当AI工具越来越多,Cadent与谷歌云为何主张「互联广告智能」

As more brands, agencies, and media companies join in on the AI chorus of "better, faster, smarter," the teams administering AI tools are singing a dissonant tune — managing a growing stack of AI-powered tools layered across planning, activation, measurement, reporting, and workflow management. It's a familiar irony to anyone in advertising: Another innovation designed to make the ecosystem easier is instead creating more systems, more interfaces, more costs, and more operational complexity.

随着越来越多品牌、代理商和媒体公司加入「更好、更快、更智能」的AI大合唱，负责管理AI工具的团队却唱起了不和谐的调子——他们要管理一整套层层叠加在规划、投放、测量、报告和工作流程管理上的AI工具。这对广告行业的人来说是个熟悉的讽刺：本该让生态系统变得更简单的创新，反而制造出了更多系统、更多界面、更高成本和更多运营复杂性。

Last month at AdExchanger's Programmatic AI 2026 event, Mike Rhodes, Director, Customer Engineering at Google Cloud, and Eoin Townsend, Chief Product & Technology Officer at Cadent tackled this irony — that the tools that promise efficiency and simplicity instead make day-to-day operations more difficult. The two leaders shared the stage and shared their vision for the next phase of AI that won't be defined by bigger models or more assistants, but by how effectively organizations connect intelligence across systems, teams and workflows.

上个月在AdExchanger举办的Programmatic AI 2026大会上，谷歌云客户工程总监Mike Rhodes与Cadent首席产品与技术官Eoin Townsend探讨了这一悖论——本该带来效率和简化的工具，反而让日常运营变得更加困难。两位领导者同台，分享了他们对AI下一阶段的展望——这一阶段的关键不在于更大的模型或更多的助手，而在于企业能多有效地在系统、团队和工作流程之间连接智能。

"What the industry doesn't need is another AI product," Rhodes said during the session. "The market is moving away from point solutions and toward the integration and orchestration of many different agents and workflows."

「行业并不需要另一款AI产品，」Rhodes在会上说道。「市场正在从单点解决方案转向对多种智能体和工作流程的整合与编排。」

AdExchanger: Three years into the AI boom, why are so many organizations still struggling to get meaningful business value from AI? Mike Rhodes: A big part of the issue is that many organizations have treated AI as another destination. They introduce a new tool or assistant and expect people to change how they work in order to see value. But most teams already move through too many systems every day. The AI initiatives that succeed are connected to trusted data and the places where work already happens.

AdExchanger提问：AI热潮已经三年了，为什么这么多企业仍然难以从AI中获得真正的商业价值？Mike Rhodes：很大一部分问题在于，许多企业把AI当成了又一个「目的地」。他们推出一个新工具或助手，就指望人们为了获得价值而改变工作方式。但大多数团队每天已经要在太多系统之间切换。真正成功的AI项目，是与可信数据以及工作实际发生的场所连接在一起的。

Eoin Townsend: Advertising has an additional challenge, which is that the ecosystem was already fragmented before AI entered the picture. We have a long history as an industry of taking exciting technologies and making them a lot more complicated than they need to be. Programmatic, identity, CTV, retail media — all of these created real value, but they also created more platforms, more dashboards, more workflows, and more complexity. AI could make that better, but it can also make it worse. If every platform launches its own AI layer and none of them connect, then we haven't solved fragmentation. We've just created smarter silos.

Eoin Townsend：广告行业还有一个额外的挑战，那就是在AI出现之前，这个生态系统本就已经很碎片化了。作为一个行业，我们一直有把令人兴奋的新技术变得比实际需要更复杂的「传统」。程序化、身份识别、CTV、零售媒体——这些技术都创造了真实的价值，但也带来了更多平台、更多仪表盘、更多工作流程和更多复杂性。AI可以让情况变好，但也可能让情况更糟。如果每个平台都推出自己的AI层，却互不连接，那我们并没有解决碎片化问题，只是制造出了「更聪明的孤岛」。

AdExchanger: Mike, you've said the industry doesn't need another AI product. What does it need instead? Rhodes: It needs connected intelligence. Advertising workflows don't happen in isolation. Campaign execution affects measurement, measurement affects optimization and optimization affects audience strategy. The problem is that intelligence often lives in disconnected systems. One tool understands reporting, another understands audiences and another sits inside a DSP or CRM. But I want to make one clear point: I'm not saying the market is headed toward one giant AI assistant that does everything. It's coordinated intelligence: different agents and systems working together across workflows. That's why orchestration matters as much as the model itself.

AdExchanger提问：Mike，你说行业不需要另一款AI产品。那到底需要什么？Rhodes：行业需要的是「互联智能」。广告工作流程从来不是孤立发生的：campaign执行会影响测量，测量会影响优化，优化又会影响受众策略。问题在于，智能往往分散在互不相通的系统里。一个工具懂报告，另一个懂受众，还有一个嵌在DSP或CRM里。但我想说清楚一点：我说的不是市场正在走向一个包揽一切的超级AI助手。而是「协同智能」——不同的智能体和系统在各个工作流程中协同工作。这也是为什么编排（orchestration）和模型本身同样重要。

AdExchanger: Eoin, why is advertising especially vulnerable to creating more complexity with AI? Townsend: Because we've largely just accepted complexity as normal for a very long time. Ad tech has become intimidating, and we in the industry almost have a perverse pride about that. Conversational AI changes that by allowing people to ask these business questions directly. But we need those answers to be grounded in the right data and the right advertising logic. Otherwise, it's just a polished response sitting on top of a fragmented system.

AdExchanger提问：Eoin，为什么广告行业特别容易在使用AI时制造出更多复杂性？Townsend：因为长期以来，我们已经把复杂性当成了常态。Ad Tech已经变得让人望而生畏，而我们这个行业对此几乎还带着一种反常的「自豪感」。对话式AI改变了这一点，它让人们可以直接提出业务问题。但前提是，这些答案必须建立在正确的数据和正确的广告逻辑之上。否则，它只是套在一个碎片化系统外面的「漂亮说辞」。

AdExchanger: The Cadent-Gemini integration is the news hook here. What problem were you trying to solve by bringing Cadent intelligence into Gemini Enterprise? Townsend: The goal was to meet users where they already work. Instead of asking people to learn another platform, we wanted them to access Cadent intelligence through a familiar environment. The integration is the headline, but the bigger story is the infrastructure work underneath it. We spent the past year unifying data, signals, and systems so the AI layer is built on a foundation that can support meaningful answers.

AdExchanger提问：Cadent与Gemini的整合是这次报道的由头。你们想通过把Cadent的智能能力引入Gemini Enterprise来解决什么问题？Townsend：目标是让工具「走到用户已经在的地方」。我们没有要求人们再学习一个新平台，而是希望他们能在熟悉的环境里访问Cadent的智能能力。这次整合是新闻标题，但更大的故事是背后的基础设施工作。过去一年里，我们一直在统一数据、信号和系统，好让AI层建立在一个能支撑有意义答案的基础之上。

AdExchanger: Everyone in AI talks about models. Why the emphasis on infrastructure? Rhodes: Because infrastructure determines whether AI can work reliably at enterprise scale. Models are going to keep changing. We'll keep seeing new capabilities and new debates about which model is best for which task. But enterprise success with AI won't fall along the lines of specific models. It's going to come down to smart orchestration. That means trusted data, secure access, permissions, governance, cost efficiency, and the ability to coordinate workflows across systems.

AdExchanger提问：大家谈AI言必称模型，为什么你们反而强调基础设施？Rhodes：因为基础设施决定了AI能否在企业级规模上可靠运行。模型会不断变化，我们还会看到新能力出现，关于哪个模型最适合哪类任务的争论也会持续。但企业在AI上的成功，不会取决于具体用了哪个模型，而会取决于聪明的编排能力。这意味着可信数据、安全访问、权限管理、治理机制、成本效率，以及跨系统协调工作流程的能力。

AdExchanger: Looking ahead, what does success look like for AI in advertising? Rhodes: Success means AI becomes less visible. Instead of being another tool, it becomes part of how work gets done — connecting data, workflows and decisions across the business. Townsend: For me, success looks like making advertising easier to work with. Advertising isn't going to become less complex overnight. But people shouldn't have to navigate that complexity every time they need an answer.

AdExchanger提问：展望未来，广告行业AI的成功会是什么样子？Rhodes：成功意味着AI变得「不可见」。它不再是又一个工具，而是成为完成工作方式的一部分——连接起整个业务中的数据、工作流程和决策。Townsend：对我来说，成功意味着让广告这件事变得更容易打交道。广告不会在一夜之间变得不那么复杂。但人们不应该每次想要一个答案，都要重新面对这种复杂性。

connected advertising intelligence 互联广告智能

本文核心概念，指打通规划、投放、测量、报告与工作流程管理各环节的AI智能，区别于孤立存在、各自为战的单点AI工具

programmatic AI 程序化AI/程序化广告AI

指应用于程序化广告采购与优化场景的AI技术统称，是AdExchanger近年重点覆盖的报道领域，本文所引Programmatic AI 2026大会即以此命名

orchestration 编排/协同调度

让多个智能体、系统或工作流按顺序协同运作的能力，被Rhodes视为比模型本身更能决定AI落地效果的因素

point solution 单点解决方案

只解决某一狭窄环节问题、彼此互不连接的独立工具；与「整合平台」相对，是当前AI工具泛滥导致运营复杂度上升的根源之一

smarter silos 更聪明的孤岛

形容看似升级为AI驱动、实则依旧彼此割裂的系统；提醒从业者警惕「给每个工具加一层AI」并不能真正解决碎片化问题

DSP（Demand-Side Platform） 需求方平台

广告主/代理商用于自动化采购展示、视频、CTV等广告库存的核心程序化工具；本文中作为「智能被困在孤立系统」的典型例子之一

CRM（Customer Relationship Management） 客户关系管理系统

管理企业与客户互动、销售线索及账户信息的系统平台；本文中与DSP并列，说明智能常年被困在互不相通的独立系统里

conversational AI 对话式AI

让用户通过自然语言直接提问获取答案的AI交互形式；是本文讨论「让更多人获取业务智能」的关键技术路径

grounded (in data) （回答/结论）基于真实数据、有据可依的

强调AI输出必须锚定在可信数据源上，而非仅呈现表面合理的回答；是判断AI工具是否真正可用于业务决策的核心标准

enterprise scale 企业级规模

指方案需要在大型组织的复杂系统、庞大数据量与多团队协作场景下依然稳定可靠运行；常与「demo效果好但难以规模化落地」形成对比

data governance 数据治理

涵盖可信数据、安全访问、权限管理与合规机制的整体框架；本文中被列为决定企业级AI能否落地的基础设施要素之一

agentic AI 智能体式AI

具备自主决策与执行能力、能跨系统连续完成任务链的AI系统；本文强调其真正价值在于「运营杠杆」，而非简单自动化

operational leverage 运营杠杆（效应）

指通过更少的人力投入撬动更大产出效率的效果；本文用以区分智能体式AI与传统自动化工具的核心价值主张差异

Gemini Enterprise Gemini Enterprise（谷歌企业级AI助手平台）

谷歌面向企业场景推出的AI助手产品；本文中Cadent将自身广告智能接入该平台，使销售与客户团队能够直接以对话方式获取业务洞察

fragmentation （生态）碎片化

指广告技术生态因平台、工具、数据源过多而彼此割裂的现状；是本文反复强调AI若不谨慎应用可能进一步恶化的核心风险

It's a familiar irony to anyone in advertising: Another innovation designed to make the ecosystem easier is instead creating more systems, more interfaces, more costs, and more operational complexity.

The two leaders shared the stage and shared their vision for the next phase of AI that won't be defined by bigger models or more assistants, but by how effectively organizations connect intelligence across systems, teams and workflows.

If every platform launches its own AI layer and none of them connect, then we haven't solved fragmentation. We've just created smarter silos.

But enterprise success with AI won't fall along the lines of specific models. It's going to come down to smart orchestration.