内容来源：https://aibusiness.com/data-governance/building-resilience-in-the-age-of-ai

内容总结：

企业AI应用：治理与韧性成投资回报关键

随着生成式AI在各行业加速部署，如何选择首个应用场景并确保长期成功，成为企业面临的核心挑战。近期一份行业报告指出，AI项目的投资回报率高低，往往不取决于所选工具本身，而更多与企业对治理和运营韧性的重视程度直接相关。

旧金山IT事件管理供应商PagerDuty发布的报告显示，在受访的1000名IT业务领导者中，近四分之三的企业在过去一年提升了运营韧性，并同步增加了相关预算。数据进一步揭示，在营收增长的企业中，超过四分之三正在加大韧性投资，而营收停滞或下降的企业中，这一比例不足一半。这表明，积极投资韧性的组织正获得更显著的回报，反之则可能迅速落后。

报告警示，缺乏良好治理与韧性措施将导致多重风险。例如，超过半数的英国和爱尔兰企业遭遇重大IT事件时，每小时损失可能超过30万美元，且声誉损害会持续蔓延。

企业面临的另一重压力在于治理体系难以跟上技术迭代的速度。PagerDuty产品开发高级副总裁戴维·威廉姆斯指出：“当前存在一场疯狂竞赛，大家急于部署AI，却未设置适当的防护措施和协调工具。”他强调，AI智能体应像人类员工一样被管理，需明确权限、监督机制和责任归属。

工程副总裁若昂·弗雷塔斯补充，AI极大提升了开发效率，以往需六个月的项目如今可能缩短至十天，但企业必须同步思考“如何保持可靠性与合规性”。

关于AI对人力资源的影响，早期“AI将取代初级岗位”的论调正在修正。首席信息官埃里克·约翰逊表示，人类在回路中仍不可或缺，AI的核心作用是增强员工能力而非简单替代。他预测，伴随AI成长的年轻一代将成为“AI原生”工作者，善于利用AI完成超乎想象的规模化任务，但岗位性质将发生演变。

威廉姆斯同时提醒，在技术过渡中需守护基础知识体系。“理解基本原理至关重要，”他表示，“我们必须通过中学、大学及在职教育，确保使用者既能掌握基础，又能对AI产出结果负责。二者缺一不可。”

综合来看，企业若希望在AI浪潮中稳健前行，需将治理框架、韧性建设与人才培育置于与技术部署同等的战略高度，方能在提升效率的同时管控风险，实现可持续增长。

中文翻译：

由谷歌云赞助
如何选择首批生成式AI应用场景
要开启生成式AI之旅，首先应聚焦于能够优化人类信息交互体验的领域。面对AI应用日益普及的趋势，治理与韧性建设至关重要。

随着各行业AI部署持续加速，运营韧性已成为区分技术规模化成功者与落后风险者的关键指标。这一结论源自旧金山IT事件管理供应商PagerDuty发布的报告，该机构通过对1000名IT业务主管的调研，探讨了AI应用如何影响企业营收。

调研结果显示，企业能否获益与其部署的AI工具类型关联有限，而更取决于对配套基础设施的重视程度。报告揭示了企业对AI韧性的投入与获得可衡量回报之间存在显著关联。数据显示，近四分之三受访企业在过去一年提升了运营韧性，并有同等比例的企业正在增加韧性建设预算。

在国家层面，韧性与营收的关联愈发清晰。报告指出，在英国和爱尔兰，超过四分之三的营收增长型企业正在增加韧性预算，而在营收停滞或下滑的企业中，这一比例不足半数。简言之，投资韧性的机构正获得更高投资回报，而忽视者正迅速落后。

治理与韧性建设的缺失会引发多重影响。除财务损失（报告显示超半数英爱企业在重大IT事件中每小时损失超30万美元）外，企业声誉还可能遭受持久损害。

技术创新的规模也催生了新挑战：如何建立相适配的治理体系。在近期圆桌会议上，PagerDuty首席信息官埃里克·约翰逊、产品开发高级副总裁大卫·威廉姆斯及工程副总裁若昂·弗雷塔斯指出，尽管企业面临持续集成AI与创新的压力，但现有治理体系尚无法匹配技术变革的速度。

"当前存在盲目追逐AI快速部署的现象，却缺乏适当的防护措施与协调工具，"威廉姆斯表示，"这正是挑战所在。"他强调AI智能体应像人类员工一样，被赋予明确的权限、监督机制与问责体系。"构建AI团队需遵循人类团队的管理逻辑：确保'聘用'合适角色、协调分工、监控绩效、淘汰低效单元并逐步赋予更多责任。若不以同等标准对待技术管理，后果可能是灾难性的。"

他补充道："我们可以为智能体编码这些规则，搭建基础设施、协调智能体、设置防护栏、提供可观测性——这些都能避免可靠性问题。但现状是这些尚未落实。"

弗雷塔斯指出AI加速开发进程带来的压力："过去需六个月的项目，现在十天就能完成。"效率提升显著，但风险也随之加剧。"应用AI时，我们必须思考如何保持可靠性与合规性，尤其是在企业级场景中。"

除基础设施外，AI的快速扩张引发了更人性化的议题：劳动力将何去何从。早期关于AI将淘汰初级岗位的论调正在修正，约翰逊强调"人在回路"依然不可或缺。"若支持站点的智能体向客户提供错误信息，企业仍需为此承担责任。关键在于增强员工能力，而非总是取代他们。"

针对入门级岗位，约翰逊认为初期恐慌实属过虑。他提出变革的将是岗位性质而非岗位存续。伴随AI成长的新生代员工将带来前所未有的技术专长。"当这些'AI原住民'进入职场时，他们将擅长驾驭AI完成超乎想象规模的任务。"

威廉姆斯持相似观点但发出警示：要融合两代人的优势，必须确保基础知识在转型中得以传承。"理解基本原理至关重要。问题在于如何通过中学、大学及在职教育体系落实这种培养？我们需要人们既掌握基础原理，又能在使用工具时对结果负责。二者兼备则前景可期，缺失任一都将陷入困境。"

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.
Governance and resilience are important in the face of increased AI use.
With AI deployments continuing to rise across industries, operational resilience has emerged as a key differentiator between those scaling the tech successfully and those who risk falling behind.
That’s according to a report from San-Francisco based IT incident management vendor PagerDuty, which surveyed 1,000 IT business leaders to answer the question of how AI uptake is affecting revenue.
The answer, it turns out, has less to do with which AI tools a company deploys and more with how seriously it takes the infrastructure around it.
The report revealed a key correlation between companies’ investment in AI resilience and achieving measurable returns.
According to the findings, most of the companies covered by the report, nearly three-quarters, increased operational resilience over the past year, with a similar number actively growing their resilience budgets.
Yet, at the national level, the link between resilience and revenue is increasingly stark. In the U.K. and Ireland, more than three-quarters of revenue-growing companies are increasing resilience budgets, compared with fewer than half of those with flat or declining revenues, according to the report.
To put it plainly, organizations that invest in resilience are pulling ahead in ROI, while those that don’t are quickly falling behind.
Failure to implement good governance and resilience has a multi-layered impact. In addition to financial loss (with the report finding that more than half of U.K. and Irish companies face losses of $300,000 or more per hour during major IT incidents, reputational damage can follow a company beyond the incident itself.
The scale of technological innovation has also given rise to another concern: that of implementing adequate governance structures.
Speaking at a recent roundtable, PagerDuty CIO Eric Johnson, SVP of product development David Williams and engineering VP Joao Freitas argued that while companies are feeling the pressure to continually integrate AI and innovate, governance structures to manage these deployments are not yet capable of managing the pace of change.
"There is this insane race to deploy AI as quickly as possible, without the appropriate guardrails or orchestration tools being in place," Williams says. "That's the challenge."
AI agents, he argued, should be managed the same way as human staff, with appropriate permissions, oversight and accountability.
"You should be treating them like you would if you were building a team of people," he said. "Make sure you hire people appropriate for the job, coordinate the work, divide it among them, monitor their performance, weed out the ones that are not performing well and give them more responsibility over time. If we're not thinking about technology the same way, it can be catastrophic."
“You can codify all these things for agents,” he added. “You can build infrastructure, orchestrate agents, provide guardrails, provide observability -- all these things that prevent reliability from becoming a problem. But that hasn't been done yet.”
Freitas pointed to the speed of AI-enabled development as another pressure point. "We had projects that previously would take six months that we can now do in 10 days," he said.
While the efficiency gains are significant, so are the risks.
"When we apply AI, we need to think about how we continue to be reliable, how we continue to be compliant, especially at the enterprise level,” Freitas added.
Beyond infrastructure, AI's rapid expansion is raising a more human question about what happens to the workforce.
The early narrative that AI would render junior roles obsolete is already being revised, and Johnson stressed that the human-in-the-loop remains essential.
"If you put an agent on your support site and it gives a customer the wrong information, then you are liable for that response," he said. "It's about augmenting staff, not always replacing them."
On the question of entry-level jobs specifically, Johnson said the initial panic was misplaced. What will change, he argued, is the nature of those roles.
With younger generations growing up alongside AI, he says these employees will be able to provide a level of tech expertise that was previously neither possible nor necessary.
"By the time they enter the workforce, we're going to have these AI natives that are really good at understanding how to manipulate AI to get things done at a scale we can't even imagine right now,” Johnson added.
Williams struck a similar note, but with a warning. Capturing the best of both generations requires ensuring that foundational knowledge isn't lost in the transition.
"Understanding the fundamentals is important," he said. "The question is how do you make sure the education is there: across secondary school, universities, as well as on the job? We need people to understand the fundamentals and be held accountable for checking the results when they use these tools. If you have those two things in place, I think we're in good shape. If you skip either one, you're in trouble."