EY Automates 40 Legacy Tax Apps Using IBM's New Agentic Platform Bob

Modernizing the SDLC with Agentic AI With IBM And EY — Ai4

Ai4

Agentic AI shifts enterprise software development from manual syntax authoring to agent orchestration, embedding automated governance directly into QA and legacy modernization pipelines.

AI-assisted summary of Modernizing the SDLC with Agentic AI With IBM And EY — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published by the Ai4 channel on YouTube on August 19, 2026, this session features IBM Software Architect Liam Allan and EY Managing Director Venkat Venkatesan. The presentation explores how IBM Bob, an agentic AI-powered software development lifecycle platform, enables enterprise-scale modernization, automated testing, and governed workflows across complex codebases.

## TL;DR IBM's Liam Allan and EY's Venkat Venkatesan discuss transitioning from isolated code-generation tools to full-lifecycle agentic AI using IBM Bob. Drawing on EY's experience modernizing over 40 tax platform applications, Venkatesan illustrates how Bob's three operational modes—Ask, Plan, and Agent—streamline architectural debates, autonomous code execution, and legacy codebase comprehension. The conversation highlights integrated governance, self-healing QA pipelines, and enterprise-wide spend and token visibility through Bobalytics. Ultimately, the speakers project that developers will evolve from manual line-by-line coders into strategic orchestrators and architectural editors managing collaborative AI agent swarms.

## Atmosphere & style The session carries the warm, engaging energy of an authentic fireside chat between technical collaborators. Allan and Venkatesan balance pragmatic enterprise architecture and governance discussions with lighthearted humor, including playful banter around British pronunciations and 'Bob the Builder' metaphors. It feels grounded and forward-looking, translating ambitious agentic AI concepts into the tangible operational realities of modern enterprise IT.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=4s) Enterprise AI adoption in regulated industries requires shifting focus from isolated local coding assistants to unified, governed SDLC automation. - [02:15](https://www.youtube.com/watch?v=JjP1052QXZU&t=135s) IBM Bob functions as a single SDLC platform connecting planning, QA, and deployment across both modern stacks and core legacy environments like Mainframes and IBM i. - [02:15](https://www.youtube.com/watch?v=JjP1052QXZU&t=135s) Automated self-healing pipelines within Bob can detect test failures in QA and deploy agents to autonomously draft and apply fixes. - [05:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=304s) EY integrated Bob across 40+ tax applications, converting governance from a dreaded weekly meeting into continuous, automated checks inside the workflow. - [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) Bob structures developer interaction across three distinct modes: Ask Mode for code inspection, Plan Mode for architectural debate, and Agent Mode for execution and deployment. - [08:16](https://www.youtube.com/watch?v=JjP1052QXZU&t=496s) Beyond raw speed, the core enterprise ROI of Bob lies in uncovering undocumented business logic in legacy codebases to de-risk modernization. - [08:16](https://www.youtube.com/watch?v=JjP1052QXZU&t=496s) AI agents accelerate onboarding by acting as on-demand technical mentors, freeing senior engineers from spending days guiding new hires. - [08:16](https://www.youtube.com/watch?v=JjP1052QXZU&t=496s) Bob pairs code modifications with automated test planning and explicit rationales to mitigate regression risks across the build pipeline. - [10:53](https://www.youtube.com/watch?v=JjP1052QXZU&t=653s) Bobalytics provides enterprise leadership with real-time visibility into token consumption, cloud costs, and developer throughput across large engineering organizations. - [10:53](https://www.youtube.com/watch?v=JjP1052QXZU&t=653s) Over the next 3 to 5 years, developers will receive an 'automated promotion' to high-level architects and agent orchestrators rather than manual syntax authors. - [10:53](https://www.youtube.com/watch?v=JjP1052QXZU&t=653s) Scaling agentic AI across enterprise environments is primarily an organizational adoption and cultural change challenge rather than a simple tooling procurement.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=4s) IBM's 2025 AI Rollout and Enterprise SDLC Challenges

Liam Allan opens the session detailing his dual perspective as both an engineer and a product manager for extensibility on IBM Bob, focusing on pushing the tool into environments where AI assistants would not typically operate. Reflecting on IBM’s enterprise rollout of AI development tools throughout 2025 across an organization exceeding 150,000 employees—roughly 70,000 of whom are software developers—he highlights the systemic gaps that emerged. While individual developers proved capable of writing code and testing locally or in personal cloud sandboxes using AI, organizations struggled to manage spend across tens of thousands of users, monitor cloud infrastructure costs, and enforce strict governance when building for highly regulated sectors such as government, banking, and aviation.

The primary breakdown in traditional AI tooling is the handoff from local development to automated deployment. In standard developer workflows, engineers create plans, write code, and verify tests locally, but the progression from local testing to automated quality assurance (QA) environments and governed enterprise deployments remains disconnected. The challenge moving into 2026 is moving beyond isolated code completion to construct a unified agentic system capable of governing access, monitoring execution, and automating the path from development to QA and production.

- **IBM Scale:** Over 150,000 total employees, including approximately 70,000 developers and 70,000 non-developers interacting with AI systems. - **Enterprise Governance Scope:** Regulated customer environments requiring strict agent tracking include defense/government, banking, and airlines. - **Core Process Gap:** Moving software smoothly and automatically from local test environments to QA and cloud deployment pipelines without manual bottlenecks.

---

[02:15](https://www.youtube.com/watch?v=JjP1052QXZU&t=135s) IBM Bob: Unified Governance, Ecosystem Integration, and Automated Remediation

Liam introduces IBM Bob (humorously acknowledging how his British pronunciation contrasts with American accents) as a unified platform spanning the entire software development lifecycle (SDLC). Unlike fragmented point solutions, Bob provides a common pane of glass for developers, DevOps engineers, business analysts, and governance officers. A cornerstone of the architecture is centralized analytics and governance, enabling leaders to observe token usage, track spend across 70,000 developers, and identify productivity trends across projects.

```mermaid flowchart LR accTitle: IBM Bob Full SDLC Architecture A["Developer / BA / DevOps"] --> B["IBM Bob (Single SDLC Tool)"] B --> C["Plan & Code (GitHub Enterprise)"] B --> D["Automated QA & Self-Healing Pipelines"] B --> E["Multi-Cloud & Legacy Deployment"] B --> F["Bobalytics (Spend, Tokens & Governance)"] ```

Because Bob is deeply integrated with enterprise systems such as GitHub Enterprise and multi-cloud environments, it can detect pipeline failures in QA and trigger automated agentic remediation workflows to fix broken builds and tests before human intervention is required. Furthermore, Bob natively integrates with both modern stacks and core enterprise legacy systems—including IBM Z Mainframes, IBM i (historically known as AS/400), Java, and IBM WebSphere—with ongoing expansion planned for platforms outside the native IBM ecosystem.

- **Supported Enterprise Environments:** Mainframe, IBM i (AS/400), WebSphere, Java, GitHub Enterprise, and diverse multi-cloud environments. - **Automated Self-Healing:** Autonomous pipelines that catch test failures in QA and assign agents to draft and apply fixes. - **Visibility:** Centralized tracking of organizational spend, cost hotspots, and developer throughput across 150,000 daily users.

---

[05:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=304s) Fireside Chat with EY: Modernizing Tax Platforms with Bob

Liam welcomes Venkat Venkatesan, Managing Director for Platforms, Products & AI at EY, who flew in specifically for the event. Venkatesan describes her responsibilities overseeing EY’s Americas tax platform, an ecosystem comprising more than 40 applications serving numerous enterprise clients. Having participated in Bob's early private preview, EY transitioned from evaluating the technology to treating Bob as a core, embedded "team member."

Venkatesan explains that adopting Bob transformed governance at EY from a periodic chore into a continuous, native part of development. Instead of developers dreading a recurring "every Tuesday governance call," compliance and quality checks are embedded directly inside the workflow. The integration also sparked internal enthusiasm and friendly competition among engineers through "Bob coins," a mechanism developers use to compare usage and output while driving ambitious development goals.

- **EY Scope:** Complex tax platform managing 40+ production applications across large enterprise client bases. - **Continuous Governance:** Shifts policy checks from weekly managerial syncs directly into the automated lifecycle. - **Gamification & Morale:** Developers track usage and challenge peers via internal "Bob coins" metrics.

---

[07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) The Three Modes of IBM Bob: Ask, Plan, and Agent

When explaining how developers interact with the platform, Venkatesan uses an automotive metaphor: Bob operates like a car equipped with three distinct gears—Ask Mode, Plan Mode, and Agent Mode. Each mode serves a specific stage in the engineering thought process, transitioning the tool from an analytical advisor to an autonomous builder.

```mermaid flowchart TD accTitle: The Three Operational Modes of IBM Bob subgraph ThreeModes["IBM Bob Three Operating Modes"] M1["1. Ask Mode<br/>(Inspects codebases, explains logic, passive analysis)"] M2["2. Plan Mode<br/>(Architectural strategy, back-and-forth debate, test planning)"] M3["3. Agent Mode<br/>(Executes code generation, builds, tests & deployment)"] end M1 --> M2 --> M3 ```

In Ask Mode, Bob inspects existing codebases and answers developer queries regarding functionality without modifying any code. In Plan Mode, Bob proposes architectural roadmaps and execution strategies for complex code modifications. This mode fosters collaborative debate; developers can reject initial suggestions, critique approaches, and iterate on design strategies before writing code. Finally, Agent Mode performs the execution, generating builds, writing tests, and running deployment routines.

---

[08:16](https://www.youtube.com/watch?v=JjP1052QXZU&t=496s) Beyond Table Stakes: Legacy Knowledge, Onboarding, and Testing Coverage

Addressing whether productivity is the only return on investment, Venkatesan calls raw speed "table stakes." The deeper value lies in code modernization and institutional knowledge recovery. Many enterprise codebases contain mission-critical business logic written twenty years ago by engineers who have long since departed. Bob analyzes these legacy repositories, explaining why and how code was originally constructed, which de-risks modernization projects that were previously avoided due to high cost and architectural complexity.

This contextual understanding accelerates new developer onboarding. Rather than pulling senior engineers away from feature development for a full week to mentor newcomers, new hires leverage Bob to navigate the codebase independently. Regarding risk mitigation, Venkatesan clarifies that Bob is not a simple line-completion or comment-finishing tool; it actively constructs comprehensive test coverage plans and provides clear rationales for every change across the build, test, and deployment phases.

- **Knowledge Retrieval:** Deciphers undocumented business rules and architectural patterns hidden in decades-old legacy code. - **Onboarding Velocity:** Reduces senior developer mentoring burdens by acting as an on-demand technical mentor for new team members. - **Risk Mitigation:** Generates dedicated test suites and coverage strategies alongside code modifications.

---

[10:53](https://www.youtube.com/watch?v=JjP1052QXZU&t=653s) Enterprise Visibility with Bobalytics and the 3–5 Year Horizon for Agentic AI

Venkatesan outlines EY's relationship as "Client Zero" for IBM Bob, noting how the platform matured from private preview into General Availability (GA), with excitement building around the analytics suite, known as Bobalytics. Liam clarifies that Bobalytics provides enterprise leaders with immediate visibility into spend, token consumption, and workload distribution, ensuring agentic operations remain accountable and cost-effective across large-scale deployments.

Looking ahead 3 to 5 years, Venkatesan predicts a paradigm shift where agents move beyond isolated task automation to pervasive collaboration, comparable to how internet connectivity shifted from a manual configuration hassle to an invisible utility. Instead of merely operating software, human engineers will manage and orchestrate teams of agents. Consequently, developers are receiving an "automated promotion"—shifting their day-to-day focus from raw syntax authorship to high-level architecture, editing, and rigorous validation.

```mermaid flowchart LR accTitle: Evolution of the Developer Role A["Traditional Developer<br/>(Code Author / Line-by-Line Coder)"] --> B["Modernized Developer<br/>(Architect, Editor & Agent Orchestrator)"] B --> C["AI Agent Swarm<br/>(Execution, Syntax, Testing, Deployments)"] ```

- **Bobalytics Capabilities:** In-depth monitoring of token usage, organizational spend, and hot project paths. - **Future Developer Archetype:** Evolving from manual syntax writers into architectural editors and agent conductors who define the "why" while agents execute the "how." - **Organizational Realities:** Emphasizes that successful AI deployment is an organizational adoption and culture-change challenge, not a simple procurement decision.

---

[15:21](https://www.youtube.com/watch?v=JjP1052QXZU&t=921s) Concluding Thoughts and Live Demonstrations

Wrapping up the fireside chat, Venkatesan compares Bob's reliable, collaborative nature to the classic "Bob the Builder" persona—equipped with a metaphorical hard hat and ready to tackle heavy construction work. Liam thanks Venkatesan for joining the session immediately after her morning flight, and directs attendees to the IBM conference booth where a QR code is displayed on screen. He invites the audience to participate in hands-on, end-to-end demonstrations using provided laptops to experience Bob building, testing, and deploying applications live.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=4s) Moving software smoothly and automatically from local test environments to QA and cloud deployment pipelines without manual bottlenecks. — Liam Allan

> [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) Bob operates like a car equipped with three distinct gears—Ask Mode, Plan Mode, and Agent Mode. — Venkat Venkatesan

> [08:16](https://www.youtube.com/watch?v=JjP1052QXZU&t=496s) productivity is table stakes. — Venkat Venkatesan

> [08:16](https://www.youtube.com/watch?v=JjP1052QXZU&t=496s) Many enterprise codebases contain mission-critical business logic written twenty years ago by engineers who have long since departed. — Venkat Venkatesan

> [08:16](https://www.youtube.com/watch?v=JjP1052QXZU&t=496s) Bob is not a simple line-completion or comment-finishing tool; it actively constructs comprehensive test coverage plans and provides clear rationales for every change across the build, test, and deployment phases. — Venkat Venkatesan

> [10:53](https://www.youtube.com/watch?v=JjP1052QXZU&t=653s) Client Zero — Venkat Venkatesan

> [10:53](https://www.youtube.com/watch?v=JjP1052QXZU&t=653s) shifting their day-to-day focus from raw syntax authorship to high-level architecture, editing, and rigorous validation. — Venkat Venkatesan

> [10:53](https://www.youtube.com/watch?v=JjP1052QXZU&t=653s) successful AI deployment is an organizational adoption and culture-change challenge, not a simple procurement decision. — Venkat Venkatesan

## Mentioned Liam Allan, Venkat Venkatesan, IBM, EY, IBM Bob, Bobalytics, GitHub Enterprise, IBM Z Mainframes, IBM i, AS/400, Java, IBM WebSphere, Bob the Builder, Ai4

## Video description Liam Allan - Software Architect, Speaker & Advocate, IBM Venkat Venkatesan - Americas Tax, Platform & Products, EY

Enterprises have spent the past two years experimenting with AI coding tools. Now the mandate has shifted from speeding up snippets to realizing governed, enterprise-scale transformation. In this session, we explore how IBM Bob, an AI-powered SDLC partner, transforms application modernization into a streamlined, automated, and scalable process. Going beyond traditional code assistants, you need a partner that analyzes entire codebases, generates modernization strategies, and validates outcomes with built-in governance and testing. Join us for a fireside chat with Venkat Venkatesan, Managing Director for Platforms, Products & AI at EY, as she shares EY's experience leveraging IBM Bob to accelerate application modernization and drive business transformation at scale. We will walk through a full modernization journey. Give insights into how an agentic AI approach can reduce modernization timelines from months to days, improve code quality, and drive enterprise transformation at scale.

[Source](https://www.youtube.com/watch?v=JjP1052QXZU)

中文简报

EY借助IBM全新智能体平台Bob实现40款传统税务应用自动化 智能体AI正推动企业软件开发从手动语法编写转向智能体编排，并将自动化治理直接嵌入QA与传统系统现代化流水线中。 IBM的Liam Allan与EY的Venkat Venkatesan探讨了如何借助IBM Bob从孤立的代码生成工具迈向全生命周期的智能体AI。结合EY对40多个税务平台应用进行现代化改造的经验，Venkatesan阐释了Bob的三种运行模式——Ask、Plan与Agent——如何简化架构研讨、自主代码执行以及对传统代码库的理解。对话重点提及了集成治理、自愈式QA流水线，以及通过Bobalytics实现的企业级支出与Token可见性。最终，发言者预计开发者将从逐行编写代码的人员，演进为管理协同AI智能体群的战略编排者与架构编辑者。 受监管行业的企业级 AI 应用正在从孤立的本地代码补全转向覆盖全生命周期的统一 SDLC 自动化与严格治理。 (0:04) IBM Bob 作为贯穿全 SDLC 的统一平台，不仅原生连接现代技术栈以及大型机（Mainframe）、IBM i 等核心遗留系统，还具备在 QA 阶段检测到测试失败时自动部署 Agent 进行修复的自愈流水线能力。 (2:15) 安永（EY）在 40 多个税务应用中引入 Bob，将过去令人头疼的每周例行业务治理会议转变为内嵌于开发工作流中的持续自动化合规检查。 (5:04) Bob 提供“提问（Ask）”、“规划（Plan）”与“代理（Agent）”三种模式，分别对应代码检查、架构方案探讨以及自主构建、测试与部署。 (7:08) 除了单纯提升开发速度，Bob 的核心企业投资回报率还在于解析遗留代码中未记录的业务逻辑以降低现代化重构风险，并通过充当技术导师来加速新员工入职培训。 (8:16)

Reading summary

IBM and EY argue that scaling AI in the enterprise software development lifecycle requires shifting from isolated code-completion tools to unified agentic platforms that govern spend, automate QA, and modernise legacy systems. Enterprise Scaling and Governance Bottlenecks (0:04) Liam Allan outlines that during IBM's 2025 rollout of AI tools across 150,000 employees—including 70,000 developers—the primary breakdown occurred during the handoff from local sandboxes to deployment. While engineers could write and test code locally, enterprises struggled to control cloud and token costs, maintain visibility across tens of thousands of users, and enforce compliance in regulated environments like banking, aviation, and government. Allan explains that isolated code-generation tools leave a manual gap between local verification and automated quality assurance. Moving into 2026, enterprise adoption requires integrating access governance, monitoring, and automated deployment pipelines into a single framework rather than relying on point solutions. IBM Bob's Architecture and Three Interaction Modes (2:15) To address fragmented toolchains, IBM developed IBM Bob, an SDLC platform that connects developer environments like GitHub Enterprise to multi-cloud and legacy systems such as IBM Z Mainframes, IBM i (AS/400), Java, and WebSphere. The system incorporates automated self-healing pipelines that detect QA test failures and assign agents to draft and apply fixes before human intervention is required. Venkat Venkatesan of EY explains that developers interact with Bob across three progressive operating modes: Ask Mode, Plan Mode, and Agent Mode. In Ask Mode, the tool inspects codebases and explains functionality without editing. In Plan Mode, it proposes architectural roadmaps and test plans, allowing developers to critique and debate strategies. In Agent Mode, it autonomously executes code generation, builds, tests, and deployment workflows. Modernising Legacy Logic and Development Workflows at EY (5:04) Serving as 'Client Zero' across its Americas tax platform of over 40 applications, EY embedded Bob directly into development routines. Venkatesan notes that this integration shifted compliance checks from weekly managerial meetings into continuous, automated background reviews, while internal gamification via 'Bob coins' helped track developer usage. Venkatesan stresses that productivity gains are secondary to recovering institutional knowledge and managing risk. Bob can analyze undocumented business logic written decades earlier by departed engineers, lowering the barrier to modernising complex legacy codebases. Additionally, the platform accelerates onboarding by answering technical questions for new hires without pulling senior engineers away from feature development, and it generates comprehensive test coverage rationales alongside code changes. Developer Role Evolution and Bobalytics Tracking (10:53) For governance and operational tracking, Allan and Venkatesan highlight Bobalytics, an analytics suite providing real-time visibility into token usage, project hotspots, and organizational expenditure across large deployments. Looking ahead 3 to 5 years, Venkatesan forecasts that software engineers will shift from manual syntax authors into architectural editors and agent conductors who define system requirements while autonomous agent swarms execute the underlying build, test, and deployment steps. Both speakers emphasize that achieving this transition requires solving internal cultural adoption challenges rather than treating AI tooling as a standard procurement exercise. While IBM Bob demonstrates automated remediation and knowledge recovery across modern and legacy architectures, long-term success relies on organizations successfully managing cultural adoption, cost accountability via analytics, and the fundamental shift of developers into supervisory architectural roles.

精炼阅读摘要

在Ai4大会上，IBM的Liam Allan与安永（EY）的Venkat Venkatesan指出，企业软件研发的矛盾已从单纯的代码生成效率，转移至全流程治理合规、多云成本控制与遗留架构现代化。演讲者认为，必须建立贯穿开发、测试与部署的端到端智能体平台，在实现流水线自动化自愈的同时，推动工程师向架构编排者转型。 规模化研发瓶颈与全流程断点 (0:04) Liam Allan结合IBM在2025年面向超过15万名员工（涵盖约7万名软件开发者及7万名非技术交互人员）推广AI工具的实际经验，指出了规模化落地中的结构性缺陷。单个工程师虽然能熟练运用AI在本地或个人云沙箱中编写并测试代码，但当应用范围扩展至数万人时，企业在全组织Token与开销控制、多云基础设施支出监控，以及面对政府、国防、金融与航空等强监管行业的合规追踪时，面临严峻挑战。 Allan强调，现有研发模式的核心断层在于从本地验证到企业级受管部署的衔接脱节。常规流程中开发者在本地完成规划、编码与测试，但后续推进至自动化质量保证（QA）环境及合规生产流水线的路径仍然支离破碎，高度依赖手工介入，阻碍了端到端交付的流畅运转。 全生命周期集成与流水线自愈修复 (2:15) 针对单点开发工具碎片化的问题，Allan介绍了覆盖完整软件开发生命周期（SDLC）的统一平台IBM Bob。该系统为开发人员、DevOps工程师、业务分析师与合规官提供单一管理视窗，并深度集成GitHub Enterprise以及多云运行环境，同时原生向下兼容IBM Z大型机（Mainframe）、IBM i（AS/400）、Java和WebSphere等关键遗留系统，后续亦计划继续拓展非IBM技术栈。 该平台架构的关键能力在于自主化流水线运作。当系统在QA阶段检测到构建中断或测试失败时，能够自动触发智能体修复工作流，由智能体自主分析根因、起草并应用修复补丁，从而在无需人工排查介入的前提下恢复流水线运转，降低流水线维护的人工瓶颈。 持续合规嵌入与三档交互机制 (5:04) 安永（EY）美洲税务平台主管Venkat Venkatesan以其管理覆盖40余个生产级应用的企业系统为例，分享了作为IBM Bob“客户零号”的深度应用实践。Venkatesan指出，该平台改变了企业原有的监管模式，将过去让开发人员视作负担的“每周二合规评审会”，转变为直接内嵌在生命周期中的原生持续合规检查，并通过内部“Bob coins”机制量化产出以调动团队积极性。 在日常工程交互方面，Venkatesan将Bob的操作逻辑类比为汽车的三种不同挡位：Ask模式仅作为代码库静态分析助手，负责解释已有逻辑而不修改代码；Plan模式则支持人机协作探讨架构路线与测试方案，允许开发者对初步提案提出反驳与迭代；最终的Agent模式才承担具体执行角色，负责自主生成代码、构建用例并运行部署程序。 遗留知识召回与工程师角色演进 (8:16) Venkatesan明确提出，单纯的代码编写提速仅仅是“基础门槛”，智能体技术的深层回报在于破解遗留技术债务与唤醒制度性知识。大量核心企业代码由数十年前已离职的工程师编写，Bob通过逆向梳理并解释无文档记录的业务逻辑与架构，大幅降低了企业此前因高昂成本与架构风险而搁置的系统现代化阻力，同时替代资深人员承担部分新人带教工作。 结合提供全局Token消耗与项目开支洞察的Bobalytics分析系统，Venkatesan预测，未来3至5年内智能体协同将如底层互联网般成为隐形基础设施。开发者的核心职能正在经历一场“自动化晋升”，由过去纠结语法的逐行代码撰写者，转变为负责确立战略意图（“Why”）、审核架构设计并统筹智能体集群执行具体任务（“How”）的系统编排师。 演讲者总结指出，将智能体全面引入软件工程绝非一次单纯的技术采购，而是一项涉及团队文化、协作习惯与责任界定的组织级转型；企业在推进自愈交付的同时，必须依赖细粒度的成本治理套件来防范资源失控，并确保强监管体系下的透明度与责任可溯性。

Detailed notes — zh

## Background 本视频由 Ai4 频道于 2026 年 8 月 19 日在 YouTube 平台发布。内容围绕 IBM 软件架构师 Liam Allan 与安永（EY）董事总经理 Venkat Venkatesan 的炉边对谈展开，重点探讨企业如何超越基础代码补全工具，借助 Agentic AI 平台 IBM Bob 实现全生命周期的应用现代化改造与严格的企业级治理。 ## TL;DR IBM 与安永分享了将 Agentic AI 落地于软件开发生命周期（SDLC）的企业级实践。面对大规模开发团队的成本盲区与合规痛点，IBM 推出打通规划、编写、测试到多云部署的平台 IBM Bob，并配备 Bobalytics 治理看板以追踪支出与工作流。安永作为“零号客户”，通过 Ask、Plan 与 Agent 三档交互模式重构遗留代码并赋能团队，见证了开发者从基层编写者向“架构师与主编”的角色晋升。对谈强调，企业 AI 落地的真正胜负手不仅在于模型技术本身，更在于组织变革、采纳度提升以及从“操作 Agent”向“管理调度 Agent”的思维转型。 ## Atmosphere & style 整场对谈兼具务实的技术深度与轻松风趣的互动氛围，两位演讲者配合默契，穿插了英美发音差异调侃与《巴布工程师》的双关幽默。交流既展现出大型企业落地 AI 时的严谨审慎（直面合规审计与成本监控），又洋溢着对技术范式变革的充沛热情，让观众在直观感受研发效能跃升的同时，清晰体会到组织文化重塑的现实分量。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=4s) IBM 在全公司规模化推广 AI 的实践表明，企业级核心痛点在于跨数万名开发者的算力支出追踪，以及高合规行业（政务、银行、航司）中的智能体合规审计与治理。 - [00:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=4s) 传统研发流程中本地开发测试与云端 QA 环境之间存在严重断层，亟需贯穿端到端软件开发生命周期（SDLC）的自动化链路。 - [02:15](https://www.youtube.com/watch?v=JjP1052QXZU&t=135s) IBM Bob 突破了传统代码补全助手的局限，构建起覆盖需求规划、自动化测试修复、合规把控到多云部署的全角色智能体协作平台，并原生支持 Mainframe、IBM i 等遗留架构。 - [05:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=304s) 安永（EY）在 40 多个企业级税务应用中将 AI 视为“团队成员”，推动合规审查从定期的被动例会转变为嵌入日常交付流的自动化治理机制。 - [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) IBM Bob 采用 Ask（只读问答）、Plan（架构方案博弈）与 Agent（自动化构建执行）三档工作模式，匹配不同的研发与设计场景。 - [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) 代码提效只是 AI 应用的基础门槛，其更深层的价值在于理解沉淀数十年的遗留代码逻辑，加速企业核心资产的系统现代化与重构。 - [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) 为防止开发加速导致质量失控，智能体需在生成代码的同时自动输出测试覆盖计划（Test Coverage Plan），端到端保障系统可靠性。 - [10:59](https://www.youtube.com/watch?v=JjP1052QXZU&t=659s) 治理平台 Bobalytics 提供针对 Token 计量、财务支出以及团队高频代码库的实时可观测性大盘，形成完整的企业级治理闭环。 - [12:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=724s) 未来 3 至 5 年的人机协作将从“操作工具”转向“管理与调度智能体”，人类专注于回答业务与战略层面的“为什么（Why）”，而智能体接管技术落地的“如何实现（How）”。 - [13:20](https://www.youtube.com/watch?v=JjP1052QXZU&t=800s) 开发者角色将迎来自动化晋升，从底层的模板代码编写者演变为系统架构师与“主编”，核心职责转化为对智能体生成方案的推敲、质疑与最终审查。 - [14:27](https://www.youtube.com/watch?v=JjP1052QXZU&t=867s) 企业 AI 规模化落地的最大瓶颈并非技术成熟度或单纯的软件采购，而在于团队采纳度、流程变革管理以及人机协同的企业文化重塑。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=4s) IBM 规模化落地 AI 工具的挑战与治理痛点 IBM 软件架构师兼可扩展性产品经理 Liam Allan 登台分享了他日常负责的核心产品，并回顾了 IBM 在 2025 年向全公司推广 AI 开发工具时积累的深刻洞察。IBM 拥有超过 150,000 名员工，其中约有 70,000 名开发者，其余则是同样庞大的非技术业务人群。在推广过程中，IBM 发现开发者借助 AI 固然能够产出优质成果，但在企业级全局管理上存在严重脱节：如何追踪 70,000 名开发者的实际 AI 支出？当应用部署到内部服务和专属云端时，AI 产生的成本如何核算？更关键的是，IBM 长期服务于政府、银行、航空公司等对合规要求极高的行业，如何对开发者及 Agent（智能体）与关键业务单元的交互实施全面合规审计与治理？ Liam 指出，以往开发者的工作流通常局限于写代码、做规划、实现逻辑，并在本地或个人云环境中运行测试。然而，行业长期缺失的关键一环在于：如何让代码无缝且全自动地从本地开发测试跨越到 QA 测试环境？如何帮助业务端加快上线节奏，同时严格管控部署权限、确保仅有授权人员与 Agent 能够访问关键资产？这些痛点正是 IBM 决定打造能够覆盖完整软件开发生命周期（SDLC）产品的核心动力。 - **庞大用户体量**：IBM 全球 150,000+ 员工，其中包含 70,000 名核心开发人员及同等体量的非技术人员。 - **治理与成本盲区**：缺乏跨团队的 Token 与算力支出追踪机制，缺少针对高合规行业（银行、政务、航司）的 Agent 级治理与审计跟踪。 - **环境割裂**：本地开发、本地测试与云端 QA 环境自动化验证之间存在断层。 --- ### [02:15](https://www.youtube.com/watch?v=JjP1052QXZU&t=135s) 贯穿全 SDLC 的智能体平台：IBM Bob 针对传统割裂的工具链，IBM 推出了全流程智能体协作平台 IBM Bob。Liam 幽默地调侃了自己的英国口音与美式发音差异，并强调 IBM Bob 并非单纯的代码补全助手，而是一个覆盖编码、DevOps 运维、合规审查等所有角色的全生命周期单一平台。 ```mermaid flowchart LR accTitle: IBM Bob 全生命周期治理与交付流 subgraph Dev["开发与规划"] A["业务分析与规划"] --> B["代码编写与本地测试"] end subgraph Pipe["CI/CD 与质量保障"] B --> C["企业级 GitHub 集成"] C --> D["QA 环境自动测试与自我修复"] end subgraph Deploy["多云部署与管控"] D --> E["合规审计与权限控制"] E --> F["多云自动化部署"] end subgraph Analytics["全流程运营监控"] G["实时分析大盘<br>(成本支出 / 使用率 / 效能分析)"] -.-> Dev G -.-> Pipe G -.-> Deploy end ``` IBM Bob 的核心价值不仅在于端到端打通了从需求规划、企业级 GitHub 集成、自动化测试到多云部署的链路，更在于其内置的分析引擎（Analytics）与治理能力。管理层可以清晰掌握 70,000 名开发者的实时使用数据，明确成本消耗分布与效能产出。在质量保障环节，当 QA 环境出现测试异常时，Agent 能自动感知故障并触发修复流水线，实现真正的闭环交付。 此外，IBM Bob 原生深度融入 IBM 生态系统，对大型机（Mainframe）、IBM i（前身为 AS/400）、Java 以及 WebSphere 等传统企业级架构提供开箱即用的支持，同时持续向外部多元生态拓展。 - **全角色协同**：覆盖开发者、DevOps 工程师、业务分析师（BA）与合规官（Governance Officer）。 - **自动化自我修复**：QA 环境测试报错时，Agent 可自动感知并执行修复 Pipeline。 - **深度生态兼容**：无缝支持 Mainframe、IBM i (AS/400)、Java、WebSphere 等关键遗留与现代架构。 --- ### [05:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=304s) 安永（EY）的实战引入：将 AI 视为团队成员 Liam 邀请了安永（EY）美洲区税务平台与产品董事总经理 Venkat Venkatesan 上台进行炉边对话。Venkat 负责安永旗下庞大的税务技术平台，该平台承载了 40 多个应用程序并服务于众多全球客户。在参与了 IBM Bob 的内测（Private Preview）后，安永团队不仅将其用于加速构建与研发，更将其定位为一名不可或缺的"团队成员"（Team Member），而非单纯的替代型工具。 Venkat 强调，引入 IBM Bob 给安永带来的最大改变之一是将治理（Governance）化为日常工程流程的有机部分，"合规不再是每周二大家必须坐在一起开的漫长例会，而是融入了生命周期的每一步"。为了激发团队热情，安永内部还引入了名为 "Bob Coins" 的机制，工程师们以此展开良性互动竞赛，大幅提升了开发团队的技术抱负与参与感。 - **业务规模**：安永税务平台包含 40 余款企业级应用，服务大量全球核心客户。 - **治理范式转移**：从每周定期开会的被动式合规评审，转变为嵌入日常交付流的自动化治理。 - **文化激励机制**：团队借助 "Bob Coins" 建立良性竞赛机制，驱动研发人员积极拥抱 Agent。 --- ### [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) 汽车的三档变速：Ask、Plan 与 Agent 模式 谈及如何正确使用 IBM Bob，Venkat 将其比喻为一辆拥有三个挡位的汽车，各模式分工明确，贯穿不同的研发场景： | 模式名称 | 核心功能与工作流 | 用户交互特征 | |:--- |:--- |:--- | | **Ask 模式** (Ask Mode) | 深度透视代码库，解答业务逻辑、架构设计与历史代码意图。 | 纯只读问答，不对代码库做任何修改。 | | **Plan 模式** (Plan Mode) | 针对平台变更提出改动方案与架构建议，生成实施蓝图。 | 双向探讨与博弈：人类开发者评估方案、提出质疑并迭代方案。 | | **Agent 模式** (Agent Mode) | 实际执行代码构建、重构实施、测试计划编写及部署协同。 | 全自动化编码与构建执行。 | 在探讨 AI 带来的价值时，Venkat 明确指出"代码提效只是入场券（Table Stakes）"，更深远的价值在于**遗留代码现代化（Code Modernization）**与**代码知识资产沉淀**。大型企业普遍沉淀了数十年历史的代码库，以往由于成本高昂没人敢碰；IBM Bob 能够深度理解 20 年前写下的业务逻辑及其原因，极大地辅助了架构师进行系统重构规划。 此外，在团队新人上手（Onboarding）方面，Bob 扮演了资深导师的角色，避免了老工程师耗费整整一周时间带教新人的负担。针对速度提升可能带来的质量与遗漏风险，Venkat 强调 IBM Bob 绝非简单的"代码补全提示词工具"，它在生成代码的同时会给出完整的测试覆盖计划（Test Coverage Plan），端到端覆盖构建、测试与部署，确保在提速的同时牢牢把控质量底线。 ### [10:59](https://www.youtube.com/watch?v=JjP1052QXZU&t=659s) EY 作为“零号客户”的演进历程与 Bobalytics 治理平台 安永（EY）作为 IBM Bob 的“零号客户”（Client Zero），深度参与了该产品从早期私有预览（Private Preview）到正式商用（General Availability, GA）的全过程演进。Venkat Venkatesan 强调，EY 长期使用多种 IBM 产品，团队在与 Bob 共同成长的过程中获得了极大的技术价值。团队对即将全面推出的 Bob Analytics（分析平台）反馈极为热烈，该平台能够让团队即时洞察 Bob 的实际调用情况与运作机制。 ```mermaid flowchart TD accTitle: IBM Bob and Bobalytics Architecture subgraph BobExecution["Agent 执行层 (IBM Bob)"] direction LR A["端到端 SDLC 任务"] --> B["代码现代化重构"] B --> C["自动化测试与验证"] end subgraph BobalyticsGov["治理与洞察层 (Bobalytics)"] direction LR D["支出监控 (Spend Tracking)"] E["Token 消耗计量"] F["高频任务与工作负载分析"] end BobExecution --> BobalyticsGov ``` Liam Allan 进一步阐释了统一管理平台 **Bobalytics** 的核心能力。在企业级软件研发生命周期（SDLC）中，仅有 AI 智能体本身还不够，必须配备严格的企业治理机制。Bobalytics 提供了全方位的可观测性看板： - **支出追踪（Spend Tracking）**：全面掌握各团队在 AI 辅助开发上的财务支出。 - **Token 消耗监控**：精确计量与分析 Token 使用效率与峰值规律。 - **工作流洞察**：直观展示团队最常处理的代码库及高频任务模块，通过单一平台实现整个 SDLC 阶段的治理闭环。 --- ### [12:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=724s) 未来 3 至 5 年：从“操作 Agent”向“管理 Agent”的范式转移 展望未来三到五年的行业格局，Venkat Venkatesan 认为当前将 Agent 视为简单任务自动化工具（Task Automation）的认知很快会被颠覆。这就如同现在人们在任何场所连 Wi-Fi 都无需思考底层网络连接原理一样，AI Agent 将成为无缝融入日常协作的默认基础设施。 未来的核心分工将发生根本性变革：人类不再是去具体“操作”某款软件，而是像团队管理者一样**调度与统筹 Agent（Managing Agents）**，协调多个 Agent 与人类工程师协同完成日常交付。在战略层面，人类的核心价值将收敛于回答“**为什么（Why）**”——审视为何要构建某个应用或业务系统；而 Bob 等 AI Agent 则全面接管技术落地的“**如何实现（How）**”。 --- ### [13:20](https://www.youtube.com/watch?v=JjP1052QXZU&t=800s) 开发者角色的重塑：从初级作者到架构师与“编辑者” 谈及开发者的职业轨迹，Venkat Venkatesan 指出这场颠覆实际上为所有工程师带来了一场“自动化晋升”。传统的开发者角色更像是一个从零敲代码的“原作者”，而在 Agent 普及后，开发者转型为**架构师（Architect）与主编（Editor）**，其核心职责变为审查、校验并挑战 Agent 生成的产物。 ```mermaid flowchart LR accTitle: Developer Role Evolution subgraph Traditional["传统模式"] Dev1["开发者 / 原始作者"] -->|"手动编写代码"| CodeBase["代码库"] end subgraph Agentic["Agentic AI 模式"] Agents["AI Agents (如 IBM Bob)"] -->|"自动化生成 / 提议"| Draft["候选方案与代码"] Draft --> Dev2["开发者 (架构师 / 主编)"] Dev2 -->|"多轮辩论与审查 (Review & Refine)"| ValidatedCode["生产级代码交付"] end ``` Liam Allan 对此高度赞同，并指出日常研发中开发者与 Bob 之间的高频交互已成为常态。很多工程师会就架构决策与 Bob 展开激烈的双向推敲：“你为什么推荐这种实现？这里不应该这样设计。”这种无需临时拉同事开会、随叫随到的“口袋顾问”协作模式，极大提升了研发决策与落地的敏捷度。 - **职责升级**：从低效的模板代码敲写转向系统架构设计与最终结果审查。 - **人机协同推敲**：利用高频反馈循环不断校准 AI 生成逻辑，确保业务代码的准确性与鲁棒性。 --- ### [14:27](https://www.youtube.com/watch?v=JjP1052QXZU&t=867s) 企业落地的真正瓶颈：超越技术采购的组织变革与文化重塑 当被问及企业高管对 AI 最常见的误解时，Venkat Venkatesan 直言：**技术的迭代速度已经极为成熟且每天都在进化**（每周一都有新突破，Bob 本身也在持续迭代），然而许多领导者误以为引入 AI 仅仅是一项单纯的技术“采购决策（Procurement Decision）”。 真正决定企业 AI 落地成败的痛点在于**技术采纳（Adoption）、变革管理（Change Management）与企业文化重塑（Culture Change）**。企业当下的重心应当放在如何将 AI 真正引入团队的现有工作流，消除员工对新工具的抵触，建立人机协作的文化土壤，而非仅仅停留在采购工具软件的表层阶段。 --- ### [15:21](https://www.youtube.com/watch?v=JjP1052QXZU&t=921s) 现场体验与结语：如“巴布工程师”般的实战交付 对谈尾声，Venkat Venkatesan 幽默地借用经典动画《巴布工程师》（Bob the Builder）的梗，调侃 IBM Bob 就像戴着工程安全帽的超级建造者一样，为现代化改造提供扎实可靠的技术支撑。 Liam Allan 在展会大屏幕上展示了现场互动的二维码，并热情邀请参会者前往 IBM 展台体验实机操作。团队在展位准备了多台笔记本电脑，现场演示如何利用 IBM Bob 从零开始端到端构建、现代化改造并部署应用程序，让现场工程师亲身感受 Agentic AI 带来的研发范式革新。 ## Notable quotes > [05:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=304s) 合规不再是每周二大家必须坐在一起开的漫长例会，而是融入了生命周期的每一步 — Venkat Venkatesan > [05:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=304s) 团队成员 — Venkat Venkatesan > [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) 代码提效只是入场券（Table Stakes） — Venkat Venkatesan > [07:08](https://www.youtube.com/watch?v=JjP1052QXZU&t=428s) 代码补全提示词工具 — Venkat Venkatesan > [12:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=724s) 调度与统筹 Agent（Managing Agents） — Venkat Venkatesan > [12:04](https://www.youtube.com/watch?v=JjP1052QXZU&t=724s) 人类的核心价值将收敛于回答‘为什么（Why）’，而 Bob 等 AI Agent 则全面接管技术落地的‘如何实现（How）’ — Venkat Venkatesan > [13:20](https://www.youtube.com/watch?v=JjP1052QXZU&t=800s) 自动化晋升 — Venkat Venkatesan > [13:20](https://www.youtube.com/watch?v=JjP1052QXZU&t=800s) 架构师（Architect）与主编（Editor） — Venkat Venkatesan > [13:20](https://www.youtube.com/watch?v=JjP1052QXZU&t=800s) 你为什么推荐这种实现？这里不应该这样设计。 — Liam Allan > [14:27](https://www.youtube.com/watch?v=JjP1052QXZU&t=867s) 采购决策（Procurement Decision） — Venkat Venkatesan > [14:27](https://www.youtube.com/watch?v=JjP1052QXZU&t=867s) 真正决定企业 AI 落地成败的痛点在于技术采纳（Adoption）、变革管理（Change Management）与企业文化重塑（Culture Change） — Venkat Venkatesan ## Mentioned Liam Allan, Venkat Venkatesan, IBM, EY (安永), IBM Bob, Bobalytics, Bob Coins, GitHub, Mainframe (大型机), IBM i (AS/400), Java, WebSphere, 《巴布工程师》(Bob the Builder), Ai4 ## Video description Liam Allan - Software Architect, Speaker & Advocate, IBM Venkat Venkatesan - Americas Tax, Platform & Products, EY Enterprises have spent the past two years experimenting with AI coding tools. Now the mandate has shifted from speeding up snippets to realizing governed, enterprise-scale transformation. In this session, we explore how IBM Bob, an AI-powered SDLC partner, transforms application modernization into a streamlined, automated, and scalable process. Going beyond traditional code assistants, you need a partner that analyzes entire codebases, generates modernization strategies, and validates outcomes with built-in governance and testing. Join us for a fireside chat with Venkat Venkatesan, Managing Director for Platforms, Products & AI at EY, as she shares EY's experience leveraging IBM Bob to accelerate application modernization and drive business transformation at scale. We will walk through a full modernization journey. Give insights into how an agentic AI approach can reduce modernization timelines from months to days, improve code quality, and drive enterprise transformation at scale. [Source](https://www.youtube.com/watch?v=JjP1052QXZU)

Source quotes

[5:04] Venkat Venkatesan: 合规不再是每周二大家必须坐在一起开的漫长例会，而是融入了生命周期的每一步 [5:04] Venkat Venkatesan: 团队成员 [7:08] Venkat Venkatesan: 代码提效只是入场券（Table Stakes） [7:08] Venkat Venkatesan: 代码补全提示词工具 [12:04] Venkat Venkatesan: 调度与统筹 Agent（Managing Agents） [12:04] Venkat Venkatesan: 人类的核心价值将收敛于回答‘为什么（Why）’，而 Bob 等 AI Agent 则全面接管技术落地的‘如何实现（How）’ [13:20] Venkat Venkatesan: 自动化晋升 [13:20] Venkat Venkatesan: 架构师（Architect）与主编（Editor） [13:20] Liam Allan: 你为什么推荐这种实现？这里不应该这样设计。 [14:27] Venkat Venkatesan: 采购决策（Procurement Decision） [14:27] Venkat Venkatesan: 真正决定企业 AI 落地成败的痛点在于技术采纳（Adoption）、变革管理（Change Management）与企业文化重塑（Culture Change）

Source quotes

[0:04] Liam Allan: Moving software smoothly and automatically from local test environments to QA and cloud deployment pipelines without manual bottlenecks. [7:08] Venkat Venkatesan: Bob operates like a car equipped with three distinct gears—Ask Mode, Plan Mode, and Agent Mode. [8:16] Venkat Venkatesan: productivity is table stakes. [8:16] Venkat Venkatesan: Many enterprise codebases contain mission-critical business logic written twenty years ago by engineers who have long since departed. [8:16] Venkat Venkatesan: Bob is not a simple line-completion or comment-finishing tool; it actively constructs comprehensive test coverage plans and provides clear rationales for every change across the build, test, and deployment phases. [10:53] Venkat Venkatesan: Client Zero [10:53] Venkat Venkatesan: shifting their day-to-day focus from raw syntax authorship to high-level architecture, editing, and rigorous validation. [10:53] Venkat Venkatesan: successful AI deployment is an organizational adoption and culture-change challenge, not a simple procurement decision.

Original transcript

[0:04] I'm actually quite excited to be here today  because I get to talk about the product [0:07] that I work on day-to-day. I'm not only as an  engineer, but also as a PM for extensibility, [0:13] making Bob work in places that it wouldn't usually  go, and that's what makes this so exciting for me. [0:19] So, I want to talk a little bit about what we  IBM discovered in 2025 as we were rolling out [0:24] AI development tools across the company,  right? IBM is more than 150,000 people today. [0:31] That's a lot of people, and I would say 70,000 of  those are developers. So, what does that mean for [0:36] the rest of them, right? 70,000 developers, but  70,000 more non-developers. How do they use AI [0:42] tools? What we've learned is that developers can  do a lot of good work with AI. But, there's some [0:48] parts that's lacking. For example, in spend, how  do we track spend across 70,000 developers, right? [0:56] How do we track how they work with their  infrastructure? When IBM is deploying apps [1:01] into our own services, into our own cloud, how do  we track the price of where AI is used there? How [1:07] do we track the governance with our own employees?  We work with the government. We work with banks. [1:12] We work with airlines. How do we track the agentic  use from our developers with those business units? [1:19] That's quite important. It's the part that we've  missed out on in 2025. Going forward this year, [1:25] that's where we're starting to aim, and that's  what makes me quite excited about this, right? [1:29] What we've discovered is that most of the  time developers aren't deploying stuff. [1:34] They just write code. They make a plan.  They implement it. They can test it locally, [1:39] and they can test it in their own cloud, right?  That's great, but the part that's missing is [1:44] how do we go from developing, testing locally  to testing in a QA environment automatically? [1:51] How do we make it easier for the business to  deploy faster? How do we make it easier to govern [1:57] how the deployments are working? Making sure  that only certain people and agents can access [2:02] that stuff. That stuff's really important for us.  So, while developers have done that and then we've [2:08] seen that in IBM, we're really hoping to build  a singular product that can cover the full SDLC. [2:15] Now, I say Bob because I have a British accent.  Some Americans say Bob. IBM Bob. Right? That's the [2:24] that's the thing. But it's Bob. IBM Bob. We are  building a tool and we have it today. You can [2:29] actually use it today that covers the full SDLC.  So, whether you are building an app, writing code, [2:37] or whether you are an engineer doing DevOps  work, whether you are a governance officer, [2:43] you can use a single tool across the board. That's  right. And not only that, the cool part about [2:50] using the singular tool is that it comes with  the governance. It comes with the analytics. [2:55] You can see, for example, how those 70,000  developers that we have are using the tool across [3:02] the board. Whether you're a developer or you know,  a BA. We can see the usage. We can see where it's [3:08] costing more, costing less, who's doing the most  work, who's not doing any work. We can see all of [3:13] that through our analytics. Um and that's really  important for us. So, across the business now, [3:20] 150,000 daily users. I'll talk more  to that in a minute. But in IBM today, [3:26] we go from dev to test through using Bob with  agentic tools, planning, integration with GitHub [3:34] in enterprise and our own tools, all the way  to deploying into many different clouds with [3:39] governance, our own governance, using a single  tool. And that goes a long way for our business. [3:46] We actually know this is the crazy crazy part is  that we can deliver faster for our customers and [3:54] ourselves using a singular tool. Mostly because  of the analytics that we get from using this tool. [4:00] We can see, for example, deployments that  go wrong. An agent is knowing and being able [4:06] to react to when things go wrong. Not only for  developers, but all the way from dev to test. When [4:12] a test goes wrong in QA, we can have pipelines  that automatically run to fix that using this [4:17] singular tool. And that's really nice for us. Now,  I actually get to work on this tool. I get to work [4:23] directly with customers. I'm here today to work  with customers and that's really the nice part. [4:29] Um I'm going to have someone up in just a moment  to talk about it. Um but if you're an existing IBM [4:34] customer and you use any of our existing products,  the best part is Bob fits in really nicely with [4:40] the IBM ecosystem. So, for example, if you are  a mainframe customer, if you are an IBM i, or [4:46] as they previously called it AS/400 customer, if  you have Java, if you have WebSphere, we work very [4:53] nicely with our existing ecosystem and we want  to keep to that. And in the future, we're going [4:57] to be expanding to more ecosystems even outside  of IBM. And that's what makes this so exciting. [5:04] Now, um I'm going to bring up a special  guest, Venkat Venkat Venkatasan. [5:09] Um she works for EY and I'm going  to ask her a couple of questions. [5:12] Please everyone, let's give a round  of applause for Venkat. [applause] [5:20] How do you do? Good, thank you. good. Meet you  all. I'm very grateful you're here. I know you [5:25] flew in just this morning, so I am really grateful  that you're here, right? Um I want to start with a [5:30] a very simple question. I'm I'm a professional and  I brought my iPad with me. Um can you kind of tell [5:36] me um what caught your attention with Bob? How did  you first see it at EY? Um so, in EY, my role is [5:43] to manage our tax platform, uh which is um like  I know tax is kind of a different area, but we do [5:50] a lot of technology in tax. And we manage a tax  platform which has around 40 plus apps in it, uh [5:57] manages several clients are part of it. And one of  the things that we started using Bob is for like [6:03] how can we accelerate our build and development  using Bob? Uh we were part of the private preview, [6:09] which was fantastic, which helped us to understand  more about Bob. And now Bob is a team member. [6:15] I love that description. Bob is a team member.  Are you finding that Bob is working with your [6:20] developers, with your teams, and not really trying  to replace them, right? It's trying to be part of [6:24] the team, right? So, that's the one thing. Like  if I have to say what has Bob helped us achieve, [6:29] there is speed, there is governance, which became  a part of our like it's not a call that we have [6:35] to every Tuesday you have to get on a governance  call. It became a part of our life cycle. That has [6:40] been the one biggest impact for us like because  of having Bob. And it has made our developers [6:46] very ambitious. Like um there is some concept  called Bob coins. So, most of our developers are [6:51] about like how many Bob coins that everyone's  using. They get to have a friendly competition [6:55] among them. Yeah. It has been really great to  use in that regard. I love that. Um I want to [6:59] just mention another thing. You mentioned that  you've used Bob in the life cycle. Are you Can you [7:04] kind of speak to Bob being used outside of just  being a development tool a little bit? Uh yes, [7:08] absolutely. So, one thing is uh for anyone who is  new to Bob or who is going to start using Bob, Bob [7:14] is like your like any car people here? It's like  a car, right? It has three gears in it. Um you [7:19] They have the ask mode and agent mode and the plan  mode. Ask mode is fantastic where you go and say, [7:26] "Hey Bob, look into the code, uh ask you answer  your questions. Hey, what is happening here? This [7:31] is something that is happening." It doesn't even  do anything. Plan mode is even fantastic where you [7:36] get to plan how you want to use Bob. Okay, I want  to do this change in my platform, in my code base. [7:43] Can you plan for me? Suggest me how I should  do it. And that is more of me being challenged [7:50] by my Bob telling me, "Hey, this is what you  should do." And then I go back and say, "No, [7:55] this is not I don't agree this is the approach."  So, it's my more of a back and forth. And the [7:59] agent mode is the actual build where that happens.  It actually builds things for me. So, it's all the [8:05] three. That's fantastic to work with. I'm I am.  That's really great. Um I'm going to pivot this [8:10] one a little directly. So, is productivity the  only thing that you get from using Bob? Is there [8:16] more to it than just being faster? Um that's great  question. So, I would call productivity as table [8:22] stakes. It's It's there. Productivity  because once you have utilized AI agent [8:27] agent to build with you, it happens. But there  are so many other things that we miss out. Like [8:33] one is the code modernization. Most of the firms,  everybody has lots of code that is there. And we [8:38] don't even think about changing them because those  are like really highly expensive projects. And [8:44] having Bob helps that a lot. That's number one.  The other thing I would call is more about like [8:51] the knowledge. There are so much knowledge in the  code base. No one knows what was written 20 years [8:56] before. It was built by someone. We get to use  it what happens. And having Bob is a big impact [9:03] because now I know why something was written, what  was written, and before going and making a change. [9:08] It helps me as someone who manages the platform  plan better. That's lovely. And you you actually [9:14] mentioned about code that's quite old, right?  Would you say that Bob is actually very good at [9:17] understanding business logic logic over time? Is  it something that it can help your team or bring [9:22] new people on? Is that something? It's more on the  latter side. It helps in bringing new people on [9:28] much faster because now my existing team, like my  developers are great, fantastic. But they don't [9:35] want to spend the time on boarding another  developer for a week or anything. And that's [9:40] where Bob helps a lot because like now Bob is  that person who helps the new team to onboard [9:45] quickly to be able to achieve. We do not have  much of a legacy code to focus on. That's great. [9:51] That's still also something that we keep an  open mind towards like using Bob for. Okay. [9:56] How does productivity risk come into play when  you're using AI? Because if it makes you faster, [10:01] where is the risk in missing something? Is that  something that you consider? So, I think that's [10:05] where you talked about and I liked what you were  saying earlier, right? It's the governance that [10:09] needs to be part of the life cycle. And Bob is  not someone who's just making changes to the code. [10:15] It tells you how to test it. There is the test  coverage plan and everything that it comes up [10:19] with. So, the productivity gain is not just like  I'm not saying and if any coders are over here, [10:25] they would know. I'm not asking Bob to say, "Hey,  finish my sentence. Finish my comment." Anything. [10:30] That's not what Bob is doing. What Bob is telling  is like, "Hey, this is what I'm doing. This is why [10:35] I'm doing. You tell me how." So, Bob gets to take  care of all the parts of it. Build, test, deploy, [10:43] all the sites around it. Yeah. So, if you know,  many businesses today are trying lots of different [10:49] tools, but what you know, for EY, what was the  singular point that said, "Let's try out Bob." [10:53] Like, how do we how did we get there? Um so, EY  is a broader organization. From the tax side, [10:59] we do have like a very excellent client. Like we  are we are one of the client zero of Bob. And then [11:06] we also work with IBM a lot. So, that's one of  the places that how we started on. We have used [11:10] several other IBM products. And as we started with  Bob, it was a journey. As we started using it, [11:17] we get to see Bob's growth as well. From  private preview to now it's generally available. [11:22] Bob analytics that's coming soon which I like.  It's there. It's so fantastic. That's what my [11:27] team says. I haven't used it myself, but they say  they they to see how Bob is being used. That's the [11:32] part that is like they get to have real insight  on their usage and how Bob is working right away. [11:38] Maybe I could touch on Bobalytics just a little  bit. So, I mentioned earlier about governance and [11:42] tracking spend. Well, our analytics platform is  called Bobalytics. So, not only are you using Bob [11:49] as the agent tool, you are also using  Bobalytics to track how spend is going, [11:54] token usage, what's being worked on the most,  right? And that's something that comes with [11:58] using a singular tool for your for the  SDLC, basically. I'm curious what you think [12:04] if we were going to have this conversation in 3  years or maybe even 5, you know, what do you think [12:09] some of the bigger changes will be in the  industry? Um, that's that's a great one, huh? So, [12:14] I think right now we are all thinking about agents  as task automation. We do have several agents. [12:20] Like here, let's have this agent to automate  the specific task. Just like how I don't know. [12:25] We ask for Wi-Fi password anywhere. We don't  even know that like how to connect to internet. [12:29] That's not a thing anymore. That was once. I  believe that's how our future is going to be. [12:34] We will have AI agents working with us to solve  this. And we won't be operating those agents. We [12:41] will be managing these agents. Because operating  agents is very similar to operating a software. [12:47] But having agents over here, it's going to be like  we are going to be managing them, conducting their [12:52] daily work along with the teams. That's how I see  it how like 3-5 years from now it's going to be. [12:58] And more of humans, we are going to think like why  do I want to build this? Is this something is this [13:04] the app that I want to build? Is this a software  that I want to build? That's where our focus is [13:09] going to be more towards where Bob and AI agents  are going to help more towards like the how side [13:15] of things. Okay. And then as a follow-up to that,  where do you think the role of developer is going? [13:20] Do you think it will change much? Do you think it  will be extended? It's a great disruption and it's [13:26] very similar to like you are like someone starts  as an author. Now the role of developer is more [13:34] towards an architect. They are an editor. They get  to see what is being coming and they all got this [13:40] automated promotion as they get to use bar because  everybody eventually wants to challenge what is [13:47] being done, how it is being handled. So the role  of a developer is slowly evolving towards being an [13:54] editor architect who manages the team of agents.  Yeah, I I actually expect a lot of my developers [14:00] to be more reviewing and making sure that what the  agents are doing is more accurate. I I actually [14:05] fully agree with you. I I have one developer  who says that they had a major conversation [14:10] with Bob and it's more about like why are you  suggesting this? No, this is not how it should be. [14:15] So that back and forth really helps a lot and it's  more like having someone at pocket dial anytime [14:22] you want. You don't have to get on the call with  another developers. Yeah. So what do you think [14:27] that leaders in general for businesses get wrong  about AI? Is there Is there still some misguidance [14:32] when it comes to using AI or is there something  that they're not seeing? What do you think? Yes, [14:37] so because we do also build a lot of AI agents and  deploy. I think I strongly believe the technology [14:44] is there and the technology is also changing every  day. Every Monday there is a new thing. Every [14:49] quarter things are improving. Bob is an example.  Bob changes every day. Bob improves a lot. [14:54] Technology is there. However, utilizing AI  is not just a procurement decision. It goes [15:00] beyond it. There is a lot in the adoption  and change management and culture change. [15:05] That is where I see we are spending a lot of  time right now on. The technology is there. [15:10] How do we adopt the technology? How do we bring  it into our environment to work with our teams? [15:15] The culture change and change management. That's  where the focus is primarily. That's great. Um now [15:21] I'm going to cut it short. I I've really asked you  such a wide range of questions and you've been a [15:25] a great host. Is there anything final that you  would like to add? Thank you. It was so nice to [15:29] talk about Bob and anyone here like and since you  said that you're from you have a different accent, [15:34] Bob the Builder song, that's literally how  Bob is. It's so fantastic to work You like [15:38] the hard hat with the little making you look  like Bob the Builder. Okay, that's great. Thank [15:42] you. I appreciate your time. Um, I would like to I  think I hope the QR code will come up. Yeah, good. [15:51] So, I want to say if you would like to see Bob  in action today, I am and my team are going to [15:55] be at the booth and we can do hands-on and  we can go end-to-end deploying something, [16:00] building something together and we can even  have laptops so you can try it yourself. [16:04] And I'd love to see you there. Thanks so  much for coming and I appreciate your time.