干什么不会被机器人抢饭碗？

在新闻和某些学术文献中，自动化往往在被描绘成人机之间的绝世之战，而在这场战争中，机器似乎注定将取得胜利，唯一的问题只在于它们什么时候举行颁奖仪式。随着职场中的自动化技术迅速从科幻小说走入商业现实，我们相信，它带来的远非人或机器人的简单二选一，而是更加微妙的抉择。针对美国经济领域各个行业的超过2,000种工作，我们进行了研究。以下是研究得出的八项发现，其中凸显了不同行业的自动化潜力——我们提供了一些技术指标，让大家看看哪些岗位最可能接受自动化，哪些岗位最不可能被自动化。

一些工作可以被自动化，不代表它就会被自动化

尽管技术可行性是自动化的必要前提，但是打造令人惊叹的商业应用还需要考虑其他因素。这其中包括开发和部署自动化软硬件所需的成本，以及劳动力供需的变动。用昂贵的机器人取代每小时挣10美元的人类厨师，在技术上或许可以实现，但在商业上可能没有太大意义，因为成本可能太高，投资回报率很差。管理和社会问题，可能也是许多医院的病人在手术后醒来想要人类护士而不是机器人照料的原因。

某些体力工作被自动化的概率最大

美国职场人士几乎有五分之一的时间都要干体力活，或是在可预见的环境中操作器械——也就是说，这些环境的设定十分常见，改变也相对容易预测。这类工作中，有超过四分之三都能用我们当下的技术自动化，尤其是制造业和食品服务业。从技术角度来看，这些行业最有可能接受自动化。在厂房中，机器人已经开始进行产品装配和打包的重复工作，而在食品服务业，一些餐厅正在测试自助下单，甚至采用机器人侍者。

数据收集和处理领域的自动化时机也已经成熟。在美国经济领域的所有岗位上，员工要花费三分之一的时间来收集和处理数据。这两项工作都很适合自动化，这将给零售、金融服务和保险等行业带来影响。我们没必要让员工离职，但他们扮演的角色可能会发生很大变化。例如，抵押经纪人会花费多达90%的时间来处理申请，自动化以后，他们可以用更多时间来给客户提供建议。

甚至连高薪岗位也会受到影响。并不只有入门员工或低收入员工才需要收集和处理数据，那些年收入超过20万美元的人也需要花费30%的时间来做这项工作。因此，各公司会很有兴趣将这些工作自动化。总体来看，工资和工作能否被自动化的关系还有很大变数。

机器人目前还不善于铺床叠被

目前来看，在不可预见的环境中进行体力工作或操作器械的工作，相对而言难以被自动化。例如在建筑工地操作吊车，或是在公共场所清扫垃圾，或是在宾馆整理床铺。最后一项属于不可预见的环境，是因为客人可能会把枕头扔到不同的地方，或者把衣服留在床上，让机器人很难自动进行客房清洁。不过这种情况可能很快就会改变，人们已经开始进行重点研究，来改善机器人在不可预见的物理环境中的表现。

是时候去当老师或洁牙师了？

最难利用当今的技术自动化的是那些管理和培养人才的工作（只有9%有自动化的潜力），决策、规划、创造性的工作（18%）或是与客户、供应商和其他股东互动的工作（20%）。这些工作涉及编写软件、创建菜单、撰写宣传材料——或是建议顾客哪种颜色的鞋子最合适，经验和阅历往往很重要。

在医疗行业中，注册护士的工作只有不到30%可以被自动化，而在洁牙师的工作中，这一比例降到了13%。在我们研究的所有行业中，最不受自动化影响的是教育。教育需要很强的专业能力，以及与他人交流的复杂技巧，迄今为止，机器在这方面的表现很不完备，只有极少数情况例外。

机器会改变工作，但它们无法完全取代人类。分析自动化的技术可行性时，最好不要看整个岗位，而要看各种工作内容所消耗的时间，以及利用现有技术可以自动化且能应用于职场的工作内容占比。整体来看，我们发现只有5%的岗位可以用现有技术完全自动化。然而，现有技术却可以让付酬工作中45%的内容自动化。此外还有超过30%的岗位，它们的工作内容有约60%可以被自动化。

请继续留意这一领域

科技会随着时间不断发展，技术可行性也在不断进步。本文中的分析目前关注的只是现有技术，但是我们会进行持续的研究，考虑科技发展的不同情况。随着科技的发展，例如机器可能会掌握自然语言，拥有与普通人类似的能力，届时在技术上可以自动化的工作类型就会增加。

自动化也会从根本上改变机构的性质。未来管理者面临的挑战，在于他们需要考虑用机器取代人力的成本，以及在改变后的办公环境中调整商业流程的复杂程度，从而确定如何用自动化改变公司，怎样发挥自动化的价值。自动化带来的主要效益，可能并非来自人力成本的降低，而是源于减少错误率、提高产出、改善质量、安全性和速度导致的生产效率的提高。 （财富中文网）

译者：严匡正

Automation is often depicted in news articles and some academic literature as a titanic struggle between man and machine, in which the machine seems destined to win and the only question is how soon to schedule the medal ceremony. As automation in the workplace moves rapidly from science fiction to business fact, we believe the changes it will bring are more nuanced than a simple choice between human and robot. We have examined 2,000-plus work activities in every industry sector across the US economy. Here are eight findings from our research, highlighting the automation potential of each sector—providing some technical indicators as to which occupations are most and least likely to go to a machine.

Just because something can be automated doesn’t mean it will be.

While technical feasibility is a necessary precondition for automation, other factors are required to build a compelling business case. These include the cost of developing and deploying the hardware and the software for automation, and the supply-and-demand dynamics of labor. Replacing human cooks earning $10 per hour with expensive robots may be possible technically, but might not make business sense because it may cost too much and not provide a good return on investment. Regulatory and social issues could also be factors many hospital patients will want a human nurse rather than a robot to care for them when they wake up after surgery.

Certain physical jobs have the highest potential to be automated.

Almost one-fifth of the time spent in US workplaces involves performing physical activities or operating machinery in a predictable environment—that is, specific actions in familiar settings where changes are relatively easy to anticipate. More than three-quarters of such activities could be automated already with today’s technology, and figure prominently in manufacturing and food service, making these sectors the most technically susceptible to automation. Robots on factory floors already do repetitive rote tasks such as product assembly and packaging, while in food service, some restaurants are testing self-service ordering or even robotic servers.

Data collection and processing are ripe for automation, too. Across all occupations in the US economy, workers spend one-third of their time collecting and processing data. Both activities are highly like to be automated and could affect industries, from retail to financial services and insurance. Workers won’t necessarily be out of the job, but their roles may very well change. For example, mortgage brokers spend as much as 90% of their time processing applications, and could instead spend more time advising clients.

Even high-paying jobs will be affected. It’s not just entry-level workers or low-wage clerks who collect and process data; people whose annual incomes exceed $200,000 spend more than 30% of their time doing so, too. That makes activities in these jobs attractive for companies to automate. Overall, the correlation between wages and automatability shows a great deal of variability.

Robots aren’t great at making beds — yet.

For now, activities that require physical movement or operating machinery in unpredictable settings are relatively challenging to automate. Examples include operating a crane on a construction site, collecting trash in public areas, or making beds in hotel rooms. The latter is unpredictable because guests throw pillows in different places, or may leave clothing on their beds, which makes it hard for a robot to carry out maid service. This might change soon, however, as significant research is being devoted to improving the performance of robots in physically unpredictable environments.

Time to become a teacher or dental hygienist?

The hardest activities to automate with the technologies available today are those that involve managing and developing people (9% automation potential), where expertise is applied to decision-making, planning, or creative work (18%), or interacting with customers, suppliers, and other stakeholders (20%). These activities, where experience and age are often an asset, can be as varied as coding software, creating menus, writing promotional materials — or advising customers which color shoes best suit them.

In health care, less than 30% of a registered nurse’s job could be automated, while for dental hygienists, that proportion drops to 13%. Of all the sectors we have examined, among the least susceptible to automation is education. The essence of teaching includes deep expertise and complex interactions with other people for which machines, so far and with few exceptions, receive an incomplete grade.

Machines will change jobs, but they won’t fully take over from humans. The technical feasibility of automation is best analyzed by looking not at occupations as a whole, but at the amount of time spent on individual activities, and the degree to which these could be automated by using technology that currently exists and adapting it to individual work activities. Overall, we find that only about 5% of occupations could be fully automated by adapting current technology. However, today’s technologies could automate 45% of the activities people are paid to perform across all occupations. What’s more, about 60% of all occupations could see 30% or more of their work activities automated.

Watch this space.

Technology continues to develop and technical feasibility will thus evolve over time. This analysis has focused only on currently available technologies, but our on-going research considers different scenarios for technology development. As technological advances such as machines being able to acquire natural language abilities that match median human capability, the numbers and types of activities that are technically susceptible to automation will increase.

Automation will fundamentally change the nature of organizations. The challenge for managers will be to identify where automation could transform their organizations, and then figure out where to unlock value, given the cost of replacing human labor with machines and the complexity of adapting business processes to a changed workplace. Most benefits may come not from reducing labor costs but from raising productivity through fewer errors, higher output, and improved quality, safety, and speed.