• Personalization
    什么是Agentic AI?AI Agent如何重塑HR行业? "Agentic AI"(代理人工智能)是您可能听说过的最新流行语,但实际上这个词在人力资源工作中的应用其实已经有一段时间了! 但究竟什么是 Agentic AI?它和传统的AI Agent有何区别?2025年是否真的是AI Agent之年? 特别呈现这篇文章与您分享! 在人工智能(AI)快速发展的今天,我们已经经历了**预测AI(Predictive AI)和生成式AI(Generative AI)**的兴起,而如今,**Agentic AI(自主智能体AI)**正成为AI的下一个进化阶段。对于HR行业而言,这一技术的到来意味着更加智能的HR系统、自动化的人才管理流程,以及更精准的数据驱动决策。 但究竟什么是Agentic AI?它和传统的AI Agent有何区别?2025年是否真的是AI Agent之年?本文将为你详细解析Agentic AI的核心概念,并探讨它如何改变HR行业。 1. 什么是Agentic AI? Agentic AI(自主智能体AI)是一种具备自主行动能力的人工智能技术,它不仅能像生成式AI一样回答问题、生成内容,还能自主感知环境、推理分析、执行任务,并从反馈中不断优化自身能力。 相比于传统的AI系统,Agentic AI最大的不同在于: 自主性(Autonomy):无需人工干预,AI代理可以独立完成任务,例如审核候选人简历、优化招聘流程等。 适应性(Adaptability):AI能够根据反馈不断优化决策,例如HR系统可以自动调整绩效评估标准,以适应不同部门的需求。 目标导向(Goal Orientation):Agentic AI可以自主制定目标,并推理如何达成这些目标,例如自动匹配候选人与职位,提高招聘效率。 2. AI Agent vs. Agentic AI:有什么区别? 在HR行业中,我们常见的AI Agent(AI代理),例如智能客服或自动化面试助手,已经在许多企业得到应用。但与Agentic AI相比,传统AI Agent仍然具有局限性。 举个例子: AI Agent:只能回答员工关于公司福利的常见问题,比如“今年的年假政策是什么?” Agentic AI:不仅能回答问题,还能主动分析员工的休假情况,自动推荐合适的休假时间,并结合公司政策优化排班,确保业务顺利运行。 3. Agentic AI如何改变HR行业? 随着Agentic AI的发展,HR的许多日常工作将发生巨变。以下是几个关键应用场景: (1)智能招聘与人才管理 Agentic AI可以帮助HR从简历筛选、面试安排到人才匹配实现全流程自动化。 ? 自动筛选简历:AI代理可通过自然语言处理(NLP)分析海量简历,并根据职位要求筛选最匹配的候选人。? 优化招聘流程:Agentic AI能够自主调整招聘策略,例如根据市场趋势调整岗位描述,优化招聘渠道,提高人才获取效率。? 智能面试安排:AI代理可以结合面试官和候选人的日程,自动安排面试,并实时调整时间,减少HR的重复沟通工作。 (2)绩效评估与员工发展 HR部门可以利用Agentic AI来优化绩效考核体系,并制定个性化的员工成长路径。 ? 智能绩效评估:AI代理可实时分析员工的工作数据,提供个性化绩效反馈,帮助管理者更公平地评估员工表现。? 个性化职业发展:Agentic AI可以分析员工的职业路径,自动推荐合适的培训课程或晋升机会,帮助企业留住优秀人才。 (3)员工体验与组织管理 AI可以提高员工满意度,并优化组织架构,提高整体效率。 ? 智能员工助手:AI代理可以主动提醒员工提交报销单、更新考勤信息,甚至预测员工的离职风险,并提前采取措施留住人才。? 企业文化管理:AI可以分析员工情绪,帮助HR团队制定更合适的企业文化建设方案。 4. 为什么2025年是“AI Agent之年”? 2025年,Agentic AI的应用将迎来爆发式增长,这主要得益于以下三大趋势: (1)AI技术的成熟与算力提升 随着大模型(如ChatGPT、NVIDIA NeMo)的不断升级,AI的推理能力越来越强,使得Agentic AI在HR场景下更加实用。 (2)企业数字化转型加速 全球范围内,企业正在加快HR数字化转型。Agentic AI能够帮助HR团队自动化重复性工作,让HR更专注于战略性任务,因此将被广泛应用。 (3)人才市场变化与HR挑战 后疫情时代,企业面临招聘难、员工流动性增加等挑战。Agentic AI可以通过智能化的人才管理系统,提高招聘效率、优化员工体验,并降低HR工作负担。 ? 预测:到2025年,超过50%的企业将引入Agentic AI,以优化HR管理流程。 5. HR如何准备迎接Agentic AI时代? 2025年将是AI Agent之年,HR行业必须抓住这一变革机遇。以下是HR团队可以采取的三大行动: ✅ 学习Agentic AI相关知识,关注AI在HR领域的应用趋势,如AI招聘、智能绩效管理等。✅ 尝试小规模部署AI代理,比如在员工服务、招聘管理等领域测试AI解决方案。✅ 与AI厂商合作,寻找适合企业的AI解决方案,如NVIDIA、微软、谷歌等提供的Agentic AI技术支持。 HR的未来,不只是管理人,更是管理智能体!Agentic AI将成为HR行业的重要助手,助力企业迈向智能化管理新时代! ? RAIHR倡导:实施负责任的AI(Responsible AI in HR, RAIHR) 随着Agentic AI在HR行业的广泛应用,我们必须关注AI的伦理、安全和公平性问题。**RAIHR(Responsible AI in HR)**倡导企业在引入Agentic AI时,遵循以下三大原则,确保AI技术的透明性、公平性和责任性: ✅ 透明性(Transparency):确保AI决策过程可解释,HR能够理解AI的筛选标准、考核指标,避免“黑箱”决策。✅ 公平性(Fairness):AI招聘和绩效评估应避免算法偏见,确保候选人和员工得到公平、公正的对待。✅ 责任性(Accountability):AI在HR领域的应用应遵循合规要求,确保数据安全,并提供人工复核机制,避免AI错误影响员工职业发展。 Agentic AI的未来,不仅是效率与智能的提升,更应是“负责任的AI”!HR行业需要共同努力,确保AI技术真正惠及企业与员工,让AI成为推动组织可持续发展的正向力量! ?? 总结:Agentic AI将彻底改变HR工作方式 ? AI Agent vs. Agentic AI:传统AI Agent只是执行预设任务,而Agentic AI能自主学习、推理和优化。? HR应用场景:Agentic AI将在招聘、绩效评估、员工体验等方面发挥巨大作用。? 2025年是AI Agent之年:技术突破、企业数字化转型、HR挑战推动Agentic AI的全面应用。 ? 实施负责任的AI(Responsible AI in HR, RAIHR)我们必须关注AI的伦理、安全和公平性问题 未来,HR不再是“人力资源管理者”,而是“AI智能管理者”!准备好迎接这场AI革命了吗? ?  
    Personalization
    2025年03月16日
  • Personalization
    Autonomous Corporate Learning Platforms: Arriving Now, Powered by AI Josh Bersin 的文章通过人工智能驱动的自主平台介绍了企业学习的变革浪潮,标志着从传统学习系统到动态、个性化学习体验的重大转变。他重点介绍了 Sana、Docebo、Uplimit 和 Arist 等供应商的出现,它们利用人工智能动态生成和个性化内容,满足了企业培训不断变化的需求。Bersin 讨论了跟上多样化学习需求所面临的挑战,以及人工智能解决方案如何提供可扩展的高效方法来管理知识和提高学习效果,并预测了人工智能将从根本上改变教学设计和内容交付的未来。推荐给大家:   Thanks to Generative AI, we’re about to see the biggest revolution in corporate learning since the invention of the internet. And this new world, which will bring together personalization, knowledge management, and a delightful user experience, is long overdue. I’ve been working in the corporate learning market since 1998, when the term “e-learning” was invented. And every innovation since that time has been an attempt to make training easier to build, easier to consume, and more personalized. Many of the innovations were well intentioned, but often they didn’t work as planned. First came role based learning, then competency-driven training and career-driven programs. These worked great, but they couldn’t adapt fast enough. So people resorted to short video, YouTube-style platforms, and then user-authored content. We then added mobile tools, highly collaborative systems, MOOCs, and more recently Learning Experience Platforms. Now everyone is focused on skills-based training, and we’re trying to take all our content and organize it around a skills taxonomy. Well I’m here to tell you all this is about to change. While none of these important innovations will go away, a new breed of AI-powered dynamic content systems is going to change everything. And as a long student of this space, I’d like to explain why. And in this conversation I will discuss four new vendors, each of which prove my point (Sana, Docebo, Uplimit, and Arist). The Dynamic Content Problem: Instructional Design By Machine Let’s start with the problem. Companies have thousands of topics, professional skills, technical skills, and business strategies to teach. Employees need to learn about tools, business strategies, how to do their job, and how to manage others. And every company’s corpus of knowledge is different. Rolls Royce, a company now starting to use Galileo, has 120 years of engineering, technology, and manufacturing expertise embedded in its products, documentation, support systems, and people. How can the company possibly impart this expertise into new engineers? It’s a daunting problem. Every company has this issue. When I worked at Exxon we had hundreds of manuals explaining how to design pumps, pressure vessels, and various refinery systems. Shell built a massive simulation to teach production engineers how to understand geology and drilling. Starbucks has to teach each barista how to make thousands of drinks. And even Uber drivers have to learn how to use their app, take care of customers, and stay safe. (They use Arist for this.) All these challenges are fun to think about. Instructional designers and training managers create fascinating training programs that range from in-class sessions to long courses, simulations, job aids, and podcasts. But as hard as they try and as creative as they are, the “content problem” keeps growing. Right now, for example, everyone is freaked out about AI skills, human-centered leadership, sustainability strategies, and cloud-based offerings. I’ve never seen a sales organization that does quite enough training, and you can multiply that by 100 when you think about customer service, repair operations, manufacturing, and internal operations. While I always loved working with instructional designers earlier in my career, their work takes time and effort. Every special course, video, assessment, and learning path takes time and money to build. And once it’s built we want it to be “adaptive” to the learner. Many tools have tried to build adaptive learning (from Axonify to Cisco’s “reusable learning objects“) but the scale and utility of these innovations is limited. What if we use AI and machine learning to simply build content on the fly? And let employees simply ask questions to find and create the learning experience they want? Well thanks to innovations from the vendors I mentioned above, this kind of personalized experience is available today.  (Listen to my conversation with Joel Hellermark from Sana to hear more.) What Is An Autonomous Learning Platform? The best analogy I’ve come up with is the “five levels of autonomous driving.” We’re going from “no automation” to “driver assist” to “conditional automation” to “fully automated.” Let me suggest this is precisely what’s happening in corporate training. If you look at the pace of AI announcements coming (custom GPTs, image and video generation, integrated search), you can see that this reality has now arrived. How Does This Really Work Now that I’ve had more than a year to tinker with AI and talk with dozens of vendors, the path is becoming clear. The new generation of learning platforms (and yes, this will eventually replace your LMS), can do many things we need: First, they can dynamically index and injest content into an LLM, creating an “expert” or “tutor” to answer questions. Galileo, for example, now speaks in my own personal voice and can answer almost any question in HR I typically get in person. And it gives references, examples, and suggests follow-up questions. Companies can take courses, documents, and work rules and simply add them to the corpus. Second, these systems can dynamically create courses, videos, quizzes, and simulations. Arist’s tool builds world-class instructional pathways from documents (try our free online course on Predictions 2024 for example) and probably eliminates 80% of the design time. Docebo Shape can take sales presentations and build an instructional simulation automatically, enabling sales people to practice and rehearse. Third, they can give employees interactive tutors and coaches to learn. Uplimit’s new system, which is designed for technical training, automatically gives you an LLM-powered coach to step you through exercises, and it learns who you are and what kind of questions you need help with. No need to “find the instructor” when you get stuck. Fourth, they can personalize content precisely for you. Sana’s platform, which Joel describes here, can not only dynamically generate content but by understanding your behavior, can actually give you a personalized version of any course you choose to take. These systems are truly spectacular. The first time you see one it’s kind of shocking, but once you understand how they work you see a whole new world ahead. Where Is This Going While the market is young, I see four huge opportunities ahead. First, companies can now take millions of hours of legacy content and “republish it” in a better form. All those old SCORM or video-based courses, exercises, and simulations can turn into intelligent tutors and knowledge management systems for employees. This won’t be a simple task but I guarantee it’s going to happen. Why would I want to ramble around in the LMS (or even LinkedIn Learning) to find the video, or information I need? I”d just like to ask a system like Galileo to answer a question, and let the platform answer the question and take me to the page or word in the video to watch. Second, we can liberate instructional design. While there will always be a need for great designers, we can now democratize this process, enabling sales operations people, and other “non-designers” to build content and courses faster. Projects like video authoring and video journalism (which we do a lot in our academy) can be greatly accelerated. And soon we’ll have “generated VR” as well. Third, we can finally integrate live learning with self-directed study. Every live event can be recorded and indexed in the LLM. A two hour webinar now becomes a discoverable learning object, and every minute of explanation can be found and used for learning. Our corpus, for example, includes hundreds of hours of in-depth interviews and case studies with HR leaders. All this information can be brought to life with a simple question. Fourth, we can really simplify compliance training, operations training, product usage, and customer support. How many training programs are designed to teach someone “what not to do” or “how to avoid breaking something” or “how to assemble or operate” some machine? I’d suggest its millions of hours – and all this can now be embedded in AI, offered via chat (or voice), and turned loose on employees to help them quickly learn how to do their jobs. Vendors Watch Out This shift is about as disruptive as Tesla has been to the big three automakers. Old LMS and LXP systems are going to look clunkier than ever. Mobile learning won’t be a specialized space like it has been. And most of the ERP-delivered training systems are going to have to change. Sana and Uplimit, for example, are both AI-architected systems. These platforms are not “LMSs with Gen AI added,” they are AI at the core. They’re likely to disrupt many traditional systems including Workday Learning, SuccessFactors, Cornerstone, and others. Consider the content providers. Large players like LinkedIn Learning, Skillsoft, Coursera, and Udemy have the opportunity to rethink their entire strategy, and either put Gen AI on top of their solution or possibly start with a fresh approach. Smaller providers like us (and thousands of others) can take their corpus of knowledge and quickly make it come to life. (There will be a massive market of AI tools to help with this.) I’m not saying this is easy. If you talk with vendors like Sana, Docebo, Arist, and Uplimit, you see that their AI platforms have to be highly tuned and optimized for the right user experience. This is not as simple as “dumping content into ChatGPT,” believe me. But the writing is on the wall, Autonomous Learning is coming fast. As someone who has lived in the L&D market for 25 years, I see this era as the most exciting, high-value time in two decades. I suggest you jump in and learn, we’ll be here to help you along the way. About These Vendors Sana (Sana Labs) is a Sweden-based AI company that focuses on transforming how organizations learn and access knowledge. The company provides an AI-based platform to help people manage information at work and use that data as a resource for e-learning within the organization. Sana Labs’ platform combines knowledge management, enterprise search, and e-learning to work together, allowing for the automatic organization of data across different apps used within an organization. Docebo is a software as a service company that specializes in learning management systems (LMS). It was founded in 2005 and is known for its Docebo Learn LMS and other tools, including Docebo Shape, its AI development system. The company has integrated learning-specific artificial intelligence algorithms into its platform, powered by a combination of machine learning, deep learning, and natural language processing. The company went public in 2019 and is listed on the Toronto Stock Exchange and the Nasdaq Global Select Market. Uplimit is an online learning platform that offers live group courses taught by top experts in the fields of AI, data, engineering, product, and business. The platform is known for its AI-powered teaching assistant and personalized learning approach, which includes real-time feedback, tailored learning plans, and support for learners. Uplimit’s courses cover technical and leadership topics and are designed to help individuals and organizations acquire the skills needed for the future. Arist is a company that provides a text message learning platform, allowing Fortune 500 companies, governments, and nonprofits to rapidly teach and train employees entirely via text message. The platform is designed to deliver research-backed learning and nudges directly in messaging tools, making learning accessible and effective. Arist’s approach is inspired by Stanford research and aims to create hyper-engaging courses in minutes and enroll learners in seconds via SMS and WhatsApp, without the need for a laptop, LMS, or internet. The company has been recognized for its innovative and science-backed approach to microlearning and training delivery. BY JOSHBERSIN 
    Personalization
    2024年02月18日
  • Personalization
    人工智能正在以比我预期更快的速度改变企业学习AI Is Transforming Corporate Learning Even Faster Than I Expected 在《AI正在比我预想的更快地改变企业学习AI Is Transforming Corporate Learning Even Faster Than I Expected》这一文中,Josh Bersin强调了AI对企业学习和发展(L&D)领域的革命性影响。L&D市场价值高达3400亿美元,涵盖了从员工入职到操作程序等一系列活动。传统模型正在随着像Galileo™这样的生成性AI技术的发展而演变,这改变了内容的创建、个性化和传递方式。本文探讨了AI在L&D中的主要用例,包括内容生成、个性化学习体验、技能发展,以及用AI驱动的知识工具替代传统培训。举例包括Arist的AI内容创作、Uplimit的个性化AI辅导,以及沃尔玛实施AI进行即时培训。这种转型是深刻的,呈现了一个AI不仅增强而且重新定义L&D策略的未来。 在受人工智能影响的所有领域中,最大的变革也许发生在企业学习中。经过一年的实验,现在很明显人工智能将彻底改变这个领域。 让我们讨论一下 L&D 到底是什么。企业培训无处不在,这就是为什么它是一个价值 3400 亿美元的市场。工作中发生的一切(从入职到填写费用账户再到复杂的操作程序)在某种程度上都需要培训。即使在经济衰退期间,企业在 L&D 上的支出仍稳定在人均 1200-1500 美元。 然而,正如研发专业人士所知,这个问题非常复杂。有数百种培训平台、工具、内容库和方法。我估计 L&D 技术空间的规模超过 140 亿美元,这甚至不包括搜索引擎、知识管理工具以及 Zoom、Teams 和 Webex 等平台等系统。多年来,我们经历了许多演变:电子学习、混合学习、微型学习,以及现在的工作流程中的学习。 生成式人工智能即将永远改变这一切。 考虑一下我们面临的问题。企业培训并不是真正的教学,而是创造一个学习的环境。传统的教学设计以教师为主导,以过程为中心,但在工作中常常表现不佳。人们通过多种方式学习,通常没有老师,他们寻找参考资料,复制别人正在做的事情,并依靠经理、同事和专家的帮助。因此,必须扩展传统的教学设计模型,以帮助人们学习他们需要的东西。 输入生成人工智能,这是一种旨在合成信息的技术。像Galileo™这样的生成式人工智能工具 可以以传统教学设计师无法做到的方式理解、整合、重组和传递来自大型语料库的信息。这种人工智能驱动的学习方法不仅效率更高,而且效果更好,能够在工作流程中进行学习。 早期,在工作流程中学习意味着搜索信息并希望找到相关的东西。这个过程非常耗时,而且常常没有结果。生成式人工智能通过其神经网络的魔力,现在已经准备好解决这些问题,就像 L&D 的瑞士军刀一样。 这是一个简单的例子。我问Galileo™(该公司经过 25 年的研究和案例研究提供支持):“我该如何应对总是迟到的员工?请给我一个叙述来帮助我?” 它没有带我去参加管理课程或给我看一堆视频,而是简单地回答了问题。这种类型的互动是企业学习的大部分内容。 让我总结一下人工智能在学习与发展中的四个主要用例: 生成内容:人工智能可以大大减少内容创建所涉及的时间和复杂性。例如,移动学习工具Arist拥有AI生成功能Sidekick,可以将综合的操作信息转化为一系列的教学活动。这个过程可能需要几周甚至几个月的时间,现在可以在几天甚至几小时内完成。 我们在Josh Bersin 学院使用 Arist ,我们的新移动课程现在几乎每月都会推出。Sana、Docebo Shape和以用户为中心的学习平台 360 Learning 等其他工具也同样令人兴奋。 个性化学习者体验:人工智能可以帮助根据个人需求定制学习路径,改进根据工作角色分配学习路径的传统模型。人工智能可以理解内容的细节,并使用该信息来个性化学习体验。这种方法比杂乱的学习体验平台(LXP)有效得多,因为LXP通常无法真正理解内容的细节。 Uplimit是一家致力于构建人工智能平台来帮助教授人工智能的初创公司,它正在使用其Cobot和其他工具为学习人工智能的技术专业人员提供个性化的指导和技巧。Cornerstone 的新 AI 结构按技能推荐课程,Sana 平台将 Galileo 等工具与学习连接起来,SuccessFactors 中的新 AI 功能还为用户提供了基于角色和活动的精选学习视图。 识别和发展技能:人工智能可以帮助识别内容中的技能并推断个人的技能。这有助于提供正确的培训并确定其有效性。虽然许多公司正在研究高级技能分类策略,但真正的价值在于可以通过人工智能识别和开发的细粒度、特定领域的技能。 人才情报领域的先驱者Eightfold、Gloat和SeekOut可以推断员工技能并立即推荐学习解决方案。实际上,我们正在使用这项技术来推出我们的人力资源职业导航器,该导航器将于明年初推出。 用知识工具取代培训:人工智能在学习与发展中最具颠覆性的用例也许是完全取代某些类型培训的潜力。人工智能可以创建提供信息和解决问题的智能代理或聊天机器人,从而可能消除对某些类型培训的需求。这种方法不仅效率更高,而且效果更好,因为它可以在个人需要时为他们提供所需的信息。 沃尔玛今天正在实施这一举措,我们的新平台 Galileo 正在帮助万事达卡和劳斯莱斯等公司在无需培训的情况下按需查找人力资源信息和政策信息。LinkedIn Learning 正在向 Gen AI 搜索开放其软技能内容,很快 Microsoft Copilot 将通过 Viva Learning 找到培训。 这里潜力巨大 在我作为分析师的这些年里,我从未见过一种技术具有如此大的潜力。人工智能将彻底改变 L&D 格局,重塑我们的工作方式,以便 L&D 专业人员可以花时间为企业提供咨询。 L&D 专业人员应该做什么?花一些时间来了解这项技术,或者参加Josh Bersin 学院的一些新的人工智能课程以了解更多信息。 随着我们继续推出像伽利略这样的工具,我知道你们每个人都会对未来的机会感到惊讶。L&D 的未来已经到来,而这一切都由人工智能驱动。
    Personalization
    2023年12月13日