The last couple of decades have seen a lot of advancement in automation technologies such as robotics, artificial intelligence, and machine learning.

As these technologies increasingly become a part of our day to day life, there has been increasing concern about their effect on the workplace.

For centuries, technological advancements have always displaced some portion of the population from the job market, and therefore, it is inevitable that these automation technologies will take some of our jobs.

The question is, which jobs will machines take over, and which ones are safe (for now)?

Previously, advances in technology displaced humans from predominantly manual jobs, jobs that required physical strength, speed, and dexterity.

Jobs that were more mental than physical were immune to machines.

If you think that the current automation technologies will stick to this script, however, think again.

Today, machines are becoming better at tasks that were previously thought to be the preserve of humans, and as a result, they are coming for all sorts of jobs, even the white collar jobs we thought they couldn’t take away from us.

For instance, according to this article by Think Growth, AI is better at analyzing sales calls than any sales manager.

Actually, for a human sales manager to analyze over 25,000 sales calls, it would take them about 9 years of non-stop working, yet this is something AI can do within a couple of hours.

Marketers are also relying on AI to come up with email marketing playbooks and content strategies. With time, these AI programs might start playing a bigger role in these occupations.

What this shows is that there is no limit to the kind of jobs that might end up being taken over by machines.

In a bid to find out where machines could replace humans, and where they can’t yet, McKinsey conducted a detailed analysis of over 800 occupations and over 2000 work-related activities to determine which jobs were most and least susceptible to automation, based on the technical feasibility of automating each job.

Below, let’s take a look at some of the findings from the McKinsey report.


When determining the automation potential of a particular job, McKinsey were looking at the possibility that currently demonstrated technologies could be used to automate a given activity.

In other words, they were checking whether the technology to automate activities related with that job exists currently.

This is what is referred to in the article as technical feasibility.

It’s good to note that while the existence of the necessary technology is a precondition for automation, it does not necessary mean that an activity will be automated.

This is because automation depends on several other factors. These include:

  • The cost of automation: This refers to how much it will cost to develop and deploy the systems – both hardware and software – that are necessary for automation.
  • Cost and relative scarcity of labor and skills: If there is an abundant supply of workers within a certain occupation, and if the cost of labor in that occupation is significantly lower than the cost of automation, then there might be no incentive for automation.
  • Benefits of automation: What benefits are to be gained from labor substitution? These include increased levels of output, increase in product quality, reduced errors, and so on.
  • Social acceptance and regulatory issues: This is the extent to which replacement of human labor with machines will be accepted by society. For instance, while it might be possible to use robots to take care of sick patients, will the patients be okay with being taken care of by robots instead of other humans?

Automation depends on the interplay between technical feasibility and these other factors.

In this article, we are going to put the greatest focus on technical feasibility of automation, though we will also touch on other factors here and there.


So, which jobs will machines and AI programs take, and which ones will be spared?

According to McKinsey, this is the wrong approach to analyzing where machines will replace us and where they cannot.

This is because there is a difference between jobs and tasks/activities.

The experts at McKinsey argue that the right question to ask is “on which tasks/activities will machines replace us?”

Analyzing the work activities related to a certain job – rather than the job in its entirety – provides a more accurate way of determining the technical feasibility of automation, because each job involves a variety of fundamentally different activities.

For instance, a job like “travel agent” involves several different tasks, some of which have a high technical feasibility for automation, while others have a very low technical feasibility for automation.

For instance, a travel agent will need good knowledge of geography to suggest the best places to visit in each country/continent.

They will also need the ability to understand and manage train and airline schedules.

These are tasks that machines are capable of performing better than humans.

On the other hand, the travel agent also needs the ability to connect with customers, understand their dreams and hopes, and sell the travel package that is best suited to them.

This task is a lot more difficult to automate.

As you can see, you cannot accurately determine whether the job of a travel agent has a high automation potential without taking a look at the tasks/activities associated with the job.

According to McKinsey, robots, machines, and AI programs will take over many tasks over the next decade – probably more than half of all the activities humans do to earn a living today.

However, in most cases, machines will take up only a portion of the activities associated with the occupation, rather than taking over the entire job.

This will lead to increased collaboration between humans and machines, without necessarily pushing humans out of the workplace completely.

Actually, this might even lead to an increase in the number of people employed within these occupations as demand for the remaining activities goes up.

A good example of this is the correlation between ATMs and employment of tellers.

When ATMs were introduced, it was believed that they would eliminate teller jobs.

Surprisingly, however, as more ATMs were installed, teller jobs increased instead of decreasing.

This is because ATMs made it cheaper to operate a bank branch, which allowed banks to open more branches, thereby increasing the demand for tellers.

In addition, the tellers could put more focus into customer relationship management, rather than the routine tasks of depositing and withdrawing customers’ money.

The same thing happened in the retail industry, where introduction of point of sale systems led to an increase, rather than a decrease, in the demand for cashiers.


According to the report by McKinsey, the most automatable activities are those that involve operating machinery or performing physical activities within a predictable environment.

Almost 20% of the time spent in workplaces in the United States involves such activities, where workers undertake specific (and often routine) actions in a setting with minimal changes, or where the changes can be easily anticipated.

McKinsey reports that there is a 78% technical feasibility of automating predictable physical activities.

Most of the activities that fall under this category are most prominent in industries such as manufacturing, retail, food service and accommodations, and so on.

Therefore, based solely on technical considerations, these industries are some of the most susceptible to automation.

Just because these industries are highly susceptible to automation does not necessary mean that you should start looking for a new career if you currently work within these industries.

For instance, if we take a look at the manufacturing industries, majority of the activities within the industry involve operating machinery or performing physical activities in a predictable environment – cutting materials, loading materials to production equipment, assembling, welding, packaging products, maintaining equipment, and so on.

If these activities form the major bulk of your work, there is a high chance that your job will be taken over by a machine.

However, the manufacturing industry also involves other activities such as customer service, administration, marketing, procurement, and so on.

These jobs do not involve operating machinery or performing physical activities in a predictable environment, and are therefore not at a high risk of being taken over by machines.

Just like manufacturing, the retail sector is highly susceptible to automation, but it doesn’t mean that you should start looking for a new career either.

A lot of activities in the retail sector are high predictable and can therefore be automated.

These include stocking stock management activities, packaging products for shipping, logistics, keeping sale records, gathering product information or customer data, and so on.

Actually, many retailers have automated some of these activities, such as stock management and logistics.

However, there are many other retail-related activities that require cognitive and social skills.

For instance, advising a client which toy they should buy for their 5 year old or which dress looks better on them requires emotional intelligence and judgement and cannot therefore be easily automated.

It’s also good to note that the fact that the technology to automate an activity exists does not necessarily mean that the activity will be automated.

Remember, there are other factors that influence automation besides technical feasibility.

For instance, many of the tasks within the food service industry are highly predictable and can therefore be easily automated.

Some restaurants have already made automation tests with things like robotic servers and self-service ordering.

You have probably even heard about the Momentum Machines hamburger cooking robot, which is capable of preparing a record 360 burgers within an hour.

All this shows that there is high technical potential for automating several cooking and food preparation activities.

Still, when you take into consideration the cost of automation, the benefits to be gained from automation, as well as the availability and cost of labor, it might not make sense to automate these activities.

For instance, considering that restaurant employees who perform these activities are typically paid about $10/hour, there aren’t really any significant cost savings to be made from automation.

Considering the impact prior technological advancements have had on the workplace, one would be tempted to assume that jobs that require low levels of skill – such as cooking in a restaurant – are the most susceptible to automation.

In some cases, however, jobs that need skills and training might have more potential for automation compared to some lower skill jobs.

For instance, jobs such as accounting, bookkeeping and auditing require more skill compared to a basic cook.

However, it costs less to automate the activities of a bookkeeper – all you need is a computer and some basic software – compared to automating the tasks of a cook, where you might need to build a specialized robot.

In addition, there are more benefits to be gained from automating bookkeeping compared to automating cooking.

As a result of this, you might notice that automation rates are higher for activities that require middle-level skills, such as data gathering and data processing activities.

This is because these activities are fairly easy to automate and result in huge benefits, such as increased efficiency and huge cost savings.

As advancements are made in machine learning and artificial intelligence, however, we might start seeing an increase in automation rates for jobs that require higher skills.


Over 30% of the time spent in workplaces in the United States is spent on activities that revolve around collecting, processing and manipulating data.

According to McKinsey, such activities have more than 60% technical potential for automation.

When most people think about data collection, processing and manipulation, they think about activities such as generating invoices, keeping record of sales, processing payrolls, administering procurement, keeping track of stock levels using barcodes, and so on.

These tasks have a high potential for automation, and many companies automated most of them a long time ago.

What many do not realize is that data collection and processing is not limited to these common examples.

A lot of jobs today, including high level jobs, involve some degree of data collection and processing.

Actually, McKinsey reports that data collection takes about 31% of the working time of people who make over $200,000 per year.

Good examples of such occupations are financial services and insurance.

Both of these occupations rely heavily on professional expertise, and it might be hard to imagine that such occupations can be automated.

However, for both occupations, about half the time is spent on collecting and processing data. Stock traders have to collect and analyze data about the stock markets.

Tellers spend much of their time verifying the accuracy of financial data.

Investment bankers have to analyze data about companies. Insurance agents have to collect client data and use it to make decisions.

All these activities have a high technical potential for automation. Automating them will free up a lot of time and allow people within these occupations to focus on activities that add more value for their clients, rather than routine processing.

Aside from data collection and processing, other activities that fall in the middle range when it comes to automation potential are activities that involve operation of machinery or performing physical activities in environments that are highly unpredictable.

Such kinds of activities include tidying up hotel rooms, operating cranes and other machinery within construction sites, setting up classroom equipment and materials, collecting trash in public areas, and so on.

With such activities, automation is difficult because the working environment is constantly changing; therefore it is difficult to create a model for the automation software to learn from.

For instance, when it comes to tidying hotel rooms, different guests will leave their rooms looking different.

Some will leave the pillows on the bed, others will leave the pillows on the seat, others will leave their clothes on the bed, and some will have clutter on the floor, and so on.

Similarly, emergency situations occur randomly, and therefore, any first responder going to offer medical care in such situations cannot possibly predict what they will find on the ground once they get to the scene of emergency.

This unpredictability makes it difficult to automate these activities.

Currently demonstrated technologies do not have the flexibility required to work in such unpredictable environments, and therefore McKinsey reports that the automation potential for such activities stands at about 25%, though this will increase once technological advancements make it possible for machines to handle such environments as easily as they can handle predictable environments.


The activities that are least susceptible to automation with currently demonstrated technologies are those that involve directly interacting with or managing people, with a 9% automation potential, and those that involve the application of expertise in planning, making decisions, or performing work that is creative in nature.

Examples of such activities include writing marketing material, coding software, taking care of sick patients, teaching, and so on.

While industries that rely on knowledge work or human interaction are the least likely to get automated, some of the activities related to these industries have some level of automation potential.

For instance, let’s take a look at industries like healthcare and education, where human interaction is absolutely critical.

When it comes to occupations such as nursing, only about 30% of their activities can be automated with currently demonstrated technologies.

With occupations such as dental hygienists, only 13% of their activities can be automated.

Still, there are some activities within the healthcare sector that are ripe for automation.

For instance, administering intravenous medications, preparing food for patients, collecting and analyzing patient data, reading radiological scans, and administering anesthesia for simple procedures take a considerable amount of working time, yet they can easily be automated.

Similarly, when we look at the education sector, a lot of activities rely on human interaction and therefore have very little potential for automation.

Still, 27% of activities in the education sector can still be automated, particularly activities that do not place in the class.

For instance, activities such as preparing and serving food for students, janitorial and cleaning services, administration services, maintenance of student information, maintenance of records, and so on are still crucial for schools to run smoothly.

These activities can be automated, which would result in a reduction in the cost of education without affecting its quality.

For now, those who work in industries that rely heavily on deep expertise, creativity and interactions with other people have the lowest chance of having their jobs taken over by machines with currently demonstrated technologies.

However, with advancements in machine learning and artificial intelligence happening rapidly, machines will gradually start coming for more and more of the activities associated with these jobs.


As we have seen, there are several activities that are unlikely to be automated with current technologies.

Still, we can expect that with time, technological advancement will make it possible to automate activities that currently have low automation potential.

For instance, various companies are working on technologies that will make it easier and safer for humans and machines to collaborate on physical tasks in unpredictable environments.

As these technologies mature, we will start seeing increased automation in these tasks.

We have also seen above that activities that require human interaction are the least susceptible to automation.

This is because machines do not understand the nuances of human communication.

Today, several companies are working on various natural language processing techniques with the aim of making machines better at understanding human speech.

Once machines learn to recognize and understand the nuances of everyday communication between people, we can expect them to take up more and more of activities that require human interaction.

Therefore, don’t be surprised when you child comes home one day with stories about their robotic teacher.


Being aware of activities where machines could potentially replace us will give business leaders the opportunity to rethink how their employees engage with their jobs and to figure out the best approach to managing their workforce in a world driven by technology.

It will also push business leaders to start thinking about how they could use machines and automation to make the execution of some activities more efficient and free up their staff to focus on strategic roles that will drive the greatest value for their organizations.

In this regard, business leaders should keep in mind that the greatest benefits of automation will not come from reducing labor costs, but from the improved productivity resulting from increased speed of production, improved quality of goods and services, and reductions in errors.

Where Machines Could Replace Humans--And Where They Can’t (Yet)

Comments are closed.