
你很可能熟悉这类说法:企业采用人工智能,绩效提升最终趋于停滞,领导者则开始怀疑自己的投资回报率。对于如何弥合采用与成果之间的差距,专家们意见不一。有人呼吁改善治理,有人主张优化部署,还有我个人最青睐的方案——更好的赋能。这些补救措施中的每一项,都取决于组织能否就“成功是什么样”达成一致,并且能够衡量它。
You’re likely familiar with some version of this narrative: companies adopt AI, performance gains eventually plateau, and leaders are left wondering about the ROI from their investments. Experts disagree on how to remedy this gap between adoption and results. Some call for better governance, others for better deployment. Or my personal favorite, better enablement. Each one of these remedies depends upon organizations being aligned on what success looks like and being able to measure it.
以学习与发展(L&D)专业人士为例。他们可能在支持你的人工智能采用计划,同时自己也面临同样的组织问题。没错,那些最有准备去建设技能的人,也在为衡量人工智能的投资回报率而苦苦挣扎。
Take learning and development (L&D) professionals. They may be supporting your AI adoption plans while facing the same organizational issues. That’s right, the people most prepared to build skills are also struggling to measure the ROI of AI.
我们为年度《人工智能在学习与发展中的应用报告》调查了来自25个以上国家的800多名职场学习专业人士。以下是我们能从他们身上学到的、关于如何从采用走向成果的经验。
We surveyed more than 800 workplace learning professionals across more than 25 countries for our annual AI in Learning & Development Report. Here’s what we can learn from them about getting from adoption to results.
没有达成一致的结果,团队就会默认追求速度。L&D团队如何使用人工智能,反映了其组织所奖励和衡量的东西。如果员工所在的组织追踪人工智能使用情况,并奖励使用更多token,那么团队的AI使用就会反映这一点。只有当这种奖励没有与明确的衡量结果相配套时,它才会成为问题。组织需要能够说:我们希望你们使用更多人工智能,以完成X。否则,他们得到的就是生产率提升,而绩效改善却微乎其微。
Without agreed-upon outcomes, teams default to speed How L&D teams use AI reflects what their organization rewards and measures. If employees are working at organizations that track AI usage and reward more tokens being used, then teams’ AI use will reflect that. This type of reward only becomes an issue when it isn’t paired with clear measurement outcomes. Organizations need to be able to say, we want you to use more AI, in order to accomplish X. Otherwise, they’re left with productivity gains that amount to little performance improvement.
在我们的报告中,75%的受访者将更快的内容创作列为他们从人工智能采用中看到的最明显影响之一。这体现为缩短制作新培训内容所需的时间,以及提高培训产出。与此同时,只有11%的人将改善在职绩效列为一项影响。
In our report, 75% of respondents named faster content creation as one of the clearest impacts they’ve seen from AI adoption. This looks like reducing the time it takes to produce new training and increasing the output of training. Meanwhile, only 11% named improved on-the-job performance as an impact.
大多数职场培训旨在提升在职表现。然而,只有35%的学习与发展团队在持续衡量这一点(相比之下,有72%的团队在持续衡量参与度和完成率)。
Most workplace training is designed to improve on-the-job performance. Yet only 35% of L&D teams are consistently measuring that (compared to 72% who consistently measure participation and completions).
这表明,领导者要么正根据这些指标来奖励和衡量学习与发展的表现,要么就是没有人就成功应该是什么样子达成共识。我们的调查反馈指向后者。超过五分之一的人将缺乏共识性的成功指标列为证明影响力的最大障碍。
This suggests that leaders are either rewarding and measuring L&D performance against those metrics, or that no one has agreed on what success should look like. Our survey responses point to the latter. More than one in five name a lack of agreed-upon success metrics as their biggest barrier to proving impact.
团队被要求使用AI,却处处碰壁 奖励并不是学习与发展团队在采用AI时面临的唯一组织约束。组织允许什么、如何治理AI,同样重要。
Teams are told to use AI, then hit roadblocks Rewards aren’t the only organizational constraint L&D teams face when they adopt AI. What an organization permits, how it governs AI, matters just as much.
这首先从团队可以访问哪些AI解决方案开始。在因领导层指令而开始使用AI的学习与发展团队中,有73%仍然需要部门或跨职能许可才能试用新的AI工具。一些团队被限制在公司批准的AI工具范围内,无法在此之外进行测试。
That starts with which AI solutions teams can access. 73% of L&D teams that started from a leadership mandate to use AI still need departmental or cross-functional permission to pilot a new AI tool. Some teams are limited to company-approved AI tools and can’t test beyond that.
团队还需要合适的数据来评估AI的影响。大多数学习与发展团队相信AI正在带来改进,但难以提供这些收益的证据。当被问及证明影响力的最大障碍时,53%的人指出了一项组织约束,包括缺乏共识性的成功指标以及获取数据或测量工具的渠道。
Teams also need the right data to evaluate AI’s impact. A majority of L&D teams are confident that AI is leading to improvements, but struggle to provide evidence of those gains. Asked for the biggest barrier to proving impact, 53% pointed to an organizational constraint, including the lack of agreed-upon success metrics and access to data or measurement tools.
这可能表现为衡量影响力所需的数据分散在互不连通的系统中,例如学习管理系统、人力资源信息系统、调查工具和AI视频平台。
That might look like data needed to measure impact living in disconnected systems, such as a Learning Management System, an HR Information System, a survey tool, and an AI video platform.
如果学习与发展团队使用AI来更快地产出培训内容,却不评估这些培训是否带来了行为改变,那么他们只是把AI当作内容工厂。在领导者就速度之外的其他成果达成共识之前,这些团队能拿出的AI采用成果,就只有更多的内容。
If L&D teams are using AI to churn out training faster, but aren’t evaluating whether that training is leading to behavioral change, then they’re just using AI as a content factory. Until leaders agree on outcomes beyond speed, more content is all those teams will have to show for AI adoption.
构建AI实现投资回报所需的结构 好消息是,约三分之一的L&D团队报告称,AI带来了绩效提升,例如缩短达到熟练水平的时间或改善在岗表现。有几个特征让这些团队脱颖而出,比如能够自由试点或构建AI解决方案,或者使用AI来评估自己的工作。这些因素往往取决于L&D无法控制的组织因素。
Build the structure AI needs to deliver ROI The good news is that about a third of L&D teams report performance gains from AI, like faster time to proficiency or improved on-the-job performance. Several traits set these teams apart, like being able to freely pilot or build AI solutions, or using AI to evaluate their work. Those factors often depend on organizational factors beyond L&D’s control.
根据我们的数据,领导者可以对AI战略做出三项结构性调整,以实现投资回报。
Based on our data, there are three structural changes leaders can make to their AI strategy to deliver ROI.
第一,先就AI应该实现什么达成一致,再就其使用如何获得奖励和衡量达成一致。如果没有明确的结果,团队就会产出最容易展示的东西,而在L&D中,这就是更快的内容。对L&D团队来说,这意味着在设计任何新培训之前,先与领导者合作,确定他们希望看到哪些绩效变化,以及如何衡量这些变化。是的,这也包括AI赋能。
First, align on what AI is supposed to accomplish, then on how its usage is rewarded and measured. Without a defined outcome, teams produce what’s easiest to show, which in L&D is faster content. For L&D teams, that means working with leaders to identify the performance changes they want to see, and how to measure them, before designing any new training. And yes, that includes AI enablement.
第二,在扩大新的AI解决方案规模或自动化更多工作流程之前,先投资于数据集成。确保跨职能团队能够获取衡量AI影响所需的数据。
Second, invest in data integration before scaling new AI solutions or automating more workflows. Make sure teams across functions have access to the data they need to measure AI’s impact.
第三,让你的AI战略与L&D合作。如果AI的目标是自动化人工工作,那么目标就应该是腾出时间用于更具战略性的能力建设,比如如何利用AI改变工作的完成方式。L&D非常适合主导这种能力建设,因为他们能够识别培训在哪些方面可以成为解决方案的一部分,在哪些方面不能。只有这样,培训才能帮助弥合AI采用与投资回报之间的差距。
Third, partner with L&D on your AI strategy. If the goal of AI is to automate manual work, then the goal should be to free up time for more strategic capability building, like how to use AI to change how work gets done. L&D is well suited to lead that capability building because they can identify where training can be part of the solution and where it can’t. Only then can training help close the gap between AI adoption and ROI.
来自选定的TNW理事会成员的专业见解,这些成员通过申请和审核流程被纳入一个收费项目。作者表达的观点仅代表其本人。
Expertise from selected TNW Council members, admitted through an application-and-review process to a fee-based program. Opinions expressed by the authors are their own.