An easy way to learn whether your dream occupation—or perhaps your current one—is at risk of being “wiped out” by AI, or to what extent it is or will be affected by AI, is to simply look for it in a list. One such list was published in 2013 by two researchers¹, sparking not only a wave of apocalyptic news headlines but also a series of further publications on the topic. It ranked occupations based on their likelihood of being automated. I’m not aware of a recent list as comprehensive as that one, but there are some, such as the one from the Burning Glass Glass Institute², which ranks selected types of occupations by their GenAI exposure score.

You can decide if that is enough. You have your answer—case closed!
Or you can be more critical and explorative, seeking a deeper understanding.
You should be aware that these studies are not the final word, nor are they set in stone. They claim to use advanced machine learning techniques, but at the end of the day, these are just probabilities. Besides, adjusting certain parameters in the research can affect the results³.
This is to say: don’t take them at face value.

Before jumping into the steps, let’s first make sure we have the basics right. Remember, an occupation consists of various tasks, each with varying degrees of automation. This means that even if a couple of tasks can be fully automated, it does not necessarily mean that the entire occupation—e.g., lawyer—can be fully automated, nor that lawyers should start looking for a completely different career.

With that in mind, here are the steps:

1. Understand AI’s strengths and weaknesses
2. Explore official occupation databases
3. Gather insights from experienced professionals
4. Be the final judge

1. Understand AI s strengths and weaknesses: For this, let’s consider the following tables, which includes a selection from key authorities in the field of automation, AI, and employment.

	AI Strengths	AI Weaknesses
Burning Glass Institute (2024)	– Streamline interactions with business software. – Generate text, answer questions and serve as conversational agents – Generate code and assist with code debugging and comprehension – Generate images, music and videos – Refine and enhance the style, coherence and quality of existing content – Summarize and classify text – Retrieve and present enormous amounts of information quickly.	– Not especially creative or original; may produce derivative content – Limited critical thinking – Low emotional intelligence – Limited factual accuracy – Challenges with mathematical functions – Likely to reproduce biases in training data
Frey and Osborne (2013)	Good with routine tasks and non-routine where big data is available	– Perception and manipulation tasks {finger and manual dexterity4, cramped work space or awkward positions (e.g. various piled objects in an environment such as home) – Creative intelligence (originality and fine arts) – Social intelligence (social perceptiveness, negotiation, persuasion, assisting and caring for others).

Mike Wooldridge’s Capabilities of AI/machines, compared to humans [Reorganized based on Wooldridge’s presentation]

Green = AI excels at it (e.g. through Large Language Models – LLMs) Red = Not present in AI to the level of humans Black = maybe a bit of it

Natural Language Processing Common sense reasoning Recalling Planning Problem solving Arithmetic Sense of agency Manual dexterity and manipulation Proprioception Vision understanding Audio understanding Hand-eye coordination Navigation Mobility (Multi-agent) coordination Logical reasoning Abstract reasoning Social reasoning Rational mental state Intentionality

Edinburgh Laboratory for Integrated Artificial Intelligence (ELIAI) about deep learning’s weakness

– Making generalizations when pre-training on all possible scenarios is impossible – Dealing with changing situations and causality – Creativity – Being able to explain predictions and decisions – Complex reasoning

Another way to look at this is by considering layers of intelligence, starting with the AI lover’s dream: a machine that can do anything, just like a human (at the top):

Varieties of general intelligence	State of the machine/AI capabilities
Machines that can do anything a human can do (General AI)	Not imminent [E.g. robotic AI is not as near as advanced as GenAI-LLMs]
Machines that can do any cognitive task a human can do	Closer but not there
Machines that can do any language-based task that a human can do	Not there yet but not far off from it
Augmented LLMs	Imminent

Credits: table elaborated by me, based on Wooldridge’s presentation. In other words, Wooldridge is the author of the table’s content.

2. Exploring an official occupation database: this is to gain a more detailed understanding of the occupation you are interested in. This means understanding what the occupation entails, the types of tasks it consists of, and the associated skills, among other factors.

Thankfully, databases such as O*NET⁵ include a wealth of useful information.

For practicality, let’s take the description of one occupation from O*NET.

As an example, the table below presents a comparison between the occupation’s required skills and AI’s strengths and weaknesses

Financial Quantitative Analyst / Skills (Doesn’t include a bunch of specific software skills / includes the top 5 skills. The complete list contains 18 skills)	State of AI capabilities
Mathematics — Using mathematics to solve problems	AI good at crunching numbers [not so with complex mathematical functions and complex arithmetic]
Critical Thinking — Using logic and reasoning to identify the strengths and weaknesses of alternative solutions, conclusions, or approaches to problems	Limited
Reading Comprehension — Understanding written sentences and paragraphs in work-related documents	Excels at summarizing huge amount of texts, but limited understanding
Complex Problem Solving — Identifying complex problems and reviewing related information to develop and evaluate options and implement solutions	(Very) limited
Active Learning — Understanding the implications of new information for both current and future problem-solving and decision-making	Limited

What does the table suggest? Do you feel that the occupation will “disappear”? Do you notice the skills you can focus on as your competitive advantage over AI?

***

If you are a good observer, you surely noticed that the Financial Quantitative Analyst occupation is marked with a bright outlook by O*NET. On the other hand, it is considered one of the most vulnerable to GenAI by the Burning Glass Institute.
Who is right? Time will tell

***

3. Gathering insights from experienced professionals: Talk to at least 2–3 professionals in the occupation of your interest and learn about the changes they are experiencing. I suggest you preferably pick those with various years of experience, as they will be in a better position to offer you a historical perspective on how tasks and skills have been evolving in the field.

4. Being the final judge: At the end of the day, even though some occupations do indeed disappear, most of the time what we have is transformation. This means that the same occupation changes in terms of the skills associated with it. The key is to understand what those changes are and what the underlying nature of those changes is.

Human beings have mastered the art of integrating information to draw logical conclusions. AI is just a tool—a very advanced one, but still a tool. It fails when exposed to what it was not trained on. Does it even learn true causal relationships? Does it even have consciousness? Google Translate doesn’t use dictionaries and doesn’t even know it is dealing with languages. It uses billions of examples, follows patterns, and produces results based on the parameters it was trained on. That’s basically the same with LLMs, which are at the core of GenAI.

On the other hand, remember that the world is becoming more analytical. Therefore, there is a growing need for high-value skills (including soft skills such as critical thinking), as well as high-value knowledge or expertise. Wikipedia-type knowledge has already been dead for many years. Last but not least, because it is more about teaming up with AI, you’d better start developing AI literacy as well.

============================================

1. “The Future of Employment: How Susceptible are Jobs to Computerisation”; Carl Benedikt Frey and Michael A. Osborne, 2013.
2. Generate AI and the workforce, Burning Glass Institute – SHRM. In another study, they provided a list of hundreds of occupations, ranked by the (change and transformation of) skills disruption.
3. One such study that challenges Frey and Osborne’s was conducted by Melanie Arntz, Terry Gregory, and Ulrich Zierahn (“Revisiting the Risk of Automation”). They took into account the spectrum of tasks within occupations, concluding that the automation risk of jobs in the USA was around 9%. That’s significantly lower than Frey and Osborne’s estimation. On the other hand, according to McKinsey, very few occupations—less than 5%—consist of tasks that can be fully automated.
4. In recent years, there have been advances in AI and robotics. Do robots already have the level of manual dexterity that human beings possess? In controlled environments, most likely they do, or they even have superior ability. In unpredictable environments, most likely they are not there yet. In any case, relying too much on finger and manual dexterity as a competitive advantage over machines might be a bit risky. According to a recent study by the World Economic Forum (2025), that skill—finger and manual dexterity—has been losing importance among employers.
5. O*NET is an occupational database developed by the US Department of Labor, widely used internationally by many institutions. There have been studies that have tested its applicability to other developed countries, and it holds up fairly well. If you are from a developing country, you might want to be cautious, as its applicability might have some limitations. Nevertheless, it is said to be a fairly good reference