going to realize that they're getting DDoS and they're going to prevent that. So, you know, I guess on the one hand, like, how do you think about, you know, your your adversaries here? Not not your adversaries, but like your clients adversaries. How are they adapting to a world where it's like, you know, it was great for Geico where they could say, hey, you have two thousand dollars. damages. The consumer is like, yeah, I did have maybe $2,500 of damages, but it's like they never really had to worry about that before because of the quantum. And now, unfortunately for them, but fortunately for the Davids of the world, like people that are actually aggrieved and have been victims, now they can fight back and you're powering that. So that's kind of one thing. But also you mentioned long horizon agents, because what does the practice of law look like attorneys are almost like the data scientists of the future. But maybe talk about what does the adversary or the adversarial process look like? What are the skate farms and GEICO's using, potentially as a result of you entering the scene? But also, big, big question, but how does law evolve and emerge just given that you're kind of manipulating words and finding facts and doing all have an important role of shepherding the case to completion, but like, what does the AI do versus not do? Yeah, I mean, ultimately, my framework for thinking about this is ultimately when you have an individual, they're going to need representation. Whether that representation, like who does the work on the representation, you can assume it's 99% AI, maybe 99.9% AI, there's always going to be that little bit that is the human touch, right? And We... do believe that people are going to make a positive difference to the outcomes of the case. And that is exactly what... [30:00] you have, right? So like what you have on the defense side or really institution side is what's always happened, which is they have more resources, they have more money, they have more time, which is actually another weapon they have, which is quite powerful. Individuals are in depth to pay their fees, And that is a very powerful tool they have. And, That is the strengths they're working with. and they're trying to rely on centralization and standardization. On the defense side, you have a attorney that is fighting your case. and pushing them to get to litigation. And I think that is going to be what happens. I think a lot more cases are going to have teeth. And all the routine back and forth, I think, is going to happen through agents talking to agents. I think that's, you know, it's crazy future world, but I think that's where we're headed to. And the difference maker is what does the attorney understand exactly? about that client, That set of facts. and their local environment of who they're going up against locally in that defense firm. And they're using that insight. to make a difference. for their client. And I don't think that ever goes away. Um, it's, it's in fact something we're seeing already where many customers on Eve are actually choosing to spend more time on their trial work. because it actually enables that. Previously, going to trial took down your entire firm. Suddenly, you're just spending all your time doing trial prep, interviewing juries. You want things to settle. Exactly. It's more of a settlement motion. That's exactly right. And instead, I think things are going more into real advocacy. How do I actually make Our point known. How do I, you know, my client is against Back of the wall. They're doing bad things, they're doing good things. How do I represent them? in what actually matters for them. I imagine this is kind of the data science side of what you're doing as well with this giant data set that's probably never been accumulated before. You mentioned all these things. It's so bespoke. There's no communal repository for all of the settlement amounts. But I've seen this with litigation finance a little bit where... If I'm a lawyer, if I'm a trial attorney, and I'm not getting paid by the hour by you, my client, well, where do I pay myself and my staff and my office rent and everything else? Well, there are companies that will show up and say, I will fund this case. That's right. right? And I'm funding this case because I'm counting on you winning it. And like, they'll do all sorts of clever things like here's the judge and here's this, like they're using data. So outside of, I think those guys have been very data science forward. But historically, like this data just like never really lived anywhere. And now that it does, you can make smarter decisions as an attorney. I'm like, do I take it to trial or do I accept the settlement? Or do I push back? Oh, Geico offered me $12,000. Should I push back and fight for 15? Should I, try to win 100? How do I make those decisions? That's exactly right. Which is why we can help them make the best decision in every single case. And know when an optimal outcome has been achieved. And I would say it's still the attorney's responsibility to push it even further. And I think that is what I meant by data science-like work, because they're using the power of software to... do work you never even thought to do. to improve the outcome. And I mean, where do you think this is a very broad question, of course, but like, where does the legal system go? I was having some some trouble with Wells Fargo. And I'm like, you know what, I've tried like for four hours to do this, I'm just going to sue them. So I went to Codex, you know, ChadGBT, I was like, sue Wells Fargo. Four hours later, it had sued Wells Fargo. It actually found a process server. I got the notice of, like, we're going to court, but, you know, the date is in February. Like another kind of similar vein is happening with patent law, where before it's so much work to go file a patent. Now it's less work to file a patent, but you still have the USPTO, which is the bottleneck. So I guess, I mean, if you were to prognosticate, like, where does the legal industry go now table stakes and you're delivering, you're part of that table, obviously for plaintiff law. But if you think about the entire practice of law in general, how does it change? If you were to fast forward five years, what does it look like? What's similar? What's different? Most likely, I think it's inevitable that you're going to have to create systems where not everything goes to the core system. And this already is a framework that is used in law. Right. Like, I mean, for some types of veterans of fair cases that goes for a different committee. for even labor and employment, that goes through EEOC in a lot of cases before it goes through. And employers have figured out different ways to mediate and have arbitration that doesn't necessarily go through the court system. Right. And this is the framework that actually naturally evolves. And I expect that to happen more and more. And I think the tweak that's going to happen is that for some pockets of it, we're probably going to go through some version of like ADR, automated dispute resolution, [35:00] Where... AI-enabled... mediators or central parties that have a neutral view of both sides are going to use AI to really speed up how they come up with judgment. And I don't think the judgment will really be replaced, but, What they will get is a set of facts that shows... historically, here's how this happened, here's how this happened, and it's up to them to remove the biases that might exist. And I think we're just going to spend more and more time on that side of things. - And do you think the small law firms get bigger? Because that would be the other thing. It's like if I could only handle 100 cases per attorney, now I can handle 500, but in the future maybe 5,000, why do I limit my office to a very, very small New Jersey location? Maybe I should go more broad. And you're going to have larger, almost corporate law entities. I think the biggest trial attorney firm is Morgan & Morgan. Is that right? And they have billions of dollars in revenue, but your average firm is quite small. But you can imagine powered by EVE and with all the changes a billion-dollar revenue business, it's kind of like venture outcomes versus kind of like what you would call a lifestyle business. You were a VC once upon a time. You know that term. Somewhat pejorative term for like you make profits. But how do you think, like, is that a reasonable take that like some of these small guys can actually, I mean, obviously you can't go public as a law firm for now, maybe in Arizona or something you can, but do you think there's a chance for some very large institutions. I think the transformation that's going to happen is a lot of these law firms are going to look more like software companies. I think they're going to finally participate in the software process. revolution that's been happening for the last 30 years they've been left out of. right? And if you kind of deep dive, you'll understand why. Let's say you have a $25,000 settlement. The attorney... gets maybe a third of that. let's call it 8,000 for simplicity. And, you know, like, About half of it goes towards labor spend. 30% goes to marketing. And the rest maybe is profit. right like that that's kind of the business as it exists today because of that Business. It is so difficult to reinvest back into servicing your next client better. The only way you can really do that is by hiring a better attorney today. What changes is now they can actually accrue a database and data set that is unique to them that gets them results. that is better than their competitors. for a subset of lock. And this is kind of specialization. And historically... This industry already went through that. 30 years ago There was a neighborhood lawyer that handled everything from employment cases to IP to what have you. Exactly. And now what we have are specializations of practice areas. You have someone that's really good at IP, someone that's good at employment. And now it's getting even more specific, right? Like someone that's good at, let's say, nerthing bird defects or something along those lines. And I can anticipate that to go even another level. Where now, someone could get really good at soft tissue injury cases that involve X and Y fact patterns. And it Advertising for those cases might be a lot more cost effective. than some other larger firm like Morgan & Morgan, just showing their billboard everywhere. So get good at something locally and then expand across the entire country because you have such high competence around that one field. That's exactly right. And so I think we're going to see that type of specialization. And I think we're also going to see the opposite, where some of the large companies, We're also working with some on the scale of Morgan Morgan. Some of them actually are coming together. So those ones that got us big... are actually going around buying other law firms, because they realized with EVE, they can very quickly consolidate what used to be a very painful reintegration process. Now it can be done on exactly what the client wants, which is, hey, the client wants a legal outcome. And Evis made just for that. So by doing that, they get to largely, we work away from a lot of the change management, uh, workforce changes, operation changes, because Eve handles a lot of that for them, they can speed up their time to making that acquisition worth it. Well, the funny thing is that there's this movement underway to outcome-oriented pricing. So like software SaaS used to be per seat per month. And I used to joke, it was like the tall, grande, venti model of software selling. It's like you go to like any SaaS company's website. This is why, you know, Claude can clone them all so readily. They all look the same. It's like $79 a month, $99 a month, $129 a month, you know, tall, grande, venti. And it was per seat. And like look at a company like Zendesk, which had $2 billion of ARR. [40:00] When they were last a public company, they were taken private by Premier and Hellman and Friedman. And now they're like, uh-oh, how many seats do you need? Like maybe zero. We need to deliver outcomes. So everybody in the software world is kind of like in a lot of AI companies, like we deliver outcomes. We charge for outcomes. The interesting thing is that the form of law that you focused on was always outcome oriented from the very beginning. But you have other types of law that are not outcome oriented where I guarantee all the clients would much, much, much prefer that it was outcome based. I mean, sometimes it's tricky. Like I have a trust in a state attorney at McDermott, Will & Emery. She's very, very good. She's like so expensive. Like I guess my outcome, the outcome is my death. So I don't know if I want that. But, you know, you would prefer to just say like there is a fixed cost. But there's often this data, you know, in their defense, the reason to charge by the hour is not so much to rip off the client. But there's data asymmetry between what I want in my head and how I articulate it. Maybe the scope of the project is changing over time. But other forms of law, I would argue they probably should look more like plaintiff's law. They should have outcome-oriented pricing as well, right? Or immigration law, that should be outcome-oriented. Everything should be outcome-oriented pricing, but it's hard because there's a data deficiency, if you will, in terms of will I win this case on behalf of the defendant? But I guess how do you think other forms of—you could probably expand well beyond plaintiff law— I don't know if you agree with me, but how do you think about just law in general changing? Because I kind of feel like plaintiff's law was ahead of everything that we're seeing right now. It's like everybody wants outcomes. Nobody wants to pay for the ingredients or the inputs. This was the only kind of field probably of business where it's like it was always outcome oriented, but kind of shouldn't other parts of law work that way? That is our perspective. exact belief as well. Right. So like the hard part has been, it is hard. you know, when you're on the defense side or even the plaintiff side, the question around In your case, is it 10 hours of work? 100 hours of work? In many cases, you really don't know until you take it on. And for plaintiffs, that is why we had to go so deep into each practice area, into achieving outcomes. And we were doing what is now coined long horizon Asians like last year. That's just what was needed to drive results. And now it's getting better and better to actually drive legal outcomes. And ultimately, I think, What is needed for democratizing of life is From a set of facts, you need to know how much compute, resources you need. to arrive at a outcome. And that happens through both aggregation of and better usage of data. and better usage of actually running the compute. to achieve that outcome. And so we're doing both, right? So like that is why we're laser focused on outcomes. And we want to drive outcomes. And eventually this David versus Goliath thing is happening. Everywhere. You know, it's not just in... in big law or in-house legal, those are just where the budgets are. So they actually can write Basically a blank check. for an interim period of time. and the The forcing function around how much is this actually worth in outcomes to me? is not known. Right. Like, I mean, you know this, like when you go fundraise, you might get hit with like a 100K check. What is the exact outcome of that? You know, like it's very hard to know what the value of this, but that is the only way they can price today. And that's because... they don't know what the expected value of their work actually is. Well, that's the asymmetry as well. It's like, I don't know. It's like, okay, you need to hire a lawyer for your Series C. And what does that mean? Well, it could be like very, very simple. It could be very complicated. So it's very hard to have a flat rate. That's right. Because they can't get burned in the low direction, right? It's like, I charge you $25,000, I spend an hour on this. But they can get burned if it's like, I have to now spend 4,000 hours of attorney time, I can lose money on this, which of course, every single firm, well, actually, as you know, in corporate law, these are loss leader products. And then you try to make it up on the M&A or IPO or something, but obviously a little bit different than plaintiff law. How do you guys charge? How did you think? Because I imagine you had a, it's not like on day one, when you started the business, you knew exactly what you were going to charge. for like what type of customer. I imagine your thinking has evolved on this as well. Absolutely. So I mean, from actually weirdly enough, from day one, We started probably with the... with the, you know, what was it? Tall, grande, venti. Tall, grande, venti. Venti model. That's what we started with because we didn't know any better, right? And then as we dug into it, we realize what they care about. is outcomes. What they care about is how do I serve more clients? So that actually forced us to be outcome-based from very early days. No, we're talking like a year and a half ago, which is pretty funny. That's early in today's world. AI times. Like dog years times 100. Yeah, exactly. Exactly. That is kind of... what happened to us, right? [45:00] There's so much room to grow where now that, We talked about that value diagram of... a third goes to attorneys, half of that is labor. That is half they don't have to spend anymore. So how do you actually turn that into something more useful? And that way, the best way to get more useful work is actually to spend it weirdly on more tokens and more... more work that Eve can do, to make more money for your firm. And that is kind of the cycle we're in today. And a lot of the market dynamics that is helping us is exactly this. We actually believe the billable auto model is the best. I don't think it's going to die completely, but I think it's going to... be on the way of the seed-based pricing model. where for some subset of law, That is probably the best way you can charge for the work because it's very difficult to know value until an outcome many, many years later. But for a lot of it, That's not true anymore. For a lot of it, you actually know what the outcomes are. and you should be able to take your work down to an amortized amount and deliver that. Back. Right. And that is what we are betting on. And this is why we focused on delivering legal outcomes from day one, End to end. And doing that in such a way that it is as simple as a lawyer saying, Here's this client I got in. Give it to Eve. He will pull you in when you're needed. And you're good to go. We've talked about how plaintiff attorneys, they are supply constrained, not demand constrained, or many of them are, because they have this infinite influx of clients that they can't serve because the math doesn't really work in terms of cost versus CAC versus LTV. It's the old-fashioned thing for any business. It applies to them as well. Where are your constraints? What holds you back from growing faster? You're obviously growing incredibly fast. What are your rate limiters? What do you need as You know, it's really fascinating because a lot of the legal benchmarks, like even benchmarks that you share publicly that Anthropic, OpenAI, many of whom we're working very closely with are working on, care a lot about like almost single outcome, right? Like here's a task, Did it kind of do mostly okay? And the world is now moving into agentic work. For us, the only thing that ever mattered was the authentic outcomes. Right. So like now that is getting more and more. The end outcome here is, Today, we're still at the point with technology and processes where there's still a decent amount of human touches needed for a case. But by doing more in our work, we've been drastically reducing that month over month. And in our industry, we need to do that per case type. So for every single case, We're learning that like a science. Like how do you actually resolve this case? And how do you build a mini attorney brain that is capable of crunching all these data points to that resolution and driving it across the board and working with all the tools they have access to including Claude and whatnot, to drive that outcome. And that is a really hard problem. So we're largely limited by engineers at this point. We have... a limitless number of things to do to drive this forward. And the other fun part is we have demand. The moment we are able to deliver on, let's say, moving vehicle accident cases end to end. the amount of value that gets unlocked for the world from that, is in the hundreds of billions of dollars. And that's just one subtype. Actually, that's a great question. Have you had a fully autonomous end-to-end case? It's like answer phone, onboard client, get check from Geico. Has that happened end-to-end? Not yet, but we're close. That is why we need the engineers. So to put things in perspective, a single case... has probably 10,000 back and forths. 10,000, something like that. Because here's how you think about it, right? A client comes in, let's say it's a PI case. Now you have medical records. So now you have to actually outbound call insurance companies and healthcare companies to get hold of the records that they the clients already had. Right. So like that is actually that takes two weeks. today. And with EVE Outbound Agents, where we actually have voice AI that can do the outbound call, and funnily enough, that's an area where we're already seeing agents talking to other agents and learning how to get past each other, which is hilarious. And that is like one step that's needed, right? And then from there, you have to then create a document and actually have a conversation with the other side. So all of this are task we are completing end to end. So today there's a little bit of touch required where Someone makes an outcome better or... does something a little better. But exactly what we're building right now is how do you actually drive this whole thing end-to-end? And we've We are seeing early signs of that where it's actually live in some customers. And have you seen, going back to the defense side and their colossus of information, have you seen their [50:00] reaction just in general change. So as an example of that, it's like, uh-oh, these guys contacted me with this. I always need to settle because they probably have pretty firm rules about this. They don't just randomly make it up as you go, I would assume, because they're dealing with this tens of thousands of times, if not more per year. So if I'm Geico, it's like, oh, you demanded this much. Well, my first attempt is going to settle here. Yeah. But if their costs go down, maybe that changes, right? So have you seen any change in reaction of your, you know, not your counterparty per se, but of your clients' counterparties? - I would say largely still, The defense side very much still works on hourly performance. business model. So they haven't quite caught up to the same pace as our attorney. So like there's a lot of, I guess, pricing asymmetry here. What defense used to do for this one law firm that we're working with is they'll wait a They push out discovery request dates. until a day before Thanksgiving. or a major holiday, and then send discovery there, because they would know the person would go on vacation, and be stuck with all this workload over that vacation, and they wouldn't do a good job at it or would rush it. And now what they use, because it's such a repeated tactic by them, they created and agent on EVE to respond the same day. So next time this came through, like they responded immediately with the responses and the defense were like, what is going on? The tactics don't. And like we're seeing there's so many variants of that that is happening now, which is causing... defense to realize, wait, this is what happens when the other side has resources. Oh, crap. And that's what's fun to see. Well, Jay, this is a fantastic conversation. I learned a lot. Didn't realize there were 10,000 touches every time you went through a case, but really fascinating and delighted to be your partner and bring Eve to the next 10,000 firms. Thank you. I mean, we still have so much more to go. We probably helped a couple hundred thousand clients a day, and we have hundreds of millions life every day. So if that's of interest, please come join Eve. Awesome. Thank you. next episode. As a reminder, the content here is for informational purposes only, should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see A16Z.com forward slash disclosures. Thank you.