Jackson Walker Codifies Partner Judgment Into AI Playbooks To Win Deals

From Prompts to Playbooks: Turning Your Best Lawyers' Judgment into AI Workflows With Jackson Walker — Ai4

Ai4

Standardizing partner redlines into structured AI playbooks delivers faster turnarounds and converts ad-hoc prompting into repeatable, firm-wide revenue.

AI-assisted summary of From Prompts to Playbooks: Turning Your Best Lawyers' Judgment into AI Workflows With Jackson Walker — original source. Check the original for full context.

Source published: 2026-08-17

Detailed notes

## Background This video was published on YouTube by the channel Ai4 on August 17, 2026. It features Greg Lambert, Chief Innovation Officer at Jackson Walker LLP, detailing how the firm moves beyond basic AI prompting by codifying partner expertise into scalable playbooks, automated workflows, and integrated DMS systems.

## TL;DR Greg Lambert presents Jackson Walker's strategic four-step framework—extract, encode, test, and scale—to transform individual legal wins into firm-wide automated AI workflows. Highlighting real-world applications, he explains how the firm turned 900 easement reviews into $200,000 of new business in 72 hours and created a 49-rule commercial real estate playbook from historical redlines. He emphasizes replacing vanity usage metrics with tangible financial and speed outcomes, while combining legal-specific AI with firm document management to eliminate internal knowledge silos.

## Atmosphere & style The presentation is pragmatic, authoritative, and deeply grounded in real-world law firm operations rather than theoretical tech hype. Lambert communicates with clear executive polish, offering an engaging insider's view that balances technical architecture with practical legal culture. Listening to the talk feels like a masterclass in law firm innovation, providing concrete operational playbooks while remaining realistically focused on risk, governance, and business value.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=3SPjBQbshII&t=4s) Jackson Walker captured over $200,000 in new legal revenue by using a Harvey database to review 900 oil and gas easement assessments within a tight 72-hour weekend deadline. - [02:35](https://www.youtube.com/watch?v=3SPjBQbshII&t=155s) Individual successes must be codified into a repeatable four-step framework—Extract, Encode, Test, and Scale—to prevent reliance on sporadic employee initiative. - [04:11](https://www.youtube.com/watch?v=3SPjBQbshII&t=251s) The firm operates a practical tech stack using Microsoft 365 Copilot, Claude Enterprise, Harvey, and specialized databases like Westlaw Advantage and Lexis Protégé to avoid legal hallucinations. - [07:28](https://www.youtube.com/watch?v=3SPjBQbshII&t=448s) Jackson Walker achieved an 84% weekly active user rate across 600+ Harvey licenses through organic peer adoption and practice group champions rather than top-down mandates. - [09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) Partner judgment can be effectively extracted using the 'Red Pen' method, leveraging 20 to 30 years of historical redlines stored in Document Management Systems. - [11:46](https://www.youtube.com/watch?v=3SPjBQbshII&t=706s) A commercial real estate partner's historical redlines were transformed into a 49-rule plain-language Markdown playbook integrated into Microsoft Word, which also serves to train junior associates. - [16:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=960s) Combining AI tools with document management systems like iManage and NetDocuments helps eliminate firm-wide 'Pardon the Interruption' (PTI) email broadcasts and resolves ~95% of internal knowledge inquiries. - [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) Firm governance must focus on cost discipline and measure tangible outcomes—such as captured revenue and matter velocity—rather than relying on vanity metrics like login counts. - [19:13](https://www.youtube.com/watch?v=3SPjBQbshII&t=1153s) The future of legal tech relies on building system-driven, intuitive interfaces around existing lawyer habits rather than expecting partners to become expert prompt engineers.

## Detailed walkthrough Here is a detailed, faithful walkthrough of the transcript:

[00:04](https://www.youtube.com/watch?v=3SPjBQbshII&t=4s) The $200,000 Easement Review: A Weekend Turnaround Story

Earlier in the year, an oil and gas client presented Jackson Walker with a time-sensitive crisis on a matter where the firm was co-counsel alongside a much larger law firm. The client possessed 900 easement assessments that required immediate evaluation to determine which ones demanded further attention. Under normal circumstances, a task of this scale would naturally be assigned to the larger co-counsel. However, the client imposed an aggressive 72-hour deadline: the full analysis had to be delivered by Monday morning, and the request came in on a Friday at midday. When pitched to the larger firm, they offered the standard response, stating that evaluating and assessing that volume of easement documents typically takes weeks or days.

Rather than turning the work away, an enterprising attorney at Jackson Walker stepped up with a novel approach. She asked to take custody of all 900 assessments to determine if she could execute the project alongside a small team. Returning to the office, she ingested the 900 documents into a Harvey database. Over the course of the weekend, the team evaluated every assessment, ordered them by level of importance, identified critical underlying issues, and presented a comprehensive analysis to the client before the Monday morning deadline. This rapid turnaround secured over $200,000 in new legal work—revenue that ordinarily would have gone entirely to the larger law firm.

- **Client Industry:** Oil and gas - **Volume & Scope:** 900 easement assessments - **Turnaround Deadline:** 72 hours (Friday midday to Monday morning) - **Primary Tool Used:** Harvey database - **Financial Impact:** Captured ~$200,000 in legal work previously out of reach

---

[02:35](https://www.youtube.com/watch?v=3SPjBQbshII&t=155s) From Heroic Luck to Firm-Wide Method

While the 72-hour easement success was a major victory, it highlights a fundamental operational risk: the outcome relied entirely on one specific attorney happening to be in the right place, possessing the right technical skill set, and knowing how to pitch a novel workflow under tight time constraints. Relying on individual initiative is essentially "luck with good branding." A single high-value win remains an isolated event unless it is codified into a structured, firm-wide practice.

The primary objective of the Chief Innovation Officer role at Jackson Walker is to transform these extraordinary wins into predictable, standard operations. Rather than relying on sporadic individual heroism, the firm employs a four-step framework to convert partner expertise into scalable AI workflows across practice groups:

``` +----------------------------------------------------------------------+ | 1. EXTRACT: Uncover partner patterns from historical redlines & DMS | +----------------------------------------------------------------------+ | v +----------------------------------------------------------------------+ | 2. ENCODE: Translate patterns into playbooks, rules & workflows | +----------------------------------------------------------------------+ | v +----------------------------------------------------------------------+ | 3. TEST: Validate rules against live matters with partner review | +----------------------------------------------------------------------+ | v +----------------------------------------------------------------------+ | 4. SCALE: Expand horizontally to other lawyers and practice groups | +----------------------------------------------------------------------+ ```

---

[04:11](https://www.youtube.com/watch?v=3SPjBQbshII&t=251s) Firm Context and Technology Architecture

Jackson Walker is a full-service, Texas-only law firm employing approximately 550 attorneys supported by 380 staff members, handling both transactional and litigation practice areas. The innovation division focuses on teaching attorneys how to utilize emerging technology effectively within their daily practice. Rather than overcomplicating their technical stack, the firm relies on a practical combination of enterprise tools and specialized legal platforms:

``` +----------------------------------------------------------------------+ | ENTERPRISE AI | | Microsoft 365 Copilot (M365 platform: OneDrive, Outlook, Word) | | Claude Enterprise (Agentic workflow pilot - 35 technical users) | +----------------------------------------------------------------------+ | v +----------------------------------------------------------------------+ | LEGAL-SPECIFIC AI | | Harvey (Firm-wide drafting, summarization, vault workflows) | | Westlaw Advantage & Lexis Protégé (Primary legal research) | +----------------------------------------------------------------------+ | v +----------------------------------------------------------------------+ | DOCUMENT STORAGE | | Document Management Systems (iManage / NetDocuments DMS) | +----------------------------------------------------------------------+ ```

While general platforms like Copilot and Harvey handle daily document generation and summaries, legal-specific databases like Westlaw Advantage and Lexis Protégé remain non-negotiable. General AI models lack the curated corpus of authoritative primary law required for litigation. Relying solely on general AI models risks legal misstatements or hallucinated case citations that could land the firm on *Above the Law* or in *The New York Times*.

---

[07:28](https://www.youtube.com/watch?v=3SPjBQbshII&t=448s) Driving Adoption Without Mandates

Jackson Walker deliberately avoided imposing top-down mandates requiring attorneys to adopt AI tools. Instead of forcing compliance, the firm scaled its deployment organically from an initial batch of 100 Harvey licenses up to over 600 active users. The adoption strategy centered on identifying individual champions within each practice group, surfacing their specific everyday pain points, and co-creating tailored AI solutions to address those challenges. As successful results spread, organic peer interest and FOMO (fear of missing out) drove widespread firm-wide demand.

``` [Initial Pilot: 100 Licenses] ──> [Practice Champions] ──> [600+ Active Licenses] │ ▼ 84% Weekly Active Rate ```

Although an 84% weekly active user rate demonstrates high usage, active utilization can be a vanity metric if attorneys only apply basic prompting to isolated tasks. Prompt engineering has natural limits; even a "69-page prompt" cannot match a structured workflow. Lasting efficiency requires moving beyond individual ad-hoc prompting toward standardized system playbooks.

---

[09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) Extracting Partner Judgment: The "Red Pen" Method

Lawyers perform significantly better when handed a red pen to critique an existing draft than when asked to create content on a blank page with a blue pen. Handing an experienced practitioner a pre-drafted document triggers their deep expertise, prompting an immediate critique based on decades of practice. Consequently, law firms do not need to invent new knowledge playbooks from scratch—the playbooks already exist inside historical work product stored in closed deal files and Document Management Systems (DMS).

Senior partners possess 20 to 30 years of accumulated negotiating judgment recorded within historical redlines. These files reveal their baseline positions, acceptable fallback positions, and non-negotiable walk-away terms. AI tools provide a cost-effective mechanism to extract these implicit patterns and transform them into explicit, structured playbooks.

``` +----------------------------------------------------------------------+ | HISTORICAL DATA | | 20-30 Years of Partner Redlines & Closed DMS Files | +----------------------------------------------------------------------+ │ ▼ +----------------------------------------------------------------------+ | AI EXTRACTION | | Surface Recurring Positions, Fallbacks & Walk-Away Terms | +----------------------------------------------------------------------+ │ ▼ +----------------------------------------------------------------------+ | HUMAN VERIFICATION | | Partner Uses "Red Pen" to Refine & Confirm 49 Rule Markdown Playbook | +----------------------------------------------------------------------+ ```

---

[11:46](https://www.youtube.com/watch?v=3SPjBQbshII&t=706s) Case Study: The 49-Rule Commercial Real Estate Playbook

To test this methodology, Jackson Walker partnered with a finance attorney specializing in borrower-side commercial real estate loan agreements. The team extracted data from six historical sets of redlines from the partner's prior deals. The analysis revealed that negotiation concentrated heavily across seven specific articles within standard loan agreements, yielding approximately 90 recurring legal positions.

The innovation team presented these 90 positions to the partner to filter out deal-specific anomalies and isolate core negotiating standards. This review yielded a finalized 49-rule playbook written in clean, plain-language Markdown files. Integrated directly into Microsoft Word, the playbook evaluates incoming loan agreements, flagging whether specific clauses meet firm standards, fall within acceptable fallback bounds, or exceed risk tolerance.

``` +----------------------------------------------------------------------+ | INPUT: 6 Sets of Historical Commercial Real Estate Loan Redlines | +----------------------------------------------------------------------+ │ ▼ +----------------------------------------------------------------------+ | ISOLATE: 7 Frequently Negotiated Articles (~90 Recurring Positions) | +----------------------------------------------------------------------+ │ ▼ +----------------------------------------------------------------------+ | FILTER: Partner Eliminates One-Offs & Sets Baseline Standards | +----------------------------------------------------------------------+ │ ▼ +----------------------------------------------------------------------+ | OUTPUT: 49-Rule Plain-Language Markdown Playbook in MS Word | +----------------------------------------------------------------------+ ```

A major secondary benefit of this plain-language Markdown format is associate training. While senior attorneys rarely have time to provide hands-on mentorship, these codified playbooks allow junior associates to study a partner's exact legal judgment and decision-making logic before drafting.

- **Practice Area:** Borrower-side commercial real estate finance - **Source Material:** 6 historical deal redline sets - **Negotiation Scope:** 7 key contract articles - **Initial Extracted Positions:** ~90 recurring positions - **Finalized Rule Count:** 49 structured rules - **Integration Platform:** Native Microsoft Word interface

---

[14:43](https://www.youtube.com/watch?v=3SPjBQbshII&t=883s) Replicating Workflows Across Practice Groups

The partner pattern extraction methodology is repeatable across different legal disciplines. For example, when working with a commercial contracts partner whose colleagues frequently wished they could "clone" them, the innovation team built two automated drafting workflows derived directly from the partner's historical redlines and work product. Once encoded, the partner reviewed and signed off on the system logic, establishing an official baseline workflow for associates across the practice group.

``` +----------------------------------------------------------------------+ | REPEATABLE PLAYBOOK CREATION PROCESS | | | | [Partner Redlines] ──> [AI Pattern Extraction] ──> [Draft Workflow] | | │ | | [Associate Guidance] <── [Partner Review/Approval] <───────┘ | +----------------------------------------------------------------------+ ```

---

[16:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=960s) Merging AI with Document Management & Retiring "PTI" Emails

Integrating AI tools directly with Document Management Systems (DMS) like iManage and NetDocuments allows law firms to search historical firm knowledge while enforcing established ethical walls and security permissions. This capability helps solve the problem of internal firm-wide broadcast emails, known as "Pardon the Interruption" (PTI) messages, where attorneys poll the firm asking if anyone has appeared before a specific judge or handled a niche legal issue.

By routing these queries through an AI-enabled DMS interface, the system automatically surfaces historical work product and identifies relevant internal experts, successfully resolving approximately 95% of inquiries. In one instance involving a serial *pro se* litigant in California, an attorney sent out a PTI email seeking prior firm experience. Within five minutes, the AI system generated a summary report identifying four Jackson Walker attorneys who had litigated against the individual, including an internal strategic strategy memo detailing how to counter the plaintiff's filing tactics.

- **System Integrations:** iManage and NetDocuments DMS - **Target Overhead:** Elimination of firm-wide "Pardon the Interruption" (PTI) email broadcasts - **Query Resolution Rate:** ~95% automated answer retrieval - **Case Example:** Identified 4 experienced attorneys and a counter-strategy memo on a serial *pro se* litigant within 5 minutes

---

[18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) Governance, Cost Discipline, and Metrics That Matter

Scaling AI across a law firm requires governance, cost discipline, and practical performance tracking. Token usage costs require active oversight to ensure the firm does not spend "$10,000 to deliver $1,000 worth of value." Furthermore, firms must look beyond high active-user percentages and measure tangible operational metrics:

---

[19:13](https://www.youtube.com/watch?v=3SPjBQbshII&t=1153s) The Future of System-Driven AI Workflows

Most law firm partners will never become expert prompt engineers. Successful innovation strategies must accept this reality and design intuitive, system-driven interfaces around existing partner habits. In the next phase of legal tech, a partner opening a new M&A transaction will interact with a familiar interface that automatically prompts them for required deal parameters, surfaces necessary contract documentation, and maintains consistent output quality across matters.

Attorneys should actively discuss these AI workflows with clients. While clients expect law firms to adopt efficiency-enhancing tools, they are often unfamiliar with the specific mechanics. Demonstrating technological efficiency ultimately expands business opportunities—as shown by the 900 easement review project, which brought in work the firm otherwise would not have captured.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=3SPjBQbshII&t=4s) evaluating and assessing that volume of easement documents typically takes weeks or days

> [02:35](https://www.youtube.com/watch?v=3SPjBQbshII&t=155s) luck with good branding — Greg Lambert

> [07:28](https://www.youtube.com/watch?v=3SPjBQbshII&t=448s) 69-page prompt — Greg Lambert

> [09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) red pen — Greg Lambert

> [09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) blue pen — Greg Lambert

> [14:43](https://www.youtube.com/watch?v=3SPjBQbshII&t=883s) clone

> [16:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=960s) Pardon the Interruption

> [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) $10,000 to deliver $1,000 worth of value — Greg Lambert

## Mentioned Greg Lambert, Jackson Walker LLP, Ai4, Harvey, Microsoft 365 Copilot, Claude Enterprise, Westlaw Advantage, Lexis Protégé, iManage, NetDocuments, Above the Law, The New York Times, Microsoft Word

## Video description Greg Lambert - Chief Innovation Officer, Jackson Walker LLP

Most law firms have bought the AI tools; far fewer have turned them into repeatable process efficiency. Jackson Walker Chief Innovation Officer Greg Lambert shares the firm's working method: extract an experienced partner's patterns from their own executed matters and redlines, encode them into AI playbooks and workflows, test against live deals, and scale across the firm. He'll walk through real results, including hundreds of easement document reviews delivered in 72 hours that captured hundreds of thousands in new work, a 49-rule loan agreement playbook built from one partner's redline history, and an AI-plus-DMS workflow that is quietly retiring the firm-wide "does anyone know\..." email. The talk covers what makes it work underneath: governance, cost discipline, and measuring outcomes in dollars rather than self-reported minutes saved. Attendees leave with a concrete, replicable model for moving from individual prompting to firm-wide process.

[Source](https://www.youtube.com/watch?v=3SPjBQbshII)

中文简报

Jackson Walker将合伙人判断固化为AI实战手册以赢得业务将合伙人的修改批注标准化为结构化的AI操作手册，不仅能加快交付速度，还能将零散的提示词应用转化为全所可复制的持续收入。 Greg Lambert介绍了Jackson Walker的四步战略框架——提取、编码、测试与规模化，旨在将个人的法律业务成果转化为全所通用的自动化AI工作流。结合实际应用案例，他阐述了该律所如何在72小时内完成900份地役权审查并带来20万美元的新增业务，以及如何根据历史修改痕迹制定出包含49条规则的商业地产实战手册。他强调应摒弃流于形式的使用量指标，转而关注切实可行的财务与时效成果，同时将法律专用AI与律所文档管理系统相结合，以消除内部知识孤岛。 Jackson Walker 律师事务所在周末 72 小时内利用 Harvey 数据库完成了 900 份石油与天然气地役权评估文件的审查，成功获取了此前难以接下的逾 20 万美元新业务创收。 (0:04) 为避免依赖偶然的个人英雄主义，该律所构建了“提取、编码、测试、规模化”（Extract, Encode, Test, Scale）四步框架，将合伙人的资深判断力沉淀为可复用的 AI 工作流。 (2:35) 律所运用“红笔法”挖掘合伙人长达数十年积累的历史修改记录（redlines），提炼出一套集成在 Word 中的 49 条规则 Markdown 实务手册，兼具辅助起草与青年律师培训功能。 (11:46) 通过将 AI 工具与 iManage 及 NetDocuments 等文档管理系统打通，律所解决了约 95% 的内部专业经验问询，逐步淘汰了全所性的“打扰邮件”（PTI emails）。 (16:00)

Reading summary

Jackson Walker's innovation leadership argues that law firms must move beyond individual, ad-hoc prompt engineering by systematically extracting partner judgment from historical work product into repeatable AI playbooks. Relying on isolated attorney initiatives represents an operational vulnerability, whereas institutionalizing expertise creates durable turnaround and revenue advantages. From Heroic Wins to Systematic Codification (0:04) Jackson Walker demonstrated the commercial viability of AI when a single attorney used a Harvey database over a 72-hour weekend to analyze 900 oil and gas easement assessments. Meeting an aggressive Monday morning deadline that a larger co-counsel declined enabled the firm to capture over $200,000 in new work. However, firm leadership categorizes one-off feats as "luck with good branding" because they depend entirely on an individual happening to possess the right technical aptitude and initiative under pressure. To convert sporadic heroics into predictable operational capacity, the firm instituted a four-step framework: extract, encode, test, and scale. The methodology uncovers partner judgment patterns from historical Document Management System (DMS) records, translates them into structured rules and workflows, validates them alongside partners on live matters, and deploys them across practice groups. The Red-Pen Method and Playbook Creation (9:54) Rather than asking attorneys to draft guidelines on a blank page, Jackson Walker relies on the principle that lawyers critique existing text far more effectively than they create from scratch. Decades of negotiating judgment—including primary positions, fallbacks, and walk-away thresholds—are already recorded inside closed deal files and historical redlines. AI tools provide the mechanism to surface these latent patterns cost-effectively. In a commercial real estate pilot, the innovation team analyzed six historical deal redlines from a borrower-side finance partner, identifying 90 recurring positions across seven standard contract articles. The partner reviewed the extraction with a "red pen" to remove deal-specific outliers, yielding a 49-rule plain-language Markdown playbook integrated directly into Microsoft Word. The tool checks incoming loan agreements against firm standards while simultaneously offering junior associates a transparent reference for partner-level decision logic. Enterprise System Integration and Institutional Search (16:00) Scaling legal AI requires anchoring tools within firm repositories like iManage and NetDocuments while enforcing ethical walls. Jackson Walker combines enterprise platforms like Microsoft 365 Copilot and a 35-user Claude Enterprise pilot with domain-specific tools like Harvey, Westlaw Advantage, and Lexis Protégé. Leadership emphasizes that specialized legal databases remain mandatory for litigation to avoid hallucinated citations or erroneous legal statements that general models risk generating. Connecting AI directly to the DMS also targets internal firm broadcasts, known as "Pardon the Interruption" (PTI) emails, where lawyers poll colleagues for niche background knowledge. The integrated system resolves approximately 95% of these inquiries automatically. In one instance, an automated query regarding a California serial pro se litigant identified four experienced firm attorneys and surfaced an internal counter-strategy memo within five minutes. Organic User Adoption and Cost Governance (7:28) Jackson Walker scaled its AI deployment from 100 initial Harvey licenses to more than 600 active users without top-down mandates. The firm identified practice-group champions, addressed specific daily bottlenecks, and let peer interest drive an 84% weekly active user rate. However, leadership notes that high usage and prompt generation can become vanity metrics, asserting that even extensive multi-page prompts cannot match structured system workflows. Sustainable scaling also demands financial governance to prevent disproportionate compute expenses, such as spending $10,000 to deliver $1,000 in value. The firm evaluates AI initiatives using business-driven metrics—captured revenue, turnaround velocity, and matter speed—rather than license counts or raw prompt volume. Jackson Walker concludes that because most partners will not become expert prompt engineers, future systems must be embedded seamlessly into everyday deal intake and drafting interfaces. While demonstrating these operational efficiencies to clients can win new business, long-term success depends on cost discipline and avoiding reliance on general models for primary legal research.

精炼阅读摘要

得克萨斯州 Jackson Walker 律师事务所提出，律所仅依赖个别律师的个人技能属于“包装精美的运气”，应当通过萃取历史修改记录中的隐性经验，将合伙人的专业判断转化为系统化的 AI 规则手册与标准工作流。告别偶发英雄主义：四步法沉淀标准流程 (0:04) Jackson Walker 此前经历过一次典型的应急委托：一家油气行业客户面临 900 份地役权评估审查，要求在周五中午至周一早间的 72 小时内完成。由于体量庞大，共同代理的大型律所按惯例表示需要数天至数周因而无法承接。该所一名律师主动带领小团队将 900 份文件导入 Harvey 数据库，在周末完成逐份评估与优先级排序，并在周一截止日前交付了全面分析。这次快速交付不仅按时解决了客户危机，还为律所锁定了原本流向大所的逾 20 万美元新增业务。然而创新团队指出，这种成果完全取决于特定律师恰好掌握对应技术技能并在紧迫时限内提出方案，本质上只是偶发性的“英雄主义”。若不能制度化，高价值案例就仅是孤立事件。为此，该所创新总监确立了四步走框架，用于跨业务组规模化转化经验：第一步“提取”（从历史修改稿与文档系统中挖掘合伙人模式），第二步“编码”（将规律转化为规则与手册），第三步“测试”（在真实案件中由合伙人复核验证），第四步“扩展”（向其他律师与业务组横向推广）。红笔效应与49条规则：萃取合伙人裁量权 (9:54) 实践表明，律师在面对空白页时难以凭空建构规则，但在手握“红笔”批注修改现有文本时能迅速调动深层经验。高级合伙人 20 至 30 年的谈判裁量权早已沉淀在历史交易修改稿中，清晰记录着其基准立场、可接受的折中方案以及不可退让的底线条款。利用 AI 工具提取这些历史规律，能够以较低成本将合伙人的隐性裁量权转化为显性结构化规则。在商业不动产借款人融资业务的落地测试中，团队调取了一位合伙人过往交易的 6 组修改记录，分析显示谈判高度集中在标准借款合同的 7 个核心条款中，共提炼出约 90 个常见立场。经合伙人剔除偶发特例后，最终确立了包含 49 条规则的纯文本 Markdown 规则手册，并直接集成在 Microsoft Word 中运行。该手册能自动审查新进合同并标记条款合规状态。此举不仅实现了规则在商业合同等领域的跨组复用，还使缺乏资深导师亲自指导的初级律师得以在起草前学习合伙人的决策逻辑。实用技术架构与无强制自下而上推广 (4:11) Jackson Walker 拥有约 550 名律师与 380 名员工，其技术架构采取企业级工具与法律垂直平台结合的务实路线。企业层部署 Microsoft 365 Copilot 及面向 35 名技术用户的 Claude Enterprise 智能体试点；专业层由 Harvey 承担全所起草、摘要与文件库工作流，并指定 Westlaw Advantage 与 Lexis Protégé 负责权威判例检索。创新团队强调，通用模型缺乏权威判例库支持，单靠通用工具检索极易出现虚假引用风险。在全所推广上，律所刻意避开自上而下的行政摊派，而是从最初 100 个 Harvey 许可起步，通过在各业务组挖掘骨干人员、针对日常痛点定制解决方案，借助同行示范效应逐步扩展至 600 多名活跃用户，实现 84% 的周活跃率。创新团队指出，提示词工程存在天然上限，再冗长的提示词也无法取代系统化工作流，律所不能将活跃度这一虚荣指标误当作真正的生产力提升。融合文档系统与以增收为核心的治理度量 (16:00) 通过将 AI 工具与 iManage 及 NetDocuments 等文档管理系统（DMS）深度整合，律所在严格遵循利益冲突隔离墙与安全权限的前提下盘活了历史知识。这直接替代了以往全所群发询问特定法官或冷门议题的打扰邮件（PTI 邮件），知识检索的自动解决率达到约 95%。例如在加利福尼亚州一起涉及连续起诉原告（serial pro se litigant）的案件中，系统在 5 分钟内即检索到 4 位曾与该原告交手的所内律师，并调取了应对其诉讼策略的内部备忘录。在运营治理层面，律所强调算力成本管控，防止出现“花费 10,000 美元仅产出 1,000 美元价值”的资源浪费。律所主张摒弃软件许可数、登录频次和提示词生成量等浅层指标，转而关注实质性的运营与商业成效，将核心评估标准锁定在增收规模、交付周转速度以及案件推进效率上。该所创新团队总结指出，绝大多数合伙人不会成为提示词工程师，法律科技应顺应合伙人现有工作习惯构建系统驱动型界面；律所主动向客户展现技术效率不仅契合客户预期，更能凭借超常规的交付速度直接捕获传统模式下无法承接的高价值业务。

Detailed notes — zh

## Background 该视频由 Ai4 于 2026 年 8 月 17 日发布在 YouTube 平台上，内容为 Jackson Walker 律所首席创新官 Greg Lambert 的主题演讲。根据页面介绍，演讲聚焦于律所如何跳出单纯购买工具与个人零散 Prompt 的局限，通过提取资深合伙人的历史工作成果与红线修改模式，将其沉淀为可复用的 AI Playbook 与工作流，并在治理、成本控制及增量创收层面建立标准化的系统落地方法论。 ## TL;DR 演讲通过分享 Jackson Walker 律所借助 Harvey 在 72 小时内完成 900 份地役权审查并斩获 20 万美元新增收入的实战，引出了将偶然胜利转化为标准化流程的“提取-编码-验证-扩展”四步法。律所立足“律师执红笔优于执蓝笔”的核心洞见，从合伙人过往真实红线记录中提炼出包含 49 条谈判规则的商业房地产贷款协议 Playbook，并无缝集成进日常审查环境以实现合伙人智慧的高效传承。此外，律所将 AI 与底层文档管理系统（DMS）深度融合，终结了低效的全员广播求助邮件，实现了全所知识资产的秒级调取。Greg Lambert 最后强调，律所必须摆脱对复杂 Prompt 的依赖并严控 Token 成本，以实际创收和交付效率等商业指标为导向，用极致的效率在存量竞争中把业务蛋糕做大。 ## Atmosphere & style 整场演讲风格务实冷峻、干货密集且极具行业穿透力，完全剥离了泛泛而谈的 AI 概念狂热。Greg Lambert 以富有见地的行业洞察娓娓道来，用“红笔与蓝笔之争”、“让胜利看起来极度平淡无奇”等辛辣而生动的比喻直击律所转型的深层痛点。分享在展示亮眼业务战果的同时，始终保持着对虚荣指标与成本陷阱的清醒克制，令听众不仅能感受到前沿技术颠覆传统法律服务的实战冲击，更能体会到一套深思熟虑、兼顾合规与商业逻辑的落地系统正静水流深地运转。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=3SPjBQbshII&t=4s) 借助 Harvey 在 72 小时内完成 900 份油气地役权评估，打破传统大所数周人工审查的周期限制，为律所直接斩获约 20 万美元新增业务收入。 - [03:12](https://www.youtube.com/watch?v=3SPjBQbshII&t=192s) 单次依赖特定律师个人技能的胜利本质上是“包装精美的运气”，首席创新官的核心使命是建立标准化方法，让成功变得平淡无奇且高度可复制。 - [03:12](https://www.youtube.com/watch?v=3SPjBQbshII&t=192s) 建立“提取-编码-验证-扩展”四步工作法，从资深合伙人的历史红线中萃取决策规则并固化为 Playbook，最终横向推广至全所业务组。 - [03:12](https://www.youtube.com/watch?v=3SPjBQbshII&t=192s) 搭建四层 AI 技术底座：底层协同（Copilot）、严防虚构判例的权威法律检索（Westlaw/Lexis+）、主力法律平台（Harvey）以及复杂智能体探索（Claude Enterprise）。 - [07:28](https://www.youtube.com/watch?v=3SPjBQbshII&t=448s) 坚决不搞行政命令式强推，而是深入各业务组挖掘“业务冠军”解决真实痛点，通过内部分享激发的错失恐惧症（FOMO）实现高达 84% 的 Harvey 周活跃度。 - [07:28](https://www.youtube.com/watch?v=3SPjBQbshII&t=448s) 高周活率在一定程度上是虚荣指标，孤立的个人 Prompt 提效存在明显上限，必须从单点个人提示词演进为系统级 Playbook 工作流。 - [09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) 提出“律师拿红笔远比拿蓝笔出色”的执业心理洞见：让律师凭空起草困难，但通过挑刺和审查历史修改痕迹（redlines）能最高效地唤醒沉淀数十年的隐性专业判断。 - [09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) 通过 AI 从 6 套历史修改红线中提炼出包含 49 条谈判规则的商业地产贷款 Playbook，嵌入 Word 环境，成功破解了年轻律师培养与合伙人经验传承的难题。 - [14:37](https://www.youtube.com/watch?v=3SPjBQbshII&t=877s) AI 与律所 DMS（文档管理系统）深度整合并在伦理防火墙合规下秒级调取智慧资产，逐步消灭了群发问询案情经验的“打扰全员（PTI）”广播邮件。 - [14:37](https://www.youtube.com/watch?v=3SPjBQbshII&t=877s) 实战展现快速协同威力：遭遇加州职业原告滥诉时，AI 检索系统在 5 分钟内即精准定位全所 4 位交手律师并调取全套反制策略备忘录。 - [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 建立严谨的 Token 成本纪律与投入产出比考量，坚决避免陷入耗费高昂大模型成本去交付低价值业务的资源陷阱。 - [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 度量 AI 价值必须抛弃自述节省时间等表面指标，重点紧盯业务营收捕获量（revenue capture）、交付周期缩短及交易周转速度等真金白银的商业结果。 - [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 律师合伙人永远不会成为提示词专家，未来应推进系统驱动型 AI（system-driven AI），在律师新建案件时自动匹配预装文档与规则架构以确保输出确定性。 - [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 极致的效率不会缩减传统计费工时，反而能帮助律所从存量市场中敏捷抢夺原本难以承接的海量业务，从而把整个营收蛋糕做大。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=3SPjBQbshII&t=4s) 72小时拿下900份地役权评估：从一次偶然的AI胜利说起今年早些时候，Jackson Walker 律所遇到了一位油气行业客户带来的棘手难题。当时律所正与另一家规模大得多的知名大所共同担任该项目的联合法律顾问（co-counsel）。客户面临的困境是：他们手头积压了整整900份地役权评估（easement assessments）文档，急需有人逐一审查评估，并筛选出究竟哪些条款和文件需要重点关注。通常情况下，这类体量庞大的常规工作会直接移交给那家规模更大的律所处理，但这次最大的障碍在于严苛的时间截点——客户在周五中午提出需求，要求在下周一清晨之前全部交付，留给律师团队的时间仅仅只有72小时。大所给出的反馈完全不出所料：按照传统的人工审查模式，完成这种规模的评估通常需要“耗费数天甚至数周时间”。就在客户一筹莫展之际，Jackson Walker 的一位女律师主动站了出来，提出了一个颠覆常规的大胆设想。她向客户申请调阅这900份评估文件，随后迅速回到办公室，将所有文件完整加载到 Harvey 数据库中，带领一个小团队紧锣密鼓地展开智能化分析与评估。在短短72小时内，这个小团队不仅完成了对全部900份文档的深度评估，还按照重要性与风险等级进行了优先级排序，精准指出了每份评估中的潜在法律问题并呈报给客户。这次闪电战为律所直接带来了约20万美元的业务收入。Greg Lambert 直言不讳地指出，按照以往的行规，这笔业务原本绝不可能落到他们头上，只会被大型律所囊括，正是因为这位极具创新思维的律师不仅具备全新的视角，而且熟练掌握了先进的 AI 工具集，才能够在极短的时间内构思方案、推进执行并交付结果。 - **客户紧急需求**：周五中午交付900份油气地役权评估，要求周一早晨（72小时内）完成筛查。 - **大所传统预估**：此类工作通常需要数天至数周的人工耗时。 - **实际交付与战果**：借助 Harvey 数据库在72小时内完成全部评估、分级与风险标注，直接斩获约200,000美元的新增业务收入。 --- ### [03:12](https://www.youtube.com/watch?v=3SPjBQbshII&t=192s) 把“奇迹”变成标准流程：四步工作法与律所AI技术栈虽然地役权评估的胜利令人振奋，但 Greg Lambert 强调，这也是他今天站在这里的核心原因与痛点所在——这次成功高度依赖于“特定律师在正确的时间、身处正确的地点、恰好具备正确的技能，并提出了正确的交易处理方式”。在本质上，这更像是一种带着优秀品牌包装的“运气”（luck with good branding）。如果不能在成功案例周围建立起一套标准方法论，那么所有的胜利都不过是一次性的偶然事件。作为首席创新官，他的核心使命就是构建系统，“让这种胜利看起来极度平淡无奇（very boring），并且具备高度的可重复性”。整个工作方法提炼为标准四步法：首先从资深律师的历史工作成果中提取模式；其次将这些模式编码为结构化的业务手册（Playbook）和自动化工作流；随后在真实的在办业务中实弹测试并验证有效性；最后横向扩展至整个业务组乃至相关业务领域。 ```text +------------------------------------------------------------------+ | Jackson Walker 知识提炼与落地四步法 | +------------------------------------------------------------------+ [1. 提取 (Extract)] 从资深合伙人历史修改红线/归档案卷中提取决策模式 │ ▼ [2. 编码 (Encode)] 合伙人以“红笔”介入修正，固化为明确的 Playbook 规则 │ ▼ [3. 验证 (Test)] 在日常办公环境（如 Word）中结合在办交易实战检验 │ ▼ [4. 扩展 (Scale)] 横向推广至初级律师、本业务组及其他相关业务团队 +------------------------------------------------------------------+ ``` 为支撑该体系的运转，Jackson Walker 构建了清晰的 AI 技术底座。作为一家拥有约550名律师、深耕德克萨斯州的全业务律所，其技术栈主要包含四大层级： 1. **底层办公与日常交互**：深度绑定微软生态，采用基于 Microsoft 365 平台的 Copilot，无缝连接 OneDrive 与电子邮件系统。 2. **专业法律检索**：采用 Westlaw Advantage 与 Lexis+ AI（Protégé）。Lambert 坦言，一年前法律检索 AI 工具表现极差，但过去一年取得了突破性进展；由于诉讼律师对权威数据语料库（corpus of data）有着刚性合规要求，通用大模型绝不能替代专业检索工具，律所绝不能冒着在《Above the Law》或《纽约时报》上因“虚构判例（hallucinated cases）”而出丑的风险。 3. **主力法律 AI 平台**：Harvey 是律所内部目前最受欢迎的工具，广泛用于日常草拟、摘要生成、文档审查、建立数据保险库（Vaults）以及定制个人工作流。 4. **前沿智能体试点**：引入 Claude Enterprise 试点项目，由约35名技术专家与资深律师组成核心小组，专门探索其他工具无法实现的复杂智能体工作流（agentic workflows），并为其搭建前端交互界面。 --- ### [07:28](https://www.youtube.com/watch?v=3SPjBQbshII&t=448s) 拒绝强制推行：从单点业务冠军到84%周活跃度的落地法则在展示工具使用数据时，Greg Lambert 经常引来同行的侧目：在全所约550名律师及380名行政支持人员中，目前已有超过600人拥有独立的 Harvey 账户权限，更令人瞩目的是其周活跃度高达 84%，且绝大多数用户每周会多次使用该工具。取得这种深层渗透率的秘诀在于律所坚决“跳过了行政强制命令”（skipped the mandate）。他们从未下达过全员必须使用的硬性指标，而是采取了稳扎稳打的试点扩容策略——从最初的100个许可逐步追加至150、200、250个，最终突破600大关。在推进路径上，创新团队深入每一个具体业务组，物色一名具备号召力的“业务冠军”（Champion），精准找出该律师在日常繁重工作中面临的真实痛点，并与之一道探索如何通过 AI 工具实现体验改良。当这位冠军律师尝到甜头并在部门内部自发分享后，团队内部迅速蔓延起“错失恐惧症”（FOMO, fear of missing out），各部门律师开始争先恐后地主动申请工具权限。然而，Lambert 清醒地指出，84%的周活在某种程度上也是一个“虚荣指标”（vanity metric）。活跃使用并不等同于正确使用。目前绝大多数律师仅将 AI 局限在孤立的个人案件或零散场景中，这很快就会触及天花板——哪怕写出长达69页的超级提示词（prompt），个人提效的上限依然十分有限。 --- ### [09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) “红笔优于蓝笔”：从资深合伙人的修改痕迹中提炼49条规则Playbook 针对律师的执业心理，Lambert 提出了一个在所内反复强调的核心洞见：“律师拿红笔时的表现远比拿蓝笔时出色得多（lawyers do better with a red pen than they do with a blue pen）。”如果让律师从零开始凭空创作（拿蓝笔），他们往往无从下手；但如果递给他们一份现成的草案让他们挑刺、修改和批注（拿红笔），他们积淀数十年的专业经验就会像泉水一样喷涌而出，因为这就是他们整个职业生涯赖以建立的肌肉记忆。这意味着构建专业 Playbook 所需的智慧其实早已存在，只是沉睡在律所的文档管理系统（DMS）和历史归档卷宗里。每一位资深合伙人在过去二三十年里积累了无数的红线修改记录（redlines），里面凝结着他们的全部职业判断：在哪些关键条款上寸步不让、在何种情境下可以接受备选方案（fallback language），以及在什么底线被触碰时会坚决掀桌放弃交易（walk away）。AI 工具的核心价值，正在于以合理的成本将这些深藏于历史红线中的隐性知识彻底结构化地提取出来。 Lambert 分享了一个极具代表性的落地案例：团队与一位金融业务合伙人合作，针对借款人商业房地产贷款协议（borrower side commercial real estate loan agreement）构建工作流。团队选取了该合伙人过往处理过的6套完整红线修改记录，通过 AI 提取分析发现，这类协议中每次必谈的核心条款共有7个大项，而在这些交易中合伙人反复采取的经常性立场大约有90个。 ```text [6套过往交易历史修改红线] │ ▼ (AI分析提取) [定位7大核心谈判章节 / 识别90个经常性谈判立场] │ ▼ (交由资深合伙人执“红笔”筛选剔除偶发案例) [提炼固化为 49 条明确规则的 Playbook (Markdown 纯文本)] │ ▼ (集成至 Microsoft Word 环境) [初级律师加载在办新协议 -> 逐条比对输出：符合 / 存疑 / 超出范围] ``` 创新团队将这90个立场整理成清单递交合伙人，让其执“红笔”明确剔除单次交易的偶发案例，最终沉淀出一份包含49条规则的高精度 Playbook。每条规则都完整承载了该合伙人的执业判断基准：标准立场是什么、可接受的备选表述是什么、放弃交易的红线在哪里，并附带真实交易中的示范条款。这套系统以极简的 Markdown 纯文本编写，直接嵌入初级律师熟悉的 Microsoft Word 环境中运行。这项创新还带来了一个意料之外的重大收益：它破解了困扰法律界数十年的“导师带徒弟”困局。初级律师在处理新协议时，只需调出 Playbook，系统就会自动逐条对照49条规则进行筛查，清晰标记出“完全符合”、“存在疑问”或“超出预设范围”的条款。初级律师得以通过 Playbook 直接窥见合伙人严密的思考逻辑，以数字化的形式实现了顶尖合伙人专业判断力的高效传承。 - **案例背景**：借款人商业房地产贷款协议（CRE Loan Agreement）。 - **提炼样本**：精选该合伙人既往6套核心交易的修改红线（redlines）。 - **提炼成果**：从7大必谈章节、90个高频立场中，由合伙人亲笔校准出包含49条明确规则的实战 Playbook。 - **运行载体**：基于 Markdown 格式编写，无缝嵌入初级律师的 Microsoft Word 审查界面。 --- ### [14:37](https://www.youtube.com/watch?v=3SPjBQbshII&t=877s) 复制与扩展：AI深度整合DMS与终结“打扰全员邮件” 这一套“提取-校准-编码”的工作法具有极强的通用性和可复制性。在另一个案例中，创新团队协助一位公认“恨不得被克隆出来”的资深合同合伙人，成功提炼并建立了两个标准起草工作流。通过合伙人亲自审核、签署确认，不仅规范了日常起草标准，还让合伙人与初级律师之间的协同变得前所未有的顺畅。更为重磅的突破发生在 AI 工具与律所底层文档管理系统（DMS，如 iManage 或 NetDocuments）的深度融合。在律所内部，DMS 历来被视为令人又爱又恨的基础设施，但在 AI 加持下，其庞大的沉淀资产被彻底激活。系统严格遵循客户隔离与伦理防火墙（ethical walls）规则，在绝对保障数据安全与权限边界的前提下，赋予律师秒级调取和盘活历史工作成果的能力。这种融合直接重塑了律所内部的知识检索机制，逐步消灭了被称为“抱歉打扰全员（PTI, Pardon the Interruption）”的低效广播邮件。过去律师常群发邮件询问“谁在某郡的某法官手下打过官司”或“谁处理过某类特殊对手”。现在，这类问题会被直接输入与 DMS 联通的 AI 系统中检索，约95%的情况下系统都能瞬间提供精准答案。 Lambert 列举了一个极其生动的实战案例：加州曾出现一名臭名昭著的职业无代表诉讼原告（serial pro se plaintiff），四处恶意滥诉。当业务团队收到相关诉讼并发起内部咨询后，AI 检索系统在短短5分钟之内便精准生成了一份深度报告，清晰指出了全所曾与该原告交手的4位在任律师，并直接调出了其中一位律师此前撰写的应对策略备忘录及全套反制方案，实现了全所智慧资产的秒级调用。 --- ### [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 治理、Token成本控制与未来愿景：用效率把蛋糕做大在演讲的尾声，Greg Lambert 重点阐述了技术落地背后的底层支撑机制——合规治理、成本约束与价值度量： 1. **治理与伦理合规**：所有 AI 交互均需经过严格的日志记录与权限审计，基于 DMS 的伦理防火墙设计隔离机制，确保律师绝无可能越权访问受限信息，且模型交互基于受控的数据快照进行。 2. **Token 成本纪律**：随着大模型消耗的攀升，“Token 已经不再便宜”，开发内部工具必须严守投入产出比，绝不能陷入“花费10,000美元成本去交付价值1,000美元业务”的资源陷阱。 3. **摒弃虚荣指标，度量真金白银**：考核 AI 项目的成败不能停留在“每周84%使用率”这种表面活跃度上，而必须紧紧盯住核心商业指标——捕获的业务营收增量（revenue capture）、案件周转交付周期（turnaround time）以及交易在系统内的流转效率。展望未来，Lambert 明确断言：“绝大多数合伙人永远都不会成为精通写提示词的专家（skilled prompters）。”律所的产品设计必须彻底绕开对复杂 Prompt 的依赖。未来的系统驱动型 AI（system-driven AI）应当是无缝内嵌的：当并购业务合伙人打开一宗新的并购案时，界面会直接提示她输入关键交易细节，系统随后自动生成并调取所需的完整文档架构，以极高的确定性交付一致的专业成果。最后，针对律师行业普遍担忧的“效率提升导致计费工时减少”的疑虑，Lambert 坚定地呼吁律师主动与客户展开深入对话。真实的商业实践表明，极致的效率不仅不会缩减业务，反而能够“把蛋糕做大（grow the pie）”——正如那900份原本属于大型律所的地役权评估案一样，唯有具备突破性的效率与敏捷度，律所才能从激烈的存量竞争中抢夺下前所未有的海量新业务。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=3SPjBQbshII&t=4s) 耗费数天甚至数周时间 — Greg Lambert > [03:12](https://www.youtube.com/watch?v=3SPjBQbshII&t=192s) 特定律师在正确的时间、身处正确的地点、恰好具备正确的技能，并提出了正确的交易处理方式 — Greg Lambert > [03:12](https://www.youtube.com/watch?v=3SPjBQbshII&t=192s) 让这种胜利看起来极度平淡无奇（very boring），并且具备高度的可重复性 — Greg Lambert > [07:28](https://www.youtube.com/watch?v=3SPjBQbshII&t=448s) 跳过了行政强制命令 — Greg Lambert > [09:54](https://www.youtube.com/watch?v=3SPjBQbshII&t=594s) 律师拿红笔时的表现远比拿蓝笔时出色得多（lawyers do better with a red pen than they do with a blue pen） — Greg Lambert > [14:37](https://www.youtube.com/watch?v=3SPjBQbshII&t=877s) 抱歉打扰全员（PTI, Pardon the Interruption） — Greg Lambert > [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) Token 已经不再便宜 — Greg Lambert > [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 花费10,000美元成本去交付价值1,000美元业务 — Greg Lambert > [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 绝大多数合伙人永远都不会成为精通写提示词的专家（skilled prompters） — Greg Lambert > [18:00](https://www.youtube.com/watch?v=3SPjBQbshII&t=1080s) 把蛋糕做大（grow the pie） — Greg Lambert ## Mentioned Greg Lambert, Jackson Walker LLP, Ai4, Harvey, Microsoft 365, Copilot, Westlaw Advantage, Lexis+ AI, Claude Enterprise, iManage, NetDocuments, OneDrive, Above the Law, The New York Times ## Video description Greg Lambert - Chief Innovation Officer, Jackson Walker LLP Most law firms have bought the AI tools; far fewer have turned them into repeatable process efficiency. Jackson Walker Chief Innovation Officer Greg Lambert shares the firm's working method: extract an experienced partner's patterns from their own executed matters and redlines, encode them into AI playbooks and workflows, test against live deals, and scale across the firm. He'll walk through real results, including hundreds of easement document reviews delivered in 72 hours that captured hundreds of thousands in new work, a 49-rule loan agreement playbook built from one partner's redline history, and an AI-plus-DMS workflow that is quietly retiring the firm-wide "does anyone know\..." email. The talk covers what makes it work underneath: governance, cost discipline, and measuring outcomes in dollars rather than self-reported minutes saved. Attendees leave with a concrete, replicable model for moving from individual prompting to firm-wide process. [Source](https://www.youtube.com/watch?v=3SPjBQbshII)

Source quotes

[0:04] Greg Lambert: 耗费数天甚至数周时间 [3:12] Greg Lambert: 特定律师在正确的时间、身处正确的地点、恰好具备正确的技能，并提出了正确的交易处理方式 [3:12] Greg Lambert: 让这种胜利看起来极度平淡无奇（very boring），并且具备高度的可重复性 [7:28] Greg Lambert: 跳过了行政强制命令 [9:54] Greg Lambert: 律师拿红笔时的表现远比拿蓝笔时出色得多（lawyers do better with a red pen than they do with a blue pen） [14:37] Greg Lambert: 抱歉打扰全员（PTI, Pardon the Interruption） [18:00] Greg Lambert: Token 已经不再便宜 [18:00] Greg Lambert: 花费10,000美元成本去交付价值1,000美元业务 [18:00] Greg Lambert: 绝大多数合伙人永远都不会成为精通写提示词的专家（skilled prompters） [18:00] Greg Lambert: 把蛋糕做大（grow the pie）

Source quotes

[0:04] evaluating and assessing that volume of easement documents typically takes weeks or days [2:35] Greg Lambert: luck with good branding [7:28] Greg Lambert: 69-page prompt [9:54] Greg Lambert: red pen [9:54] Greg Lambert: blue pen [14:43] clone [16:00] Pardon the Interruption [18:00] Greg Lambert: $10,000 to deliver $1,000 worth of value

Original transcript

[0:04] So earlier this year, we had an oil and gas client that had come to us with a problem. It [0:11] was on a matter that we were co-consul on with a much larger firm. And the problem was they [0:17] had 900 easement assessments that they needed someone to go through and evaluate for them [0:25] and determine which ones they needed to to look at. uh not a big problem. Um normally that would [0:31] go to the larger firm. Uh the problem came in was that the timeline that the client needed this in [0:40] was about 72 hours. They needed it by that Monday morning and this was Friday [0:46] uh midday when they came to so uh they pitched it to the larger firm. the firm gave them the [0:54] typical answer of usually this type of evaluation and assessment takes you days, weeks, uh usually. [1:03] Um but we had one uh of our attorneys who really kind of stepped up and had a different uh idea. [1:16] So, she asked if she could have the 900 assessments and take a look at them and then pitch [1:23] uh whether or not she could handle the work uh and with a small team. And so she went back to [1:29] the office, took those 900 assessments, loaded them up into a Harvey database, and then uh [1:38] started evaluating those for the assessment with the small team. She started working at on those [1:47] and within 72 hours had done a full assessment on those, had ordered them by level of importance, [1:55] had identified issues with the assessments and presented that to the client. That resulted in [2:02] about $200,000 worth of work. And quite frankly, that's work that normally we would not have [2:09] gotten. a larger uh firm would have gotten that work and not us. Uh but we had that very creative [2:17] attorney who had a different way of looking at it and she had the experience with the tool sets to [2:22] know how to propose this, work it through and get it completed in a short period of time. [2:35] So the problem with that uh story is one of the reasons why I think I'm I'm up here today. Um and [2:44] that is that this depended upon that one attorney to be in the right place and have the right skill [2:50] set at the right time and pitch uh the the right uh way to work that that deal in order to get it. [2:59] Um, that was just kind of luck and I'm branding that a little bit. So, [3:05] it's luck with good branding. So, I like uh take take a look at it that way. [3:12] So, a win like that really stays as a one-off until you build a method around it. And my job [3:21] at Jackson Walker is to make winds like that look very boring, um, look very repeatable and create [3:29] a system around those to make them work. And the method itself has really four steps. And [3:36] you'll watch it through this entire presentation where we extract a senior lawyer's uh, patterns [3:44] out of their own work product. Then we encode those patterns into a playbook and a workflow [3:52] and we test the playbook against live matters and see what holds up and we scale it. Then we [3:59] scale it sideways to the next lawyer across the practice group and across related practice groups. [4:11] So just a little bit of quick context uh so that you know who's who's up here on the stage. Um, [4:16] Jackson Walker is a full-ervice firm. We are Texas only. We have about 550 attorneys. Um, [4:24] litigation, transactional, uh, the whole works. I'm the chief innovation officer there. And my, [4:33] uh, the easiest way that I can explain what I do is, uh, I I work uh, to make and teach lawyers how [4:42] to use these tools and to use them well. So our stack uh is really kind of a pretty basic stack. [4:49] I think most most law firms have some semblance of this. Um we use u uh Harvey uh or Microsoft [4:58] C-Pilot uh to start off uh that uses our M365 uh platform. So all of our one drive, our emails, um [5:08] I was having a conversation earlier. Uh law firms love Microsoft. Uh we we tend to use Microsoft [5:14] products and co-pilot is pretty much a a necessity just because of the the way that we are uh we [5:22] embed the Microsoft tools into our practice. Um the next up is we use West Law Advantage [5:28] and Lexus Protege for legal research. um legal research tools quite honestly for the past couple [5:35] of years uh well up until about a year ago have been horrible. Um but the last last year or so [5:41] these have gotten much much better. Um and so the earlier panel u mentioned something about if you [5:48] haven't uh reviewed something in six months review it again because it's probably much better than it [5:53] was. Uh same thing with them. Um Harvey is far and away our most popular tool. Um it it's used [6:01] for everyday kind of AI interactions used for for drafting for summarization for reviews building [6:08] vaults uh creating their own individual workflows uh on that. So um probably our most popular tool [6:17] and then lastly we're using Claude Enterprise as a pilot. We have about 35 uh technicians and [6:26] lawyers who are working on that. Uh the idea is to build agentic workflows uh that we can't do in any [6:33] of the other tools and then create frontends for those uh in order for us to to use them. [6:42] The one thing I do want to mention because you hear a lot of a lot about it is uh [6:48] as great as co-pilot and great as Harvey gets and as great as Claude Enterprise gets uh we're not [6:55] getting rid of Weslaw or Lexus. Um the the issue there being especially for our litigators, uh [7:01] we need that corpus of data and as good as the uh generic or the general tools get, they're just not [7:09] legal tools and we don't want to be on uh in an article on Above the Law or the New York Times uh [7:17] for citing uh hallucinated cases. So those tools stay um whether we like it or not, unfortunately. [7:28] Now, when I show this uh to a lot of firms, I get I get some eye rolls. [7:32] Uh we have 550 attorneys. We have probably another 380 staff members. [7:37] Um and right now we have about 600 uh users of Harvey uh who each have a license to to that. [7:47] And then uh we check the metrics and this goes up pretty much every month. um 84 84% of them are [7:56] users every week. Most of those are users multiple times a week. So, it's a tool that gets used a lot [8:07] and we really wanted to look at how we did that. [8:13] We skipped the the mandate. We didn't require that anyone get it. In fact, [8:16] we started off with 100 licensed to to Harvey, went up to 150, 200, 250, and now uh we're [8:24] a little over 600 uh license. What we did was we went practice group by practice group and we found [8:32] a champion within each practice group and then we found an issue that that person was having [8:38] with the work that they were doing every day. And then we worked with them to determine if [8:44] this was something that we could improve their experience uh and through the use of the AI tools. [8:50] And in many cases it worked great. We had that one champion who then shared that information with [8:57] someone else. There was a lot of fear of missing out uh on it and next thing I know it's like [9:03] everybody's wanting the tool. Uh lots of training uh going in on it. um which may be our biggest [9:11] biggest issue going forward is you know how do we scale you know one and me and [9:16] and do the training across um multiple practices. Now um the one thing that you uh one thing that [9:26] we run into and that 84% can really be kind of a vanity me metric uh for us because it's one [9:32] thing for the people to be using it but are they using it correctly? Um most of the 84% [9:38] are using it on their individual matters, their individual use case and that's great but that [9:45] comes with a ceiling uh on it as well. Uh there's only so much even a 69page prompt can do for you. [9:54] So, one of the things that I say probably too much uh at my firm uh when I talk with lawyers is that [10:03] they do better with a red pen than they do with a blue pen. If you give them something to edit, [10:08] they're great. If you give them something to create, not so much. Um so, if you hand a [10:14] lawyer a draft to mark up, the expertise just pours out of them. It's what it's what [10:19] they do. It's what they're trained to do. it's what they they've done built their career on. [10:25] And what that means is there's something kind of wonderful uh that we have access to and that is [10:31] that the playbooks that we want to create are actually already there. It's just we have to [10:37] surface them. Um every senior partner has spent 20 or 30 years of producing red lines. Um those [10:45] red lines contain their judgment. um the positions that they'll always take, the fallback language [10:51] that they'll accept when when needed, and the terms that makes them walk away from a deal. [10:57] All of that is in closed files. It's in our document management system, and it's [11:02] just waiting for someone to come and get it. So, what's new is that the AI tools have finally made [11:09] that knowledge extractable at a reasonable cost. So, the method itself goes like this. We extract [11:16] the patterns from the uh the reviews that the um partner has worked on from their own documents. [11:22] We in encode them with explicit rules to follow. And this is the part that really matters that we [11:30] give that attorney a red pen to go back and then correct uh what what it is, what works, [11:37] what doesn't work. Um and that makes it concrete for us. So I want to give a real example here. [11:46] You can hit the right button. Um so this is is an example of a real build that we did with [11:53] a finance partner. This was a borrower side commercial real estate loan agreement. Um we [11:59] started off with six sets of red lines uh that we had with the uh doctor or with the um documents [12:07] from his previous work. Um we extracted the information using the AI tool on those uh [12:15] on those seven or on those six um um deals. We found that there were seven of the articles of [12:23] these loan agreements that get negotiated every single time. And across those seven [12:28] deals, the partner was taking roughly about 90 recurring positions on those. So we had the list, [12:37] we gave it to the partner of that 90 and then we asked the partner which one of these are actually [12:42] the ones that you you work on, which one of these are one offs for the deal. Um which one of these [12:49] you have to make the determination uh at at that point. Um, and what it survived was a 49 rule [12:58] playbook. And each one of those rules carries the full full shape of the of the judgment of [13:05] the partner. That's our standard position. That's the fallbacks we can live with. That's the point [13:10] where we walk away. That's the actual language uh from real deals showing us what it looks like. And [13:17] the uh the production itself works in Microsoft Word. And there's one other thing that this [13:23] gives you. Um I didn't don't have it up there. These rules are in plain language. These are in [13:29] markdown files. Um one of the side benefits that we discovered was that we could give the associate [13:36] the playbook and they could look at that and would have a good idea of what that partner that they're [13:42] working with would accept, what they wouldn't accept, what what their judgment told them. [13:51] And I know we we dream of the the I guess good old days of [13:56] uh partners taking the associates under under their wing and teaching them all this. [14:01] Um but I've been doing this for 30 years and I'd never seen that. Um so uh [14:06] we we finally have figured out a way uh to to make that happen. Um so this works in Microsoft Word. [14:14] The associate pulls up the new deal. They've used the playbook and they run it against the the deal [14:20] and it runs through those uh 4849 uh rules that they have and it gives them [14:26] this meets it. This this is questionable. This is outside the scope. Um it makes it very easy. [14:37] The other things that we do let's see [14:43] so again it's the same method. We're we're finding that these methods are transferable. [14:48] Um we can do one and then all of a sudden we can see that the same process can help us find another [14:56] uh playbook to work with with another partner. Um and so the uh [15:03] couple of examples here uh we have a contract partner. Um and [clears throat] this one is one [15:09] where everyone said they wish they could clone them. Um same thing we we built two uh draft [15:14] workflows. Um each one was extracted from their own work, reviewed by them, approved, signed off, [15:21] uh workflows uh created and reviewed and then the partner works with the associates. So as [15:28] things change, then we can go in and change the the workbook or the workflow itself. [15:35] So again, notice that it was the same method. It's it's the same every time. Uh we use [15:42] apply this in different practice groups. We're running on different tools. Um the expertise [15:47] is already there. It's just a matter of putting that expertise in front of the partner, giving [15:52] them a red pen and saying this is what I do. Now the uh one thing that I love that has happened [16:00] recently is that we have merged our AI tools with our document management system. If you're [16:05] in law firms, this is either the the best thing that h that has happened to you or the bane of [16:11] your existence is to put things into IM manage or into NetDucks um is required. But the nice [16:19] thing about the AI tools is that it enables us to go in and look at the previous work. It's [16:24] structured by the client. It's structured by the client's work. It's protected. uh all the guard [16:29] rails are there so that people cannot get into things that uh they ethically cannot get into. [16:35] So those walls still stay there. Um but it enables us to surface a lot of things using the AI tools [16:41] that we would have not been able to uh otherwise or or at least not as fast. [16:49] And so one of the things that we've done is uh I know this doesn't happen at your place but [16:54] at our place we do this thing it's called pardon the interruption emails uh P pti and it's uh hey [17:01] does somebody know somebody in this county who's been before this judge or do you know [17:06] so what we started doing was actually sending all of those through our system and then surfacing [17:13] the an or the the result for that and sending it back uh to the person that had questioned it and [17:18] like 95% of the time it gives them a great answer for it. Um, and in addition to that, [17:24] uh, it it also one there was one really good case where there was a pro prosay plainif who [17:30] was a serial uh, filer in California. Email went out asking if we if anyone had worked had uh, [17:38] had faced this person before. Uh, within five minutes we had a report that showed [17:42] the four attorneys who had done it. Um, one of our attorneys had actually written a memo on [17:48] how to deal with it and then how they were they were counteracting the serial pliff's uh work. [18:00] So there um now the boring part is uh all of this of course has to be uh governed um the workflows [18:10] itself uh work through the firm. Um we make sure that things are logged that things are access when [18:16] it's accessed. Uh we design ethical walls in in the document management system and the AI itself [18:24] works from uh from snapshots. And you heard on the earlier panel that uh cost discipline is is a [18:31] thing. Uh to tokens are not as cheap as they used to be. Um, and so we have to make sure that as [18:38] we're building these tools that we're doing them in an efficient manner, that we're not spending [18:43] $10,000 to s to to get $1,000 of work. And third, you have to measure. And the measure is again stay [18:52] away from the vanity metrics. Um, look at revenue capture, look at time, uh, turnaround time, [18:59] uh, how matters move through, uh, the system. And so those are the types of of measurements [19:05] that really really matter more than it does 84% of our people who use Harvey every week. [19:13] So where does this all go? So here's my projection. Um most partners will never [19:18] become skilled prompters. Love them. It's just not what they want to do. And we design around that. [19:25] um the next phase of the the system uh driven AI, you know, you picture a partner opening up a [19:31] a new M&A matter, running that through a familiar uh interface. She supplies the the information, [19:38] the details of the deal. It prompts her for the documentation that is needed. It actually [19:44] prompts her with the documentation that's needed. Um and then the design delivers this consistency [19:52] uh of output. Um the other thing is that um clients really want you to be using this, [20:03] but they're they're just as unfamiliar uh with with the processes uh as as we are. Um [20:10] the thing that we really encourage our attorneys to do is to have that conversation with them. [20:16] um because we have found that efficiency actually ends up growing the pie because those 900 matters [20:23] would not have come to us. Um and so the uh the ability to have that conversation, [20:30] that hard conversation with your client is what you need to have. So I am out of time. Um thank [20:38] you. I'm Greg Lambert. I'll be in the hallway if you need me. Thank you very much. [applause]