内容来源：https://aibusiness.com/agentic-ai/bob-ibm-ai-coding-assistant

内容总结：

IBM押注生成式AI：发布智能编程助手“鲍勃”，聚焦混合云与垂直领域应用

在近日于波士顿举行的IBM Think 2026用户大会上，这家拥有115年历史的科技巨头正式发布了其生成式AI编程助手“鲍勃”（Bob），并将其定位为企业迈入AI辅助软件开发领域的便捷入口。该工具被形容为一位“支持型、协作型”的AI编码伙伴，能够在将最终决策权留给人类的同时，自动化大量软件开发流程。

据安永全球税务及咨询业务高级经理文卡泰桑介绍，其团队在私人测试阶段已使用鲍勃数月。她将鲍勃比作“数字工作者”，强调其与传统“结对编程”类似，人类开发者与AI像同事一样协同工作。鲍勃的核心能力在于能根据任务性质，自动将编码需求路由至Anthropic的Claude模型、法国AI公司Mistral的开源模型以及IBM自家的Granite轻量级模型，体现了IBM多模型、多云的战略思路。

大会期间，IBM董事长兼CEO克里希纳将AI置于公司未来战略的核心。他透露，IBM已将AI和自动化应用于自身全部运营，并由此实现了40亿美元的生产力提升。他表示，客户关注的焦点已从“预算有多少”转变为“AI在业务流程中嵌入得有多深”。

除鲍勃外，IBM还发布了150个预构建AI代理以支持混合云和大型机环境，并宣布了Watsonx与流数据平台Confluent的集成。分析人士指出，IBM明智地避开了与主流大模型厂商的正面竞争，转而专注于为金融、农业等高度监管行业的忠实客户提供轻量化、垂直化的AI解决方案。SanjMo咨询公司创始人莫汉认为，IBM聚焦细分客户需求的策略，而非打造通用模型，是正确之举。

与此同时，大型金融服务公司SEI的CTO沃马克表示，该公司正计划与IBM咨询合作，构建一批AI代理以改造其基于云的业务流程。他强调，SEI也采取了多模型策略，未来可能同时使用Watsonx、Claude或OpenAI的模型。不过，沃马克也坦言，目前AI项目的投资回报率（ROI）尚待观察，“人们仍在评估AI所承诺的愿景”。

大会的另一位亮点是退役网球巨星阿加西。他登台介绍了其与IBM合作开发的、基于Watsonx的移动端网球数字教练应用。阿加西表示，AI的能力让他看到了以创新方式深度提升网球运动的巨大潜力。

中文翻译：

由谷歌云赞助
选择您的首个生成式AI应用场景
要开始使用生成式AI，首先应聚焦于那些能够改善人类与信息交互体验的领域。
对于这家历史悠久的科技供应商来说，软件开发生命周期平台是企业进入AI编程领域的一个便捷切入点。

波士顿——来认识一下Bob，一个兼具支持性与协作性的AI编程助手，它能自动化大部分软件开发流程，同时让人类掌握主导权。
“它就是‘巴布工程师’，”全球税务与咨询公司安永税务技术与转型实践高级经理文卡特·文卡泰桑说道，她亲切地引用了一部关于建筑工人和他那帮会说话的机器的著名英国动画电视节目。“当我第一次开始使用Bob时，我就知道它不仅仅是一个简单的编程助手。”

可以说，在IBM Think 2026用户大会开幕前约一周发布的Bob，是本次大会的明星，它出现在各种小组讨论和演讲中，甚至出现在这家老牌科技供应商董事长兼首席执行官阿尔文德·克里希纳的开场主旨演讲里。

在某种程度上，Bob是IBM对标Claude Code和Codex的产品。Claude Code是来自IBM合作伙伴兼竞争对手Anthropic的广泛使用的编程代理，Codex则来自另一家生成式AI领导者OpenAI。

不同之处在于，Claude是Bob核心使用的AI模型之一。Bob会根据任务性质，将编程任务分配给Claude、法国AI供应商Mistral的开源模型，以及IBM自家的轻量级模型系列Granite。

文卡泰桑和她的团队正在为安永构建一个庞大的全球税务平台。在IBM于上月底将Bob正式公开发布并在大会上向全球介绍它之前，他们已经私下试用Bob的测试版好几个月了。

“我称它为智能代理，”文卡泰桑在周二（大会正式开幕日）的一次采访中说道。“它在每个阶段都能帮到你。就像你和它一起工作一样。”

在编程界，结对编程是一种传统实践，即两名开发者共享一个工作站，共同协作编写代码。
“Bob就像那样。它就像一个数字员工。你们俩一起工作，”文卡泰桑说道。

拥有自己吉祥物的Bob，当然不是本周展出的唯一IBM AI新作。

克里希纳将AI定位为这家拥有115年历史的公司未来战略的核心。他在主旨演讲中表示，IBM自身已在其所有运营中应用了AI和自动化，并实现了40亿美元的生产力提升。

“当你与不同的客户交谈，与不同地区和行业领域的人交流时，会发现这是一个巨大的变化，”克里希纳说。“这不再关乎你的预算是多少。问题归结为：AI在你的业务流程中嵌入得有多深？”

其他新进展包括：在Watsonx Orchestrate中发布了150个预构建智能代理，用于混合云和大型主机环境，以及对Concert AIOps平台的重大扩展。IBM还推出了新一代Watsonx Orchestrate代理管理系统，并在以110亿美元收购Confluent后，整合了Watsonx和Confluent的流数据平台。

对于许多观察者而言，IBM明智地选择战略性扩展其生成式AI产品（以Granite和Watsonx模型系列为主导）至混合云和大型主机，同时保留众多云AI产品和服务，并坚持多模型、多云策略。

IBM与全球许多最大、最古老的金融机构以及其他受严格监管行业的公司有着长期的合作关系，这些公司非常看重大型主机的数据隐私和安全性。IBM是一个值得信赖的品牌，拥有忠实的客户群，这些客户正在同步推进本地IT运营和更现代化的云与AI技术。

咨询公司SanjMo的创始人兼分析师桑吉夫·莫汉在大会接受采访时表示，对于仍在制造和销售大型主机硬件及软件的IBM来说，大型主机仍然是一个利润中心。

“如果你是一家金融服务公司或农业公司，对于全球70%的交易，一切仍然通过大型主机流转，”莫汉说。“大型主机是一个增长中的业务，而非垂死的业务。”

同样，莫汉认为，IBM决定不与最大的生成式AI供应商展开竞争，而是专注于更小、更轻量的模型，这也是正确的。

他说：“IBM的意思是说，在如此激烈的竞争下，‘如果我们专注于客户所在的非常细分领域，那么我们就可以基于需求所在之处开辟一条道路，而不是创建一个通用模型。’”

与此同时，另一家IBM客户——宾夕法尼亚州的大型金融服务公司SEI——的IT部门则不走大型主机路线，而是计划与IBM的专业服务部门IBM Consulting合作，构建一批AI代理。不过，该公司的会计部门仍然在IBM大型主机上运行。

IBM Consulting最近在SEI的一次竞争性招标中胜出，将为涵盖众多业务流程的基于云的操作设计代理，包括替换陈旧的光学字符识别文档系统。

“他们来是为了真正帮助我们自下而上地重新审视我们的工作流程，重新设计这些工作流程，在适用之处应用AI，并可能构建出这些代理，”SEI的首席技术官扎卡里·沃马克说。“这些代理将部署在我们的框架上，可能需要我们尚未拥有的技术。未来，运营将从手动操作应用程序转变为编排智能代理。”

沃马克表示，这些技术可能包括Watsonx模型和代理、Claude或OpenAI的模型。

“我们的方法绝对是多模型的，”他说。“所有这一切都将成为我们持续构建的‘马具’的一部分。”

至于投资回报率，这还有待观察，正如SEI将用来衡量它的指标一样。

“现在还为时尚早。关于投资回报率的问题问得很好，”沃马克继续说道。“我认为人们仍在评估那里存在的潜力。”

IBM的一位客户是退役网球巨星安德烈·阿加西及其体育娱乐公司。

阿加西周二出现在主舞台上，谈论其由Watsonx驱动的球拍运动数字教练移动应用程序，该应用计划于今年晚些时候发布。

阿加西表示，在他打球时，准备是克服他身体限制的关键，因为在职业网球巡回赛中，他既不是身材最高大也不是速度最快的选手。

他说，他必须依靠他的教练和训练师来帮助他保持高水平的发挥。

“而现在，当你突然看到AI的能力，看到我与IBM的合作，你开始意识到……我们有许多方法可以利用它来以美妙的方式提升这项运动，并将其推向深远的未来，”他说。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
For the long-established tech vendor, the software development lifecycle platform is an accessible entry point for enterprises into the realm of AI coding.
BOSTON -- Meet Bob, the supportive and collaborative AI coding helper that automates much of the software development process while leaving the human in charge.
“It’s Bob the Builder,” said Venkat Venkatesan, senior manager in the tax technology and transformation practice at EY, the global tax and consulting firm, affectionately referring to the well-known animated British TV show featuring a builder and his helpful talking machines. “When I first started using Bob, I knew it was not a simple coding assistant.”
You could say that Bob, released about a week before the start of IBM’s Think 2026 user conference, was the star of the show, popping up in panels and presentations and even the opening keynote by Arvind Krishna, the long-established tech vendor’s chairman and CEO.
In a way, Bob is IBM’s answer to Claude Code, the widely used coding agent from IBM partner/competitor Anthropic, and Codex from OpenAI, another generative AI leader.
Except that Claude is one of the AI models at the heart of Bob, which routes coding tasks to Claude, open source models from France-based AI vendor Mistral, and Granite, IBM’s own family of lightweight models, depending on the nature of the job.
Venkatesan and her team, who are building out an extensive global tax platform for EY, had been using Bob in private beta for a few months before IBM made it generally available late last month and introduced it to the wider world at the conference.
“I would call it an agent,” Venkatesan said in an interview on Tuesday, the official start of the conference. “It helps you during every phrase. It’s like you’re working with it.”
In the coding community, pair programming is a traditional practice in which two developers share a single workstation to collaboratively write code.
“Bob is like that. It’s like a digital worker. You both work together,” Venkatesan said.
Bob, who has his own mascot, was certainly not the only new IBM AI creation on display this week.
Krishna positioned AI at the core of the 115-year-old company’s go-forward strategy, saying in his keynote that IBM itself has applied AI and automation across all of its operations and realized $4 billion in productivity gains.
“As you talk to different clients, as you talk to different geographies and industry sectors, this is the big change,” Krishna said. “It’s no longer about how much your budget is. The question comes down to, how deeply is AI embedded in your business processes?”
Among other new developments was the release of 150 prebuilt agents in Watsonx Orchestrate for hybrid cloud and mainframe environments and a major expansion of the Concert AIOps platform. IBM also touted a new generation of the Watsonx Orchestrate agent management system and an integration of Watsonx and Confluent’s streaming data platform, after IBM's $11 billion acquisition of Confluent.
For many observers, IBM is wise to strategically extend its generative AI offerings -- led by the Granite and Watsonx lines of models -- to hybrid cloud and mainframes, while retaining many cloud AI products and services and maintaining a multi-model, multi-cloud approach.
IBM has a long tenure with many of the world’s biggest and oldest financial institutions and other companies in highly regulated industries that value the data privacy and security of mainframe computers. It’s a trusted brand with loyal customers that are moving forward with both on-premises IT operations and more modern cloud and AI technologies.
Remarkably for IBM, which still builds and sells mainframe hardware and software, mainframes are still a profit center, said Sanjeev Mohan, founder and analyst at the SanjMo advisory firm, in an interview at the conference.
“If you’re a financial services company or an agricultural company, and for 70% of global transactions, everything still flows through mainframes,” Mohan said. “Mainframes are a growing business, not a dying business.”
Likewise, IBM’s decision not to engage competitively with the biggest generative AI vendors and instead focus on smaller and lighter-weight models was correct, Mohan said.
“What IBM is saying is that, with so much competition, ‘If we focus on very niche areas where their clients are, then we can cut a swath based on where the need is rather than create a generic model,’” he said.
Meanwhile, the IT department of another IBM customer, Pennsylvania-based SEI, a large financial services company, is not going the mainframe route in favor of a fleet of AI agents it plans to build with IBM Consulting, the professional services wing of IBM. The company's accounting section, however, still runs on IBM mainframes.
IBM Consulting recently won a competitive request for proposals at SEI to design agents for cloud-based operations for a multitude of business processes, including replacing dated optical character recognition document systems.
“They’re coming in to really help us reexamine our workflows bottom-up and reengineer those workflows and apply AI where applicable and potentially build out those agents,” said Zachary Womack, CTO at SEI. “Those agents would be deployed on our framework and may require tech that we don’t have. In the future, operations goes from banging away on applications to orchestrating agents.”
That tech could include Watsonx models and agents, Claude or OpenAI models, Womack said.
“We definitely are multi-model in our approach,” he said. “All that will be part of the harness we are continuing to build out.”
As for ROI, that remains to be seen, as are the metrics SEI will use to determine it.
“It’s still early days. The question of ROI is a good one,” Womack continued. “I think people are still evaluating the promise that’s there.”
One IBM customer is retired tennis superstar Andre Agassi and his sports entertainment company.
Agassi appeared on the main stage Tuesday to talk about his Watsonx-powered racket sports digital coaching mobile app, set to be released later this year.
Agassi said that when he played, preparation was key to overcoming his physical limitations as neither the biggest nor fastest player on the pro tennis tour.
He said he had to count on his coaches and trainers to help him perform at a high level.
“And now, all of a sudden, when you start seeing the capability of an AI, my partnership with IBM, you start to realize … we have multiple ways to use this to enhance this game in a beautiful way, and take it deep in the future,” he said.