4 What Are Al Agents & How They Work 4 什么是人工智能代理及其工作原理
8 The State of Al Agents in 2025 8 2025 年人工智能代理的现状
12. Al Agent Use Cases That Drive Value 12 推动价值的人工智能代理应用案例
19 How To Get Started with Al Agents 19 如何开始使用人工智能代理
26 Common Pitfalls & How to Avoid Them 26 个常见陷阱及避免方法
35 The Future of Work with AI Agents 35 AI 代理的未来工作
41 Conclusion 41 结论
Al agents. A concept that once seemed like science fiction but is now transforming how businesses operate in 2025. AI 代理。一个曾经看似科幻的概念，如今正在改变 2025 年企业的运营方式。
For some, they represent an opportunity to automate mundane tasks. For others, they raise anxious questions: “Will an AI agent replace my entire job? Can I simply delegate everything while these digital assistants impress my boss? Could AI agents run every aspect of my professional and personal life?” 对一些人来说，它们代表着自动化枯燥任务的机会。对另一些人来说，它们引发了焦虑的问题：“AI 代理会取代我整个工作吗？我能否完全委托一切，而这些数字助理却让我的老板印象深刻？AI 代理能否管理我职业和个人生活的方方面面？”
The reality? No, Al agents cannot do 100% of your work-and that’s actually good news. The true power of AI agents lies not in replacement but in partnership. As Kieran Flanagan, HubSpot’s SVP of Marketing, puts it: 现实是？不，AI 代理不能完成你 100%的工作——这其实是个好消息。AI 代理的真正力量不在于取代，而在于合作。正如 HubSpot 市场高级副总裁 Kieran Flanagan 所说：
“The companies that are being most successful [with AI agents]… they’re not replacing entire roles. They’re promoting people. You’re still accountable for the end results, you’re still accountable for reviewing this, putting a nice bow on it, presenting this. But now you have this extra tool and these agents can do something for you.” “那些在使用 AI 代理方面最成功的公司……他们并没有替代整个职位。他们是在提升员工。你仍然要对最终结果负责，仍然要负责审查、完善和展示。但现在你有了这个额外的工具，这些代理可以为你做些事情。”
In this guide, we feature exclusive insights from HubSpot’s Chief Marketing Officer, Kipp Bodnar, and Senior Vice President of Marketing, Kieran Flanagan, as they cut through the hype to reveal what’s actually possible with AI agents in 2025. Drawing from real implementation experience across our organization, they’ll show you where to start, which use cases deliver measurable value today, and how to build an effective human-AI collaboration strategy. 在本指南中，我们特别邀请了 HubSpot 的首席营销官 Kipp Bodnar 和高级营销副总裁 Kieran Flanagan，剖析炒作背后的真相，揭示 2025 年 AI 代理真正能实现的可能性。基于我们组织内的实际实施经验，他们将向您展示从哪里开始，哪些用例今天能带来可衡量的价值，以及如何构建有效的人机协作策略。
What Are Al Agents & How They Work 什么是 AI 代理及其工作原理
In the simplest terms, Al agents are software systems that can perform tasks autonomously on your behalf. But that definition hardly captures what makes today’s AI agents so transformative for businesses in 2025. 最简单来说，AI 代理是能够代表您自主执行任务的软件系统。但这个定义远不能体现出 2025 年 AI 代理对企业带来的变革性影响。
66
When you think about it, an agent, most of them will have memory because you need that to have agent behavior. And most of them will have access to tools because they can actually do things on your behalf. They’re autonomous. 仔细想想，一个代理，大多数都会有记忆，因为你需要记忆来实现代理行为。大多数代理还会访问工具，因为它们实际上可以代表您执行操作。它们是自主的。
explains Kieran Flanagan. 基兰·弗拉纳根解释道。
99
Unlike traditional chatbots or AI assistants that simply respond to immediate queries, modern AI agents can handle multi-step processes without continuous human guidance. This distinction is crucial to understanding their potential. 与传统的聊天机器人或仅对即时查询作出回应的人工智能助手不同，现代人工智能代理能够在无需持续人工指导的情况下处理多步骤流程。这一区别对于理解它们的潜力至关重要。
The Anatomy of an Effective Al Agent 高效人工智能代理的构成
True AI agents in 2025 combine several key capabilities that enable them to function as digital teammates rather than just tools. 2025 年的真正人工智能代理结合了多项关键能力，使其能够作为数字团队成员而不仅仅是工具发挥作用。
Memory and Context Awareness 记忆与上下文感知
The most effective Al agents maintain context throughout a task, remembering previous interactions and instructions. This persistent memory allows them to build on past work and understand nuance in ways that simple queryresponse systems cannot. 最有效的人工智能代理能够在整个任务过程中保持上下文，记住之前的交互和指令。这种持久的记忆使它们能够在过去的工作基础上继续发展，并以简单的查询响应系统无法做到的方式理解细微差别。
An agent helping with customer support doesn’t just answer single questions-it remembers the entire conversation history, recognizes when issues are related to previous tickets, and can maintain context even across multiple sessions. 协助客户支持的代理不仅仅回答单个问题——它会记住整个对话历史，识别问题是否与之前的工单相关，并且能够在多次会话中保持上下文。
Tool Integration 工具集成
What truly separates agents from earlier AI systems is their ability to use tools and access various data sources. Modern agents can interact with your CRM, analytics platforms, email systems, and other software to gather information and take action. 真正将智能代理与早期人工智能系统区分开来的是它们使用工具和访问各种数据源的能力。现代智能代理可以与您的客户关系管理系统（CRM）、分析平台、电子邮件系统及其他软件进行交互，以收集信息并采取行动。
“These agents are going to use those tools to interact with other systems, your ERP, your CRM, whatever that might be,” notes Flanagan. This integration capability means agents aren’t limited to what they “know”-they can actively retrieve information and manipulate systems to accomplish goals. “这些智能代理将利用这些工具与其他系统交互，无论是您的企业资源计划系统（ERP）、客户关系管理系统（CRM）还是其他系统，”Flanagan 指出。这种集成能力意味着智能代理不局限于它们“知道”的内容——它们可以主动检索信息并操作系统以实现目标。
Multi-Step Reasoning and Planning 多步骤推理与规划
Perhaps the most impressive aspect of today’s Al agents is their ability to break down complex tasks into logical steps and work through them methodically. They don’t simply execute preprogrammed flows-they can reason about how to approach novel situations. 当今人工智能智能代理最令人印象深刻的方面，或许是它们能够将复杂任务分解为逻辑步骤，并有条不紊地逐步完成。它们不仅仅执行预设流程——还能推理如何应对新颖的情况。
66
Kipp Bodnar points out that this isn’t science fiction anymore: “I think it’s a peek at what’s to come. I think it shows that the future is still a little ways away,” acknowledging both the remarkable progress and the ongoing development in this space. Kipp Bodnar 指出，这已经不再是科幻小说：“我认为这是一瞥未来的景象。我认为这表明未来仍有一段距离，”他既认可了这一领域的显著进展，也承认了其持续的发展。
Kipp Bodnar
Beyond Chatbots: The Evolution to True Agents 超越聊天机器人：向真正智能代理的演进
The confusion between chatbots and AI agents is understandable but important to clarify. While both use similar underlying language models, their capabilities differ substantially: 聊天机器人和人工智能代理之间的混淆是可以理解的，但需要澄清。虽然两者都使用类似的底层语言模型，但它们的能力有很大不同：
A chatbot takes your question and delivers an answer. An agent takes your goal and delivers a result. 聊天机器人接收你的问题并给出答案。
This distinction becomes clear when looking at how they handle complexity. Ask a chatbot to “find the best-performing blog posts from the last quarter and draft social media updates for each,” and you’ll likely get general instructions or an example. An AI agent will actually retrieve the analytics data, identify the top performers, draft customized social posts for each, and even schedule them if integrated with your publishing tools. 当涉及复杂任务时，这一区别就很明显了。向聊天机器人提出“找出上季度表现最好的博客文章，并为每篇文章起草社交媒体更新”这样的请求时，你很可能只会得到一般性的指导或示例。而 AI 代理则会实际检索分析数据，识别表现最佳的文章，为每篇文章起草定制的社交媒体内容，甚至如果与你的发布工具集成，还能安排发布时间。
Chatbot vs. Al Agent: Same Request, Different Outcomes 聊天机器人 vs. AI 代理：相同请求，不同结果
CHATBOT 聊天机器人
Provides Information Only 仅供参考
“Find my best-performing blog posts from last quarter and draft social media updates for each” “查找我上季度表现最好的博客文章，并为每篇文章起草社交媒体更新内容”
Here’s how you can do that: 操作步骤如下：
Go to your analytics dashboard 进入你的分析仪表盘
Filter for last quarter’s posts 筛选上季度的帖子
Sort by page views/engagement 按页面浏览量/互动量排序
For each post, draft social updates 为每个帖子起草社交更新内容
Include post links and hashtags Here’s an example template… 包含帖子链接和标签 这是一个示例模板…
COMPLETES ENTIRE WORKFLOW 完成整个工作流程
66
The way that we think about it is everyone knows about all these LLMs, so ChatGPT and Claude and everything. And they’re very good at basically predicting the next token and creating content," Flanagan explains. "But the beauty of agents is you can exploit that feature to help these LLMs kind of navigate a problem on their own. So it’s not a chat anymore. You give it a task and you can leave the room. And then this agent’s gonna try to autonomously figure out how to get there. Flanagan 解释说：“我们的想法是，大家都知道这些 LLMs，比如 ChatGPT 和 Claude 等。它们非常擅长预测下一个词元并生成内容。但智能代理的美妙之处在于，你可以利用这一特性帮助这些 LLMs 自主地解决问题。所以这不再是简单的聊天。你给它一个任务，然后可以离开。这个代理会尝试自主地找出完成任务的方法。”
Kieran Flanagan
This autonomy-the ability to work independently toward a goal-is what makes agents so potentially valuable for businesses looking to automate complex processes without sacrificing quality or requiring constant oversight. 这种自主性——独立朝着目标工作的能力——正是智能代理对希望自动化复杂流程且不牺牲质量或需要持续监督的企业来说极具价值的原因。
The State of Al Agents in 2025 2025 年人工智能代理的现状
In 2025, AI agents stand at a fascinating intersection: revolutionary enough to transform business processes, yet still evolving toward their full potential. For business leaders looking to separate reality from hype, understanding the current landscape is essential. 在 2025 年，人工智能代理处于一个引人入胜的交汇点：它们足够革命性，能够改变业务流程，但仍在向其全部潜力发展。对于希望区分现实与炒作的商业领袖来说，了解当前的格局至关重要。
66
“It’s a peek at what’s to come. I think it shows that the future is still a little ways away,” notes Kipp Bodnar, reflecting the measured optimism that characterizes the most successful AI implementations today. “这是一瞥未来的景象。我认为这表明未来仍有一段距离，”Kipp Bodnar 指出，这反映了当今最成功的人工智能实施所体现的审慎乐观态度。
\author{
Kipp Bodnar
}
Where Al Agents Excel Today 当今人工智能代理的优势所在
The most effective Al agents in 2025 aren’t trying to replace entire departments or roles. Instead, they’re handling specific workflows with clearly defined parameters and objectives. Particularly strong use cases include: 2025 年最有效的人工智能代理并不是试图取代整个部门或职位。相反，它们处理具有明确参数和目标的特定工作流程。特别强大的应用场景包括：
Research and data compilation processes that would take humans hours to assemble manually. Agents can pull information from multiple sources, organize findings, and present them in useful formats-all without the cognitive fatigue humans experience during repetitive tasks. 研究和数据汇编过程，这些过程如果由人类手动完成需要数小时。代理可以从多个来源提取信息，整理发现，并以有用的格式呈现——所有这些都不会像人类在重复任务中那样产生认知疲劳。
Background task management that doesn’t require real-time human oversight. Kieran Flanagan explains, “The latency doesn’t matter if it’s a task that you just want to put on in the background. And you have this little army of agents doing things.” 不需要实时人工监督的后台任务管理。Kieran Flanagan 解释道：“如果只是想把任务放在后台运行，延迟并不重要。你还有一支小型代理军团在执行任务。”
Low-precision tasks where 90%90 \% accuracy is acceptable. These make ideal starting points for organizations new to agent technology, as they offer meaningful value with minimal risk. 低精度任务，其中 90%90 \% 的准确率是可以接受的。这些任务是刚接触代理技术的组织的理想起点，因为它们以最低的风险提供了有意义的价值。
The adoption of AI agents follows a familiar technology curve, with larger enterprises leading the charge. AI agents are transforming how businesses interact with their customers. 54% of global companies are using conversational AI in some way or the other to provide faster and more personalized service. However, more sophisticated agent implementations-those that handle complex multi-step processes-remain at an earlier adoption stage. AI 代理的采用遵循了一个熟悉的技术曲线，大型企业率先推动。AI 代理正在改变企业与客户互动的方式。全球 54%的公司以某种方式使用对话式 AI，以提供更快、更个性化的服务。然而，更复杂的代理实现——那些处理复杂多步骤流程的——仍处于较早的采用阶段。
This adoption landscape highlights a crucial opportunity: those implementing more advanced agents today are positioning themselves at the leading edge. 这一采用格局凸显了一个关键机遇：那些今天实施更先进代理的人正将自己置于领先地位。
66
As Kieran notes, “If you’re listening to the podcast and you’re following along and you’re thinking, ‘I’m gonna create an agent and do something,’ you’re in the fast movers. You still have an opportunity to be an incredibly fast mover here.” 正如基兰所说，“如果你正在听播客并跟随进展，心想‘我要创建一个代理并做点什么’，那么你就是快速行动者。你仍然有机会成为这里极其快速的行动者。”
Kieran Flanagan 基兰·弗拉纳根
The Human-Al Partnership Model 人机合作模式
The most successful implementations of AI agents in 2025 follow a clear pattern: they enhance human capabilities rather than attempting to replace them. This creates a multiplier effect where employees can focus their time on higher-value work. 2025 年最成功的 AI 代理实施遵循一个明确的模式：它们增强人类能力，而不是试图取代人类。这创造了一种乘数效应，使员工能够将时间集中在更高价值的工作上。
Effective Al agent strategies recognize that humans and Al bring complementary strengths to the table: 有效的 AI 代理策略认识到人类和 AI 各自带来了互补的优势：
Al agents excel at execution, processing vast amounts of data, maintaining consistency, working without breaks, and handling repetitive tasks without losing focus. AI 代理擅长执行，处理大量数据，保持一致性，无需休息地工作，并且能够专注地处理重复性任务。
Humans excel at judgment, creativity, relationship building, complex decision making, and providing the critical oversight that ensures AI outputs align with business objectives. 人类擅长判断、创造力、建立关系、复杂决策，以及提供关键的监督，确保 AI 输出符合业务目标。
“The companies that are being most successful [with AI agents]… they’re not replacing entire roles,” explains Flanagan. “You’re still accountable for the end results, you’re still accountable for reviewing this, putting a nice bow on it, presenting this. But now you have this extra tool and these agents can do something for you.” “那些在使用 AI 代理方面最成功的公司……他们并没有替代整个职位，”Flanagan 解释道。“你仍然需要对最终结果负责，仍然需要审查这些内容，给它们做个漂亮的包装，进行展示。但现在你有了这个额外的工具，这些代理可以为你做一些事情。”
"Is This an Agent Job?" Decision Tree “这是不是代理的工作？” 决策树
A framework for determining which task to automate with Al agents 一个用于确定哪些任务适合用 AI 代理自动化的框架
Current Limitations and Challenges 当前的限制和挑战
Understanding what Al agents can’t yet do is just as important as knowing their strengths. Current limitations include: 了解人工智能代理尚不能完成的任务，与了解其优势同样重要。目前的局限包括：
Speed and latency issues. Today’s agents often work slower than humans at individual tasksthough they make up for this by working continuously and on multiple tasks simultaneously. 速度和延迟问题。现有的代理在单个任务上的处理速度通常比人类慢，尽管它们通过持续工作和同时处理多个任务来弥补这一点。
Trust and control mechanisms. Organizations are still navigating the right balance of autonomy versus oversight. “One of the big question marks around autonomous agents is what’s the right UX pattern so people feel comfortable with them,” Flanagan points out. “Should it just come back and say it’s booked or should you actually be able to see the agent complete the task so you feel comfortable with the agent doing something on your behalf?” 信任与控制机制。组织仍在探索自主性与监督之间的最佳平衡。Flanagan 指出：“关于自主代理的一个大疑问是，什么样的用户体验模式能让人们感到安心。它是直接反馈任务已完成，还是应该让用户看到代理完成任务的全过程，从而让用户对代理代表自己执行操作感到放心？”
Integration complexity. Connecting agents to legacy systems and proprietary databases remains a challenge, often requiring technical expertise. The most successful implementations start with systems that have modern, accessible APIs. 集成复杂性。将代理连接到遗留系统和专有数据库仍然是一个挑战，通常需要技术专长。最成功的实施通常始于拥有现代且易于访问 API 的系统。
High-precision requirements. Tasks demanding near-perfect accuracy still require significant human oversight. For now, these remain better suited to human-AI collaboration rather than full agent autonomy. 高精度要求。需要近乎完美准确性的任务仍然需要大量的人类监督。目前，这些任务仍然更适合人机协作，而非完全自主的智能代理。
The 2025 Inflection Point 2025 年拐点
What makes 2025 particularly significant in the AI agent timeline is the convergence of several key factors: improved foundation models, expanding tool integration capabilities, and growing organizational readiness. 2025 年在 AI 代理时间线上之所以特别重要，是因为多个关键因素的汇聚：基础模型的改进、工具集成能力的扩展以及组织准备度的提升。
Unlike previous waves of AI hype, today’s agent technologies are delivering measurable business impact -even in their still-evolving state. Organizations implementing agents now are developing the expertise, infrastructure, and cultural readiness that will position them for competitive advantage as these technologies continue to mature. 与以往的 AI 热潮不同，现今的代理技术即使仍在不断发展中，也已带来可衡量的商业影响。现在实施代理的组织正在培养专业知识、建设基础设施并提升文化准备度，这将使它们在这些技术持续成熟的过程中获得竞争优势。
As Kipp Bodnar observes, “We’re still in very short supply of just smart, rational humans applying logic and leveraging technology.” This insight captures the essence of what makes AI agents valuable in 2025 -they don’t replace the need for human intelligence, but rather amplify it through strategic automation. 正如 Kipp Bodnar 所观察到的，“我们仍然非常缺乏那些能够运用逻辑并利用技术的聪明、理性的人类。”这一见解抓住了 2025 年 AI 代理价值的本质——它们并不取代人类智能的需求，而是通过战略性自动化来增强人类智能。
Al Agent Use Cases That Drive Value 推动价值的 AI 代理用例
While the potential applications for AI agents span virtually every business function, successful implementations in 2025 share a common thread: they focus on specific, high-impact use cases rather than attempting to automate entire departments. Let’s explore the most effective applications of Al agents across marketing, sales, and operations. 虽然 AI 代理的潜在应用几乎涵盖每一个业务职能，但 2025 年成功的实施有一个共同点：它们专注于具体的、高影响力的用例，而不是试图自动化整个部门。让我们来探讨 AI 代理在市场营销、销售和运营中的最有效应用。
Marketing Applications 市场营销应用
Marketing teams have emerged as early adopters of AI agent technology, finding particular success in areas requiring both creativity and data-driven precision. 市场营销团队已成为 AI 代理技术的早期采用者，在需要创造力和数据驱动精确性的领域取得了特别的成功。
Content Production and Optimization 内容生产与优化
Al agents have revolutionized content workflows by handling the labor-intensive aspects of content creation while leaving strategic decisions to human marketers. A single content strategist can now effectively manage the output previously requiring an entire team. AI 代理通过处理内容创作中劳动密集的部分，彻底改变了内容工作流程，同时将战略决策留给人类营销人员。现在，一个内容策略师就能有效管理过去需要整个团队完成的产出。
66
I think it’s the best time in human history to monetize good ideas," Kieran Flanagan explains. "The thing AI unlocks is the ability to bring that to life in a really incredibly fast way, and you can iterate much faster with Al . “我认为这是人类历史上将好点子变现的最佳时机，”Kieran Flanagan 解释道。“AI 所开启的是以极快的速度将想法变为现实的能力，而且你可以用 AI 更快地进行迭代。”
Kieran Flanagan 基兰·弗拉纳根
Effective implementations include agents that: 有效的实现包括以下代理：
Transform long-form content into multiple channel-specific formats 将长篇内容转换为多种渠道特定格式
Generate SEO-optimized blog posts from rough outlines 根据粗略大纲生成 SEO 优化的博客文章
Create personalized email sequences based on customer segment data 根据客户细分数据创建个性化电子邮件序列
Analyze content performance and recommend optimization strategies 分析内容表现并推荐优化策略
HubSpot’s marketing team has seen particular success with agents that analyze YouTube content and transform it into platform-specific social media posts. As Kieran demonstrates, these agents can extract key insights from video transcripts, identify the most compelling “spicy takes” or educational moments, and format them according to best practices for platforms like LinkedIn. HubSpot 的营销团队在使用分析 YouTube 内容并将其转化为特定平台社交媒体帖子的代理方面取得了特别成功。正如 Kieran 所展示的，这些代理可以从视频转录中提取关键见解，识别最引人注目的“辣点评”或教育时刻，并根据 LinkedIn 等平台的最佳实践进行格式化。
Audience Research and Insights 受众研究与洞察
Al agents excel at continuous market monitoring and insight generation, providing marketers with deeper understanding of their audience. 人工智能代理擅长持续的市场监测和洞察生成，为营销人员提供对受众更深入的理解。
Forward-thinking companies deploy agents to: 具有前瞻性的公司部署代理来：
Monitor social conversations around specific topics and identify emerging trends 监控特定话题的社交对话并识别新兴趋势
Analyze competitor content strategies and identify potential white space 分析竞争对手的内容策略并识别潜在的空白领域
Compile customer feedback across channels into actionable insight reports 汇总各渠道的客户反馈，形成可操作的洞察报告
Test multiple content approaches simultaneously to identify winning formulas 同时测试多种内容方案，以确定最佳方案
One particularly valuable application involves agents that analyze website performance and automatically generate A/B testing hypotheses. These agents study both your site and competitor sites, then recommend specific tests likely to improve conversion rates. 一个特别有价值的应用是分析网站性能并自动生成 A/B 测试假设的智能代理。这些代理会研究您的网站及竞争对手的网站，然后推荐可能提升转化率的具体测试方案。
Campaign Analytics and Optimization 活动分析与优化
The ability to process vast amounts of performance data makes AI agents ideal for campaign optimization. Unlike traditional analytics that often provide retrospective insights, agent-powered systems can make real-time adjustments. 处理大量性能数据的能力使得 AI 代理非常适合进行活动优化。与通常提供回顾性洞察的传统分析不同，代理驱动的系统可以进行实时调整。
Innovative implementations include: 创新的实现包括：
Dynamic budget allocation across channels based on performance metrics 基于性能指标的跨渠道动态预算分配
Continuous ad copy and creative optimization based on engagement patterns 基于参与模式的持续广告文案和创意优化
Automated performance reporting with actionable recommendations 自动化绩效报告及可执行建议
Anomaly detection that flags unusual patterns requiring human attention 异常检测，标记需要人工关注的异常模式
Sales Applications 销售应用
Sales teams are finding AI agents particularly valuable for activities that historically consumed significant time without directly generating revenue. 销售团队发现，AI 代理在处理那些历史上耗时但未直接产生收入的活动时特别有价值。
Lead Qualification and Prioritization 潜在客户资格评估与优先排序
Al agents have transformed lead management by ensuring sales teams focus their energy on the most promising opportunities. 人工智能代理通过确保销售团队将精力集中在最有前景的机会上，彻底改变了潜在客户管理。
Leading organizations deploy agents that: 领先的组织部署了以下代理：
Enrich lead data by gathering information from multiple public sources 通过从多个公共来源收集信息来丰富潜在客户数据
Score and prioritize leads based on likelihood to convert 根据转化可能性对潜在客户进行评分和优先排序
Identify ideal time windows for outreach based on prospect behavior 根据潜在客户行为确定理想的联系时间窗口
Route leads to appropriate team members based on expertise alignment 根据专业匹配将潜在客户分配给合适的团队成员
Kipp describes a particularly effective implementation: Kipp 描述了一个特别有效的实施案例：
66
When someone would come into their website and create an account, they would have agents go online and start researching this person… Given the information that they found, they come up with hypotheses on how this person is going to use their product. 当有人进入他们的网站并创建账户时，他们会让代理上线开始研究这个人……根据他们找到的信息，提出关于此人将如何使用他们产品的假设。
Kipp Bodnar
Sales Outreach Customization 销售外展定制
The days of generic sales outreach are disappearing as AI agents enable hyper-personalized communications at scale. 随着 AI 代理实现大规模的超个性化沟通，通用销售外展的时代正在消失。
Successful applications include: 成功的应用包括：
Generating customized outreach messages based on prospect data 基于潜在客户数据生成定制的推广信息
Creating tailored value propositions highlighting relevant product benefits 创建突出相关产品优势的量身定制价值主张
Developing personalized case studies and social proof examples 开发个性化的案例研究和社会证明示例
Crafting follow-up sequences based on prospect engagement patterns 根据潜在客户参与模式制定后续跟进序列
“Since it’s being posted from Linkedln, the agent will know about it and then it can track it on your behalf and give you a report 24 hours later,” explains Dharmesh Shah via Kipp, highlighting how agents can provide ongoing engagement intelligence. “由于这是从 LinkedIn 发布的，代理会知道这件事，然后可以代表你跟踪，并在 24 小时后给你一份报告，”Dharmesh Shah 通过 Kipp 解释道，强调了代理如何提供持续的参与情报。
Meeting Preparation and Follow-up 会议准备与跟进
Perhaps the most time-saving application for sales teams involves meeting workflow automation. 也许对销售团队来说，最节省时间的应用是会议工作流程自动化。
High-impact implementations include agents that: 高影响力的实施包括以下代理：
Compile comprehensive prospect research before meetings 在会议前编制全面的潜在客户调研
Generate meeting agendas based on prospect needs and sales cycle stage 根据潜在客户需求和销售周期阶段生成会议议程
Create detailed meeting summaries with action items 创建包含行动事项的详细会议总结
Draft personalized follow-up communications with relevant resources 起草个性化的后续沟通，附带相关资源
Operational Applications 运营应用
Beyond customer-facing functions, AI agents are streamlining internal operations across organizations. 除了面向客户的职能外，AI 代理正在简化组织内部的运营。
Internal Knowledge Management 内部知识管理
The challenge of making organizational knowledge accessible is being addressed through AI agents that can both catalog and retrieve information intelligently. 通过能够智能编目和检索信息的 AI 代理，正在解决使组织知识可访问的挑战。
Leading organizations deploy agents to: 领先的组织部署代理以：
Create training materials from existing content 从现有内容创建培训材料
Process Automation 流程自动化
Administrative workflows that previously required significant manual effort are being transformed through AI agent automation. 以前需要大量人工操作的行政工作流程正在通过 AI 代理自动化进行转变。
Successful implementations include: 成功的应用包括：
Expense report processing and approval 费用报销处理与审批
Calendar management and meeting scheduling 日程管理与会议安排
Travel booking and itinerary optimization 差旅预订与行程优化
Procurement request handling 采购请求处理
Customer Service Augmentation 客户服务增强
While complete automation of complex customer service remains challenging, Al agents are significantly enhancing human support capabilities. 虽然复杂客户服务的完全自动化仍具挑战性，但人工智能代理显著提升了人工支持能力。
Effective applications include: 有效的应用包括：
Automated response drafting for common customer inquiries 常见客户咨询的自动回复草拟
Real-time agent assistance with product information and solutions 提供产品信息和解决方案的实时客服支持
Post-interaction summarization and categorization 互动后总结与分类
Proactive identification of at-risk customers requiring intervention 主动识别需要干预的高风险客户
Each of these use cases demonstrates the central theme of successful AI agent implementation in 2025: focused automation of well-defined processes that free humans to apply their uniquely human capabilities where they add the most value. 每一个用例都展示了 2025 年成功实施 AI 代理的核心主题：专注于自动化明确定义的流程，从而释放人类去发挥其独有的人类能力，在最能创造价值的地方发挥作用。
Transform Your Business with Breeze AI 用 Breeze AI 变革您的业务
- Breeze Copilot: - Breeze Copilot：
Your AI sidekick that understands your business, using your CRM data to help you work smarter wherever you are in HubSpot. 您的 AI 助手，了解您的业务，利用您的 CRM 数据帮助您在 HubSpot 的任何地方更智能地工作。
- Breeze Agents: - Breeze Agents：
AI-powered experts that automate complex workflows across marketing, sales, and service without expanding your headcount. 由人工智能驱动的专家，自动化处理营销、销售和服务中的复杂工作流程，无需增加员工数量。
- Complete AI Solution: - 完整的人工智能解决方案：
Experience the transformative power of HubSpot’s integrated AI platform designed for real business impact today. 体验 HubSpot 集成 AI 平台的变革力量，专为实现真正的业务影响而设计。
Learn More 了解更多
Content Agent 内容代理
Reach your audience with engaging podcasts 通过引人入胜的播客触达您的受众
Generate episode 生成剧集
Hi, how can I help? Type your request below or start with one of these suggestions: 你好，我能帮你什么？请在下面输入你的请求，或者从以下建议开始：
Prepare for a meeting 准备会议
Research a company 调研一家公司
How To Get Started with AI Agents 如何开始使用 AI 代理
Implementing AI agents doesn’t require a massive organizational overhaul or specialized technical expertise. The most successful companies take an incremental approach, starting with targeted use cases that deliver immediate value while building internal capability. 实施 AI 代理不需要大规模的组织改造或专业的技术专长。最成功的公司采取渐进式的方法，从能够立即带来价值的针对性用例开始，同时建立内部能力。
Al Agent Implementation Roadmap AI 代理实施路线图
A Step-by-Step Guide to Successful Deployment 成功部署的分步指南
The first step in any successful Al agent implementation is identifying the right opportunities to begin with. As Kieran Flanagan advises: 任何成功的 AI 代理实施的第一步都是确定合适的切入机会。正如 Kieran Flanagan 所建议的：
66
If you actually take a role, let’s say your role is a BDR, and you split out your tasks into low precision, high precision, you can actually start to pick ones that are in that low precision category. And that’s like some places to experiment with agents. 如果你真的承担一个角色，比如你的角色是 BDR，你将任务分为低精度和高精度，你实际上可以开始选择那些属于低精度类别的任务。这些就是一些可以用代理进行实验的地方。
This precision-based framework provides an excellent starting point: 这个基于精度的框架提供了一个极好的起点：
Low-Precision Tasks: 低精度任务：
Activities where 90% accuracy is acceptable and errors have minimal consequences. These make ideal first candidates for agent automation. 准确率达到 90%且错误后果较小的活动。这些是代理自动化的理想首选。
High-Precision Tasks: 高精度任务：
Activities requiring near-perfect accuracy where errors could have significant consequences. These should remain humanled for now, though agents can still assist. 需要近乎完美准确率且错误可能带来重大后果的活动。这些目前应由人工主导，尽管代理仍可提供辅助。
When evaluating potential use cases, consider these four criteria: 在评估潜在的使用场景时，请考虑以下四个标准：
(1) Frequency: Tasks performed regularly create more significant impact when automated. （1）频率：定期执行的任务在自动化时会产生更显著的影响。
2. Time Intensity: Focus on tasks that consume disproportionate human time relative to their strategic value. 2. 时间强度：关注那些相对于其战略价值消耗大量人力时间的任务。
3 Structured Data: Processes with clearly defined inputs and outputs are easier to automate successfully. 3. 结构化数据：具有明确定义输入和输出的流程更容易成功实现自动化。
4 Clear Success Metrics: Choose applications where you can measure concrete improvements. 4 个明确的成功指标：选择可以衡量具体改进的应用。
Low-Precision vs. High-Precision Task Framework 低精度与高精度任务框架
Where to Start with AI Agents in 2025 2025 年 AI 代理的起步方向
1 LOW PRECISION 1 低精度
Start here with AI agents 从这里开始使用 AI 代理
90%90 \% accuracy is acceptable 90%90 \% 的准确率是可以接受的
Errors have minimal consequences 错误的后果很小
High frequency, repetitive tasks 高频率、重复性的任务
Background processes are ideal 后台进程是理想的选择
EXAMPLES: Content drafting, research, data compilation 示例：内容起草、研究、数据汇编
² HIGH PRECISION ² 高精度
Human oversight required 需要人工监督
Near 100%100 \% accuracy required 接近 100%100 \% 的准确度要求
Once you’ve identified promising use cases, follow this practical implementation roadmap: 一旦确定了有前景的用例，请按照以下实用的实施路线图进行：
1 Start Simple and Build Success 1 从简单开始，逐步取得成功
Begin with a single, well-defined use case rather than attempting to implement multiple agents simultaneously. Early wins build organizational confidence and provide valuable learning. 从一个单一且定义明确的用例开始，而不是试图同时实施多个代理。早期的成功能够建立组织的信心并提供宝贵的经验。
2) Select the Right Technology Approach 2）选择合适的技术方案
Today’s market offers multiple paths to implementing AI agents: 如今市场上有多种实现 AI 代理的途径：
No-code platforms: Enable business users to create basic agents without technical expertise 无代码平台：使业务用户无需技术专长即可创建基础代理
Low-code frameworks: Provide greater flexibility while minimizing development requirements 低代码框架：在减少开发需求的同时提供更大灵活性
Custom development: Delivers maximum customization for enterprise-specific needs 定制开发：为企业特定需求提供最大程度的定制化
Choose an approach that matches your team’s technical capabilities and the complexity of your use case. 选择与您的团队技术能力和用例复杂性相匹配的方法。
(3) Build with Human Oversight in Mind (3) 以人为监督为核心进行构建
The most successful agent implementations maintain appropriate human supervision. Design workflows where agents handle the heavy lifting but humans retain approval authority for critical decisions. 最成功的代理实现都保持了适当的人类监督。设计工作流程时，让代理负责繁重的工作，但关键决策仍由人类审批。
Kipp notes, “The latency doesn’t matter if it’s a task that you just want to put on in the background.” This insight highlights why background processes make excellent starting points-they provide value without requiring immediate response times. Kipp 指出：“如果只是想让任务在后台运行，延迟并不重要。”这一见解强调了为什么后台进程是很好的起点——它们能提供价值，同时不需要即时响应。
4. Test Extensively Before Full Deployment 4. 在全面部署前进行充分测试
Conduct thorough testing using historical data and controlled scenarios before putting agents into production environments. This helps identify potential issues and refines agent performance. 在将代理投入生产环境之前，使用历史数据和受控场景进行彻底测试。这有助于识别潜在问题并优化代理性能。
Integration 集成
Connecting AI Agents with Existing Tools 将 AI 代理与现有工具连接
An AI agent’s effectiveness often depends on its ability to access relevant data and systems. AI 代理的有效性通常取决于其访问相关数据和系统的能力。
Successful integration requires attention to three key areas: 成功的整合需要关注三个关键领域：
1 Data Access and Security 1 数据访问与安全
Ensure agents have appropriate access to the information they need while maintaining security protocols. This often involves: 确保代理能够访问所需的信息，同时保持安全协议。这通常包括：
Creating specific API connections to internal systems 创建与内部系统的特定 API 连接
Establishing clear data usage boundaries 建立明确的数据使用界限
For maximum impact, agents should fit seamlessly into existing workflows rather than creating separate processes. Consider: 为了达到最大效果，代理应无缝融入现有工作流程，而不是创建独立的流程。请考虑：
Where agent outputs will be delivered 代理输出将被传送到何处
How employees will review and utilize agent work 员工将如何审查和利用代理的工作
Which existing systems need to connect to your agent 哪些现有系统需要连接到您的代理
3) User Experience Design 3）用户体验设计
The way users interact with agents significantly impacts adoption. Kieran emphasizes this challenge: 用户与代理的交互方式对采用率有着重大影响。Kieran 强调了这一挑战：
66
One of the big question marks around autonomous agents is what’s the right UX pattern so people feel comfortable with them… should it just come back and say it’s booked or should you actually be able to see the agent complete the task so you feel comfortable with the agent doing something on your behalf? 关于自主代理的一个大疑问是，什么样的用户体验模式才能让人们感到舒适……它是否应该仅仅回复“已预订”，还是你实际上应该能够看到代理完成任务，这样你才会对代理代表你执行操作感到放心？
99
Kieran Flanagan
The most successful implementations provide appropriate visibility into agent activities without overwhelming users with unnecessary detail. 最成功的实现方式是在不让用户被不必要的细节淹没的情况下，提供对代理活动的适当可见性。
Time saved by automating previously manual tasks 通过自动化先前的手动任务节省的时间
Volume of work processed compared to pre-agent baseline 处理的工作量与代理前基线相比
Cost per transaction or activity 每笔交易或活动的成本
2) Quality Metrics 2）质量指标
Accuracy of agent outputs compared to human benchmark 代理输出与人工基准的准确性
Error rates and types of errors occurring 错误率及错误类型
Consistency of deliverables 交付成果的一致性
3) Business Impact Metrics 3）业务影响指标
Revenue influenced by agent-supported activities 代理支持活动影响的收入
Customer satisfaction changes in agent-supported processes 代理支持流程中的客户满意度变化
Employee satisfaction and productivity improvements 员工满意度和生产力的提升
Start by establishing a pre-implementation baseline for these metrics, then track changes as agents become integrated into workflows. 首先建立这些指标的实施前基线，然后随着代理融入工作流程，跟踪变化情况。
By following this structured approach-assessing opportunities, implementing thoughtfully, integrating properly, and measuring results-organizations can successfully incorporate Al agents into their operations while minimizing risks and maximizing returns. The key is starting small, focusing on quick wins, and building on success incrementally rather than attempting wholesale transformation. 通过遵循这一结构化方法——评估机会、审慎实施、正确整合和衡量结果——组织可以成功地将 AI 代理纳入其运营，同时最大限度地降低风险并提高回报。关键是从小处着手，专注于快速成果，并在成功的基础上逐步推进，而不是试图进行全面转型。
Common Pitfalls & How to Avoid Them 常见陷阱及避免方法
Even with careful planning, organizations implementing AI agents often encounter challenges. Understanding these common pitfalls—and how to avoid them—can significantly improve your chances of success. 即使经过仔细规划，实施 AI 代理的组织仍常常遇到挑战。了解这些常见陷阱及其避免方法，可以显著提高成功的可能性。
Common AI Agent Pitfalls 常见的 AI 代理陷阱
Over-automation Without Proper Oversight 缺乏适当监督的过度自动化
Unrealistic Expectations About Capabilities 对能力的非现实期望
Poor Implementation Strategies 糟糕的实施策略
Resistance to Adoption and Change Management 对采用和变革管理的抵制
Data Quality and Integration Challenges 数据质量与集成挑战
Lack of Clear Success Metrics 缺乏明确的成功指标
Overlooking Ethical and Compliance Considerations 忽视伦理和合规性考量
Over-automation Without Proper Oversight 缺乏适当监督的过度自动化
One of the most frequent mistakes is automating too much too quickly, without maintaining appropriate human supervision. 最常见的错误之一是过快地自动化过多内容，而没有保持适当的人为监督。
The Pitfall 陷阱
Organizations eager to realize efficiency gains may automate entire workflows without establishing proper oversight mechanisms. This can lead to errors compounding unchecked, customer dissatisfaction, or regulatory compliance issues. 渴望实现效率提升的组织可能会在没有建立适当监督机制的情况下自动化整个工作流程。这可能导致错误不断累积，客户不满，或监管合规问题。
How to Avoid It 如何避免
Implement a graduated autonomy approach where agents earn increasing independence as they demonstrate reliability. As Kipp points out, “The latency doesn’t matter if it’s a task that you just want to put on in the background.” Focus initial automation on these background tasks where human review can occur before outputs reach external stakeholders. 实施分级自治方法，让代理在展示出可靠性后获得越来越多的独立性。正如 Kipp 所指出的，“如果是你只想放在后台运行的任务，延迟并不重要。”初期自动化应集中在这些后台任务上，在输出到外部利益相关者之前可以进行人工审核。
Best Practice 最佳实践
Create clear review protocols specifying which agent actions require human approval, which need periodic sampling, and which can proceed independently. 制定明确的审核协议，规定哪些代理操作需要人工批准，哪些需要定期抽查，哪些可以独立进行。
Unrealistic Expectations About Capabilities 对能力的不切实际期望
Enthusiasm for AI agent potential often leads to expectations that exceed current technological capabilities. 对 AI 代理潜力的热情常常导致对其能力的期望超出当前技术水平。
The Pitfall 陷阱
Setting unrealistic goals for what agents can accomplish leads to disappointment, wasted resources, and potentially abandonment of valuable use cases. 为代理设定不切实际的目标会导致失望、资源浪费，甚至可能放弃有价值的应用场景。
How to Avoid It 如何避免
Conduct thorough capability assessment before defining agent responsibilities. Recognize that today’s agents excel at well-defined tasks with clear parameters but struggle with highly nuanced decision-making that requires contextual judgment. 在定义代理职责之前，进行彻底的能力评估。要认识到，现今的代理擅长于具有明确参数的明确定义任务，但在需要上下文判断的高度细微决策方面表现不佳。
Best Practice 最佳实践
Start with use cases that play to current strengths-data processing, pattern recognition, and execution of clear processes-rather than scenarios requiring deep situational awareness or emotional intelligence. 从发挥当前优势的用例开始——数据处理、模式识别和执行明确流程——而非需要深度情境感知或情商的场景。
Poor Implementation Strategies 不良实施策略
Flawed implementation approaches can undermine otherwise promising AI agent initiatives. 有缺陷的实施方法可能会破坏本来有前景的 AI 代理项目。
The Pitfall 陷阱
Common implementation mistakes include inadequate testing, poor integration with existing workflows, and insufficient attention to user experience. 常见的实施错误包括测试不足、与现有工作流程整合不良以及对用户体验关注不够。
How to Avoid It 如何避免
Adopt iterative implementation methodologies with frequent testing and refinement. Pay particular attention to how agents integrate with human workflows. 采用迭代实施方法，进行频繁的测试和改进。特别关注代理如何与人类工作流程集成。
Best Practice 最佳实践
Create transparency into agent actions by providing appropriate visibility into processes while avoiding information overload. Balance automation benefits with human comfort levels. 通过提供适当的流程可见性，创造对代理行为的透明度，同时避免信息过载。在自动化收益与人类舒适度之间取得平衡。
Resistance to Adoption and Change Management 采用阻力与变革管理
Even the most technically sound agent implementation can fail if employees resist adoption. 即使是技术上最完善的代理实现，如果员工抵制采用，也可能失败。
The Pitfall 陷阱
Introducing AI agents without adequate stakeholder engagement often triggers resistance based on job security concerns, distrust of capabilities, or frustration with learning new workflows. 在没有充分利益相关者参与的情况下引入 AI 代理，往往会因员工担心工作安全、对能力缺乏信任或对学习新工作流程感到沮丧而引发抵制。
How to Avoid It 如何避免
Position agents as enhancing human capabilities rather than replacing them. Actively involve end users in the implementation process to build trust and incorporate their insights. 将代理定位为增强人类能力而非取代人类。积极让最终用户参与实施过程，以建立信任并吸收他们的见解。
Best Practice 最佳实践
Celebrate early wins where agents demonstrably make employees’ jobs easier rather than threatening them. Share success stories and create internal champions who can advocate for the benefits they’ve experienced. 庆祝代理明显使员工工作更轻松的早期成功，而非威胁员工。分享成功案例，培养内部拥护者，宣传他们所体验到的好处。
Data Quality and Integration Challenges 数据质量与集成挑战
Al agents are only as good as the data they can access. 人工智能代理的能力取决于其可访问的数据质量。
The Pitfall 陷阱
Agents built on incomplete, inaccurate, or poorly integrated data will produce disappointing results, undermining confidence in the technology. 基于不完整、不准确或整合不良数据构建的代理将产生令人失望的结果，削弱对该技术的信心。
How to Avoid It 如何避免
Conduct thorough data readiness assessments before implementation. Address data quality issues and create reliable integration pathways before deploying agents that depend on that information. 在实施之前进行彻底的数据准备评估。在部署依赖该信息的代理之前，解决数据质量问题并创建可靠的集成路径。
Best Practice 最佳实践
Start with use cases where data is already well-structured and accessible. Address data quality issues incrementally rather than attempting to solve all data problems simultaneously. 从数据已经结构良好且易于访问的用例开始。逐步解决数据质量问题，而不是试图同时解决所有数据问题。
Lack of Clear Success Metrics 缺乏明确的成功指标
Without defined metrics, it’s impossible to evaluate agent success objectively. 没有明确的指标，就无法客观评估代理的成功。
The Pitfall 陷阱
Vague goals like “improve efficiency” without specific, measurable targets make it difficult to assess ROI and refine agent performance. 像“提高效率”这样模糊的目标，没有具体、可衡量的指标，难以评估投资回报率并优化代理表现。
How to Avoid It 如何避免
Establish clear baseline metrics before implementation and define specific success criteria for each agent use case. 在实施之前建立明确的基线指标，并为每个代理用例定义具体的成功标准。
Best Practice 最佳实践
Include both efficiency metrics (time saved, volume processed) and quality metrics (accuracy, error rates) in your evaluation framework. Review and adjust these metrics as your agent program matures. 在评估框架中同时包含效率指标（节省时间、处理量）和质量指标（准确率、错误率）。随着代理项目的成熟，审查并调整这些指标。
Overlooking Ethical and Compliance Considerations 忽视伦理和合规性考虑
Al agents operate within broader ethical and regulatory contexts that cannot be ignored. 人工智能代理在更广泛的伦理和监管环境中运作，这些环境不可忽视。
The Pitfall 陷阱
Failing to consider potential bias, privacy concerns, or compliance requirements can lead to reputational damage or legal exposure. 未能考虑潜在的偏见、隐私问题或合规要求，可能导致声誉受损或法律风险。
How to Avoid It 如何避免
Incorporate ethical review into your agent development process. Ensure all implementations comply with relevant regulations and internal policies. 将伦理审查纳入您的代理开发流程。确保所有实施均符合相关法规和内部政策。
Best Practice 最佳实践
Develop an AI ethics framework specific to your organization that guides agent development and deployment decisions. Review this framework regularly to account for evolving best practices and regulatory requirements. 制定一个针对贵组织的人工智能伦理框架，指导代理的开发和部署决策。定期审查该框架，以适应不断变化的最佳实践和监管要求。
By anticipating these common pitfalls and implementing preventive strategies, organizations can significantly improve their AI agent initiatives’ likelihood of success. Remember that agent technology is still evolving-patience and a willingness to learn from early implementations will pay dividends as capabilities continue to advance. 通过预见这些常见陷阱并实施预防策略，组织可以显著提高其人工智能代理项目的成功率。请记住，代理技术仍在发展中——耐心和从早期实施中学习的意愿将在能力不断提升的过程中带来回报。
Want more cuttingedge insights on AI, marketing, and business growth? 想要获取更多关于人工智能、营销和业务增长的前沿见解？
Subscribe to the Marketing Against the Grain YouTube channel for weekly conversations with industry leaders like Dharmesh Shah and expert analysis from hosts Kipp Bodnar and Kieran Flanagan. Don’t miss out on the latest strategies and perspectives that are shaping the future of marketing and Al . 订阅“逆势营销”YouTube 频道，每周与行业领袖如 Dharmesh Shah 进行对话，并由主持人 Kipp Bodnar 和 Kieran Flanagan 提供专业分析。不要错过正在塑造营销和人工智能未来的最新策略和观点。
Subscribe 订阅
Inside GPT-4.5: The Surprising AI Upgrade Marketers Can’t Ignore. 深入了解 GPT-4.5：营销人员无法忽视的惊人人工智能升级。
How Do You Integrate AI Into the Workplace? | Kraken CMO 如何将人工智能整合到职场？| Kraken 首席营销官
This AI Analyzes Your Website Like a $10k Consultant [Tutorial] 这款人工智能像价值一万美元的顾问一样分析你的网站【教程】
The Future of Work with AI Agents 人工智能代理的未来工作
As Al agents become increasingly integrated into business operations, they’re reshaping not just how work gets done, but the very nature of professional roles and skills. Understanding these emerging dynamics will help organizations and individuals prepare for the evolving workplace of tomorrow. 随着人工智能代理越来越多地融入业务运营，它们不仅改变了工作的完成方式，也在重塑职业角色和技能的本质。理解这些新兴动态将帮助组织和个人为未来不断演变的职场做好准备。
How AI Agents Will Change Skills and Hiring 人工智能代理将如何改变技能和招聘
The rise of Al agents is creating a fundamental shift in how companies evaluate talent and how professionals position themselves in the job market. Kieran Flanagan captures this transformation with remarkable insight: 人工智能代理的兴起正在从根本上改变公司评估人才的方式以及专业人士在就业市场中的定位。基兰·弗拉纳根对此转变有着深刻的洞察：
66
In the future, people might be hired because of the agents that they have helping them do their role. 未来，人们可能会因为他们拥有帮助完成工作的代理而被雇佣。
\author{
Kieran Flanagan 基兰·弗拉纳根
}
This represents a profound evolution in professional value-from being judged solely on personal capabilities to being evaluated on one’s ability to effectively leverage AI agents as force multipliers. The implications are far-reaching: 这代表了职业价值的深刻演变——从单纯依靠个人能力被评判，转变为基于有效利用 AI 代理作为倍增器的能力进行评估。其影响深远：
New Technical Literacy Requirements 新的技术素养要求
While deep coding knowledge won’t be necessary for most professionals, understanding how to effectively prompt, direct, and collaborate with AI agents will become a core competency across functions. 虽然大多数专业人士不需要深入的编码知识，但如何有效地提示、指导和与 AI 代理协作将成为各职能部门的核心能力。
The Agent Builder Advantage 代理构建者的优势
Employees who can construct and refine their own agents to address specific workflow challenges will have significant advantages over those who can only use pre-built solutions. 能够构建和优化自己代理以解决特定工作流程挑战的员工，将比只能使用预制解决方案的员工拥有显著优势。
Emphasis on Uniquely Human Skills 强调独特的人类技能
As agents handle more routine and analytical tasks, the premium on distinctly human capabilities—creativity, empathy, ethical judgment, and strategic thinking—will increase substantially. 随着代理处理更多常规和分析性任务，对独特人类能力——创造力、同理心、伦理判断和战略思维——的需求将大幅增加。
Organizations are already beginning to adapt their hiring processes to identify candidates who demonstrate aptitude for effective human-AI collaboration. Forward-thinking companies are less concerned with candidates who can perform repetitive tasks that agents will soon handle and more interested in those who show potential to thrive in this new paradigm. 各组织已经开始调整其招聘流程，以识别那些展现出有效人机协作能力的候选人。具有前瞻性的公司不再过分关注能够完成代理即将处理的重复性任务的候选人，而更看重那些有潜力在这一新范式中茁壮成长的人。
Training Your Team to Work Alongside Al Agents 培训您的团队与 AI 代理协同工作
Successful integration of AI agents requires thoughtful preparation of your workforce. The goal isn’t teaching employees to accept replacement, but rather helping them evolve into more strategic roles as agents handle routine aspects of their work. 成功整合 AI 代理需要对您的员工进行周密准备。目标不是教员工接受被替代，而是帮助他们随着代理处理工作中的常规部分，转变为更具战略性的角色。
Kipp Bodnar highlights this ongoing need for human judgment: Kipp Bodnar 强调了对人类判断力的持续需求：
We’re still in very short supply of just smart, rational humans applying logic and leveraging technology. 我们仍然非常缺乏那些运用逻辑并利用技术的聪明、理性的人类。
99
Effective training approaches focus on: 有效的培训方法侧重于：
Agent Literacy 代理素养
Ensuring employees understand agent capabilities, limitations, and appropriate use cases. 确保员工了解代理的能力、局限性以及适用的使用场景。
Output Evaluation 输出评估
Developing critical assessment skills to effectively review and refine agentgenerated work. 培养批判性评估技能，有效审查和完善代理生成的工作。
Workflow Integration 工作流程整合
Helping employees redesign their work processes to incorporate agents at optimal points. 帮助员工重新设计工作流程，在最佳环节融入代理。
Organizations finding the most success with this transition emphasize that agents aren’t replacing employees but rather changing how they spend their time. The narrative should consistently highlight how agents free people to focus on higher-value activities where human judgment adds the greatest value. 在这一转型中取得最大成功的组织强调，智能代理并不是取代员工，而是改变他们的时间分配方式。叙述应始终突出智能代理如何解放人们，使其能够专注于那些人类判断能带来最大价值的高价值活动。
Building a Personal Portfolio of AI Agents 构建个人 AI 代理组合
Just as professionals today cultivate their network of human collaborators, the workforce of tomorrow will develop personalized collections of AI agents tailored to their specific needs and working styles. 正如当今的专业人士培养他们的人类协作者网络，未来的劳动力将发展出针对其特定需求和工作风格的个性化 AI 代理集合。
This personal agent portfolio will become an increasingly valuable professional asset. Consider how different roles might assemble complementary agents: 这一个人代理组合将成为日益宝贵的职业资产。考虑不同角色如何组建互补的代理：
Marketing Professionals 市场营销专业人员
Content generation agents, market research agents, performance analytics agents, and audience insight agents working in concert to amplify creative capabilities. 内容生成代理、市场调研代理、绩效分析代理和受众洞察代理协同工作，以增强创意能力。
aill Sales Professionals 销售专业人员
Prospect research agents, outreach customization agents, meeting preparation agents, and followup management agents combining to enhance relationship-building abilities. 潜在客户调研代理、外联定制代理、会议准备代理和跟进管理代理结合起来，提升关系建立能力。
Product Managers 产品经理
Customer feedback analysis agents, market trend monitoring agents, feature prioritization agents, and roadmap communication agents collaborating to improve decision quality. 客户反馈分析代理、市场趋势监控代理、功能优先级代理和路线图沟通代理协同工作，以提升决策质量。
As this trend accelerates, professionals will invest in building, training, and refining their personal agent teams as a form of career development. The ability to assemble an effective “AI staff” that complements your strengths and addresses your weaknesses will become a valuable skill in itself. 随着这一趋势加速发展，专业人士将投入时间构建、训练和完善他们的个人代理团队，作为职业发展的形式。组建一个能够补充自身优势并弥补弱点的高效“AI 团队”的能力，将成为一项宝贵的技能。
The Evolving Relationship Between Humans and AI 人类与人工智能关系的演变
Perhaps the most profound change ahead is the evolving relationship between knowledge workers and Al agents. We’re witnessing a transition from Al as tools to Al as collaborative partners. 也许未来最深刻的变化是知识工作者与人工智能代理之间关系的演变。我们正见证人工智能从工具向协作伙伴的转变。
Kieran captures the creative potential of this relationship: “It’s how it unlocks your creativity and the speed you can iterate with it that I just think is the most exciting thing.” 基兰捕捉到了这种关系的创造潜力：“它释放你的创造力，以及你与它迭代的速度，我认为这是最令人兴奋的事情。”
This partnership model acknowledges the complementary strengths of humans and Al : 这种合作模式承认了人类与人工智能的互补优势：
Human Strengths 人类的优势
Creative thinking 创造性思维
Contextual understanding 情境理解
Ethical judgment 伦理判断
Emotional intelligence 情商
Interpersonal trust-building 人际信任建立
Complex decision-making 复杂决策制定
Al Agent Strengths 人工智能代理优势
Information processing 信息处理
Pattern recognition 模式识别
Consistent execution 一致的执行
Tireless operation 不知疲倦的操作
Scalable application of learned knowledge 已学知识的可扩展应用
The most successful professionals in this new landscape won’t be those who resist AI integration or those who abdicate responsibility to AI. Rather, they’ll be the ones who develop sophisticated collaboration models that maximize the unique contributions of both human and artificial intelligence. 在这个新环境中，最成功的专业人士不会是那些抵制人工智能整合的人，也不会是那些将责任完全推给人工智能的人。相反，他们将是那些开发出复杂协作模型，最大化人类与人工智能各自独特贡献的人。
Preparing Your Organization for the Future 为您的组织做好未来准备
Organizations preparing for this future should focus on several key priorities: 为迎接未来，组织应关注几个关键优先事项：
Culture Shifts 文化转变
Moving from fear-based resistance to opportunity-based enthusiasm by showcasing concrete examples of how agents enhance rather than threaten professional roles. 通过展示具体案例，说明代理如何增强而非威胁专业角色，从而从基于恐惧的抵制转向基于机遇的热情。
Skill Development 技能发展
Investing in training that emphasizes both technical agent literacy and the distinctly human skills that will remain irreplaceable. 投资于培训，强调技术代理素养和将保持不可替代的独特人类技能。
Infrastructure Readiness 基础设施准备
Building the technical foundations to support increasingly sophisticated agent deployments and integrations. 构建技术基础，以支持日益复杂的代理部署和集成。
Governance Frameworks 治理框架
Establishing clear principles for how agents will be deployed, their limitations, and the human oversight mechanisms that ensure quality and ethical compliance. 建立明确的原则，规定代理的部署方式、其限制以及确保质量和伦理合规的人类监督机制。
The organizations that embrace these priorities today will develop significant competitive advantages as Al agent capabilities continue to advance. Rather than being disrupted by these changes, they’ll be positioned to harness them for both business performance and employee satisfaction. 今天拥抱这些优先事项的组织将在人工智能代理能力不断提升的过程中获得显著的竞争优势。他们不会被这些变化所颠覆，而是能够利用这些变化提升业务绩效和员工满意度。
As we look ahead, the future of work with Al agents offers tremendous potential for organizations willing to thoughtfully navigate this transition. By focusing on the unique value of human-AI collaboration rather than replacement narratives, businesses can create working environments where both their people and their technology achieve their fullest potential. 展望未来，配备 AI 代理的工作模式为愿意审慎应对这一转型的组织带来了巨大的潜力。通过关注人机协作的独特价值，而非替代论述，企业能够创造出既让员工又让技术发挥最大潜能的工作环境。
Conclusion 结论
The journey to implementing AI agents in your business isn’t about replacing people-it’s about reimagining how work gets done. As we’ve explored throughout this guide, 2025 represents a pivotal moment in the evolution of AI agent technology: capable enough to deliver real business value, yet still requiring thoughtful human partnership. 在企业中实施 AI 代理的过程并非是为了取代人力，而是重新构想工作的完成方式。正如我们在本指南中所探讨的，2025 年是 AI 代理技术发展的关键时刻：技术已足够成熟，能够带来真正的商业价值，但仍需与人类进行深思熟虑的合作。
The organizations gaining the most significant advantages today aren’t those with the most advanced technology, but those with the clearest strategy for human-AI collaboration. They understand that AI agents excel at handling repetitive, data-intensive tasks while humans bring irreplaceable creativity, judgment, and relationship-building skills to the table. 当今获得最大优势的组织，并非拥有最先进技术的企业，而是拥有最清晰人机协作战略的企业。他们明白，AI 代理擅长处理重复且数据密集的任务，而人类则带来不可替代的创造力、判断力和建立关系的能力。
66
As Kieran Flanagan puts it: “It’s how it unlocks your creativity and the speed you can iterate with it that I just think is the most exciting thing.” This encapsulates the true promise of AI agents-not as replacements, but as amplifiers of human potential. 正如基兰·弗拉纳根所说：“它激发你的创造力，以及你能够以多快的速度进行迭代，这才是我认为最令人兴奋的地方。”这恰恰体现了人工智能代理的真正承诺——不是作为替代品，而是作为人类潜能的放大器。
Key Takeaways 关键要点
1
Start small with low-precision tasks. Focus initial efforts on processes where 90%90 \% accuracy is acceptable and errors have minimal consequences. 从低精度任务的小规模开始。将初期工作重点放在 90%90 \% 准确度可接受且错误后果最小的流程上。
2
Design for human-Al partnership. The most successful implementations maintain appropriate human oversight while automating routine elements of complex workfiows. 设计以人机合作为核心。最成功的实施方案是在自动化复杂工作流程中常规环节的同时，保持适当的人类监督。
3
Measure both efficiency and quality. Track not just time saved, but also improvements in output quality, consistency, and business impact. 衡量效率和质量。跟踪的不仅是节省的时间，还包括产出质量、一致性和业务影响的提升。
Abstract 摘要
Build organizational capabilities. Develop your team’s skills in effectively directing, evaluating, and collaborating with AI agents. 构建组织能力。培养团队在有效指导、评估和协作 AI 代理方面的技能。
5
Maintain realistic expectations. Understand current capabilities and limitations to avoid disappointment and recognize the significant value Al agents can deliver today. 保持现实的期望。了解当前的能力和局限，避免失望，并认识到 AI 代理今天能够带来的重大价值。
The rapid evolution of Al agent technology means that what feels revolutionary today will become commonplace tomorrow. Organizations that begin developing expertise now will build significant competitive advantages as these technologies continue to mature. 人工智能代理技术的快速发展意味着，今天看起来革命性的事物明天将变得司空见惯。那些现在开始积累专业知识的组织，将在这些技术不断成熟的过程中建立显著的竞争优势。
As Kipp Bodnar observes, “We’re still in very short supply of just smart, rational humans applying logic and leveraging technology.” This insight captures the essence of successful AI implementation-it’s ultimately about augmenting human intelligence rather than replacing it. 正如 Kipp Bodnar 所观察到的，“我们仍然非常缺乏那些能够运用逻辑并利用技术的聪明、理性的人类。”这一见解抓住了成功实施人工智能的本质——归根结底是增强人类智能，而非取代它。
The future of work isn’t humans versus Al , but humans and Al together, each contributing their unique strengths. By starting your AI agent journey today with clear-eyed strategy and realistic goals, you’re positioning your organization to thrive in this new landscape of possibilities. 未来的工作不是人类与人工智能的对立，而是人类与人工智能的协作，各自发挥独特优势。通过今天以清晰的战略和切实的目标开启你的人工智能代理之旅，你正为你的组织在这片充满可能性的新领域中蓬勃发展奠定基础。
Now is the time to begin. The practical frameworks and implementation roadmap in this guide provide everything you need to take those crucial first steps toward a more productive, creative, and effective future with Al agents as your partners. 现在正是开始的时机。本指南中的实用框架和实施路线图为你提供了迈出关键第一步所需的一切，助力你与人工智能代理作为合作伙伴，共同迈向更高效、更具创造力和更有效的未来。
导出提示
谨慎下载超过 300 页以上 PDF，容易内存溢出。
PDF Pro 下载需借用浏览器的打印功能，出现打印界面后，在【打印机】处选择【另存为 PDF】，最后点击【保存】即可。
Ctrl
