内容来源：https://aibusiness.com/agentic-ai/enterprises-overwhelmed-and-attracted-by-ai-technology

内容总结：

纽约报道——在OpenAI于周四发布其旗舰生成式AI模型最新版本GPT 5.2（距GPT 5.1发布仅一个月）之际，科技界普遍认为，AI技术的发展速度已远超企业的应用步伐。自三年前ChatGPT引爆生成式AI热潮以来，市场焦点已从基础模型转向更具自主性的智能体（Agentic AI）。面对微软、AWS、谷歌和OpenAI等厂商接连不断的新产品，企业常感应接不暇。

“炒作和噪音太多了，而像（OpenAI CEO）萨姆·阿尔特曼这样的人并没有让事情变得更简单，”智能体AI供应商EverWorker的首席执行官彼得·瓜真蒂在AI商业峰会上表示。他坦言，这种快速迭代的态势“令人感到恐惧和无所适从”。

尽管新技术看似诱人——通用汽车助理建筑师娜奥米·塔德塞在参观展会时感叹“AI无处不在”，并对其提升效率的潜力感到着迷——但她也对AI可能取代工作岗位表示担忧。瓜真蒂建议，企业不应盲目全速推进，而应“调低音量”，从小处着手，聚焦于能带来高回报、低风险的简单应用场景，例如解决现有业务痛点、人力短缺或无效开支等领域。

专家强调，数据是AI成功落地的基石。Retool销售工程师大卫·斯旺指出：“没有数据可观测性，AI应用就难以建立信任。”美国航空数据工程与分析经理阿努拉达·马拉达普进一步说明，数据必须与具体应用场景精准匹配，“上下文对AI至关重要”。

然而，许多企业在匆忙引入AI时，往往忽略了数据治理与准备工作。马拉达普观察到，当前应用进程“并非太慢，而是因缺乏底层治理基础而显得混乱”。为降低混乱，瓜真蒂建议企业应向内部全员开放AI工具权限。

NBC环球目的地与体验首席数字技术官克里斯·克雷纳在会议中分享了实践经验：企业应致力于让AI工具“去神秘化”，明确将其定位为服务于特定业务目标的工具，从而缓解各岗位员工在实际工作中遇到的阻力。

（本文由谷歌云赞助支持）

中文翻译：

由谷歌云赞助
如何选择首个生成式AI应用场景
要开启生成式AI之旅，首先应关注能够优化人类信息交互体验的领域。企业可通过微调逐步适应AI工具的落地。

纽约讯——OpenAI于周四发布其旗舰生成式AI模型最新版本GPT-5.2，距GPT-5.1面世仅一月之隔。这似乎印证了科技界许多人的观点：AI技术的发展速度已超越其应用普及的节奏。

自三年前OpenAI推出ChatGPT引爆生成式AI热潮以来，市场焦点已从基础AI模型转向智能体AI。当微软、AWS、谷歌和OpenAI等厂商接连推出新产品时，企业往往刚接触前一项技术就不得不迎接下一轮变革。

层出不穷的AI模型与智能体工具有时令人应接不暇。
"当前存在大量炒作与噪音，而像（OpenAI首席执行官萨姆·奥特曼）这样的人并未让局面变得更清晰，"智能体AI供应商EverWorker的首席执行官彼得·瓜根蒂在Informa人工智能峰会期间接受AI Business采访时表示。该公司专注于帮助企业通过无代码方式创建、定制和部署智能体AI员工。

"他们总在重复陈词滥调，"瓜根蒂继续说道，"这确实令人感到恐慌和无所适从。"

由于新AI产品发布伴随的喧嚣与营销攻势，企业可能产生落后焦虑，而旁观者则容易对技术产生高深莫测的向往。

通用汽车助理建筑师娜奥米·塔德塞对此深有感触。
"AI已无处不在，"塔德塞在采访中指着人工智能峰会展厅内数十家展示工具的AI厂商说道，"制造业和汽车领域已深度融合AI技术……这些工具能帮助人们大幅提升效率，非常吸引人。"

"虽然我不完全认同'AI取代工作岗位'这类说法，"她补充道，"但显然所有人都在拥抱AI，奋力向前迈进。"

然而EverWorker的瓜根蒂指出，企业在推进AI时应保持审慎，而非全速冒进。
"我的建议是降低外界干扰音量，"他表示，"从小处着手，保持专注。当前业务中哪些环节存在痛点？哪些岗位面临招聘困难？哪些开支未能创造实际价值？"

"应从这些简单直接、高回报低风险的场景开始实践。"

除简化起步方案外，企业更需聚焦数据基础。
"缺乏数据可观测性，AI应用就难以建立信任，"低代码软件与AI开发平台Retool的销售工程师大卫·斯旺在12月10日的专题讨论中强调。

数据可观测性之所以关键，在于它能确保企业将AI技术应用于恰当的领域。
"上下文对AI至关重要，"美国航空数据工程与分析经理阿努拉达·马拉达普解释道，"将数据与正确场景匹配是成功的关键。"

她进一步指出，企业不仅要关注数据，更需确保数据处于"就绪状态"。验证数据是否适配应用场景是判断其成熟度的重要方式。事实上，许多企业在匆忙引入AI技术时，往往忽视了对数据适用性、预处理及治理机制的考量。

当多数人认为AI应用进展缓慢时，马拉达普却持相反看法：
"我们推进得过快，未能充分考虑底层治理流程与基础建设，"她坦言，"这导致局面更趋混乱而非缓慢。"

瓜根蒂提出，让组织内部全员接触AI工具是缓解混乱的途径之一。
"目前存在大量针对特定部门或场景的定制化智能体。"

NBC环球正致力于降低员工使用AI工具的门槛。该公司目的地与体验业务首席数字技术官克里斯·克雷纳在12月10日的炉边谈话中，探讨了如何运用生成式与智能体AI解决组织内部痛点。

"无论是一线运营主管还是后台AI数据科学家，每个人的工作中都存在摩擦点，"克雷纳表示，"我们正努力消除技术神秘感，并明确传达：这是我们将要使用的工具，而这是我们运用它的视角。"

您可能还喜欢

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.
Enterprises can implement small changes to ease their adoption of AI tools.
NEW YORK -- OpenAI's release on Thursday of GPT 5.2, the latest version of its flagship generative AI model, just a month after GPT 5.1, appears to back up the contention of many in the tech world that AI technology is advancing at a faster pace than its adoption.
Over the last three years, since the popularization of generative AI following OpenAI's release of ChatGPT, the market has shifted from hype about AI models to hype about agentic AI. At times it seems that enterprises are barely taking in one technology before vendors such as Microsoft, AWS, Google, and OpenAI release their next product.
Sometimes, the multitude of AI models and agentic AI tools can feel overwhelming.
"There's a lot of hype and noise, and I don't think people like [OpenAI CEO Sam Altman] are making it easier," said Peter Guagenti, CEO of agentic AI vendor EverWorker, during an interview with AI Business at Informa's AI Summit conference.
EverWorker's focus is on helping businesses create, customize and deploy agentic AI workers without using code.
"They are talking in platitudes and all these other things," Guagenti continued. "It's really scary and overwhelming."
Due to the noise and promotional energy surrounding new AI releases, enterprises may feel like they are behind, while to others, the technology may seem advanced and tantalizing.
It certainly felt compelling for Naomi Taddesse, an assistant architect at General Motors.
"There's a lot of AI everywhere," Taddesse said in an interview, referring to the dozens of AI vendors showcasing their tools on the AI Summit floor. "There is a lot of AI integrated within manufacturing, within automative … things that will help or aid you do things a lot faster. It's very intriguing."
"The whole taking away jobs, all that stuff … I don't necessarily love that," she continued. "But I can see that everyone is going toward AI, and everyone is trying to move ahead."
However, in moving toward AI, enterprises need to be intentional, instead of moving full speed ahead, Guagenti, of EverWorker, said.
"What I advise is to turn the volume down," he said. "Start small, stay focused. Where in your business do you have pain today? Where in your business can't you hire today? Where in your business do you have excess expenses that you know are not adding value?"
"Start with these simple, straightforward use cases where you have a high reward and a low risk," he continued.
In addition to starting with simple applications, enterprises should focus on their data.
"Without data observability, I do not believe there is ever trust between AI applications," said David Swan, a sales engineer at Retool, developer of a low-code software and AI development platform, during a panel at the conference on Dec. 10.
Data observability is also critical because it helps ensure that the use cases enterprises are applying AI technology to are grounded correctly.
"Context really matters for AI," said Anuradha Maradapu, manager of data engineering and analytics, data governance at American Airlines, in an interview. "Pointing the data to the right use case really matters."
She added that not only should enterprises focus on data, but they should also ensure that their data is ready. One way to determine if the data is ready is to verify that it fits the application. Indeed, a certain lack of consideration for ensuring that data is the right data for the job and has been properly prepared and governed is reflected in the sometimes hurried pace with which many businesses are adopting AI technology.
While many say that adoption is slow, for Maradapu, it's somewhat the opposite.
"We are moving so fast without considering all the underlying governing processes and foundations put in place," she said. "So that it's actually becoming chaotic more than slow."
One way to make things less disorderly is for enterprises to give everyone within their organization access to AI tools, Guagenti said.
"There are all these bespoke agents that are departmental or use case focused," he said.
One organization that says it's working on making the tools less confusing for workers is NBC Universal.
In a fireside chat at the conference on Dec. 10, Chris Crayner, EVP, chief digital and technology officer at NBC Universal Destinations and Experience, discussed how enterprises can use generative and agentic AI to address challenging points within the organization.
"It doesn't matter if you're a front-line attraction lead for us, or you're an AI data scientist at the back," Crayner said. "It doesn't really matter. There's friction in everybody's job."
"We're trying to demystify it and also be transparent that this is a tool that we're going to use, but here's the lens that we're trying to use it through," he continued.
You May Also Like