Just a few days ago Anthropic dropped a study that looked at millions of real AI interactions to understand which jobs are most exposed to AI and which are less affected. Instead of speculating about what AI could do in theory, the researchers looked at how people actually use AI tools in professional contexts today. Based on this data, they developed a new measure called “observed exposure” – essentially estimating how much of the work inside different professions is already being supported or automated by AI.
The results are quite revealing.
Jobs Most Exposed to AI Take Over
Jobs most exposed to AI today tend to be those that involve digital, cognitive, and text-based work, such as:
- customer support and operations
- software development
- data analysis
- marketing and content creation
- finance and research roles
These are roles where a significant share of everyday tasks involves analyzing information, producing text, writing code, or handling structured data – exactly the kinds of tasks large language models are designed to assist with.
In some of these occupations, AI tools are already capable of assisting with 30–50% of the underlying tasks, and in specific roles the coverage can be even higher. For example, programmers, customer service representatives, and data entry specialists show some of the highest levels of AI task coverage in the data.
However, one of the most interesting insights from the report is that we are still far from using the full potential of AI.
In the study’s graph, the blue area represents the share of tasks that AI could theoretically perform, while the red area shows what people are actually using AI for today. Across almost every professional category, the red area is much smaller than the blue one (see image below).
For example, in computer and mathematical occupations, AI could theoretically support over 90% of tasks, yet real usage today covers roughly one third of that potential. In other words, the technology is already capable of helping with far more work than we currently allow it to.
This gap between capability and adoption may turn out to be one of the biggest productivity opportunities of the coming decade.
Jobs Less Exposed to AI Takeover
On the other end of the spectrum, the study shows that some professions remain much less exposed to AI – at least for now. These tend to be jobs that require physical presence, manual work, or direct interaction with the physical environment, for example:
- construction
- installation and repair
- transportation
- food services
- healthcare support roles
What these occupations have in common is that they rely heavily on physical coordination, hands-on work, and situational awareness in real environments. Fixing a pipe, preparing food in a busy kitchen, transporting goods, or assisting patients in a medical facility involves constant interaction with the physical world – something current AI systems simply cannot do on their own.
Large language models are extremely good at processing information, generating text, analyzing data, and supporting digital workflows. But they cannot climb a ladder, repair a heating system, lift a patient, or adapt their movements to an unpredictable physical environment.
Because of this, the study finds that occupations dominated by manual or physical tasks show significantly lower exposure to AI assistance today.
However, it is important to interpret this carefully.
Lower exposure does not necessarily mean these jobs are permanently safe. It simply reflects the current technological reality: generative AI primarily operates in the digital world, not the physical one. For now.
As robotics, automation, and AI-powered machines continue to develop, even some of these roles may eventually be reshaped. But compared to knowledge work, the timeline for disruption in these fields is likely to be slower and more gradual.
What Does it Mean to All of Us
(no matter the profession we have today)
Another important point this research shows is that it doesn’t say jobs will disappear. It shows that tasks inside jobs are changing.
It all reminded me of a mentoring conversation I had about 6-7 years ago. A woman came to me worried that her job is at risk. She felt pressure to pivot quickly into something “in demand”, and she was seriously considering retraining as a software engineer.
I told her something that made her freeze in a pause:
“You see how fast the market is shifting. You believe you need to be in the center of it – so now you choose software engineering. But imagine that with this speed of change, software engineering becomes highly automated in five years. What will you do then?”
She froze for a moment and said:
“I didn’t think about it like this…this hits hard. You’re right. With my logic, I would be looking for something else. But I don’t want to be a forever junior in a role, because I jump into something new every 5 years.”
That day we didn’t design a career plan around chasing demand. I helped her reflect on different questions instead:
- What actually energizes you?
- What strengths show up naturally in your work?
- What is that you really enjoy doing and why?
- What values do you want your career to express?
By the end of the session, I told my mentee:
“So now, after half an hour of exploration, it sounds like you actually don’t want to become software engineer. It is not about the role. What you want is to be relevant. And this is great! What do you feel you need to do, independently of the role you are in, to keep being relevant?”
She paused for a moment, and then I seen this mental lightbulb switching on, as her eyes got that spark that we usually have when we finally get the answer we were looking for for so long:
“I need to constantly learn new things! In which ever role I will be, I need to learn new tricks, new tools, new approaches. And then I will be changing together with my job, instead of running away into the next “flashy” thing. And it now, for the first time, sounds like a plan! I can do this!”
I do believe that if we would constantly learn and constantly improve in our roles – we can change together with them. But once again – this is a human capacity. AI will not learn for you. But it can, for sure, empower your learning speed.
Another proof to stay in your role, even if you are an engineer who’s job is now fast-tracked to be automated with AI, is the current trend of new roles emerging. Companies now look for professionals with the same technical skills, but with a twist:
- prompt engineers (only today I got two requests from recruiters in my circle for a referral for this type of role)
- AI/ML engineers
- LLM engineers
- context engineers
- data & intelligence engineers
No one yet gives you a solid opportunity to get a degree in these roles or provides an extensive and science backed training in these professions. So what can you do to keep up?
My answer is – be curious, try to learn constantly, and adapt in your current roles as you go. This is a “human” skill of being eager and fast to learn, being curious, being agile is what will set you apart and set you up for success. And it being a “human” skill makes it available to you already.
That’s the approach I believe in: we need to connect to that “human” in us first and not be scared of the “artificial”.
Focus on Developing Your “Human” Skills
The good news is I believe technology may actually push us to become more human.
As AI takes over routine analysis, documentation, and operational tasks, the value of human contribution shifts toward things machines still struggle with. And in addition to being an agile learner, here are the core “human” skills I beleive we need to focus on:
- emotional intelligence
- leadership that inspires and influences
- complex collaboration
- judgment in ambiguity
- creativity and strategic thinking
- curiosity and growth mindset
Trying to chase “safe” jobs is a losing strategy.
The market will keep shifting. Even jobs that look safe today may change tomorrow. We might all run to become plumbers… until robotics arrives there too.
So the real question is not which job is safe. The real question is:
How do you remain relevant no matter how the market changes?
And the answer, in my view, lies much deeper than the job title you currently hold.
The future belongs to people who never stop developing themselves.
Psychologist Carol Dweck described this beautifully through the idea of the growth mindset – the belief that our abilities can be developed through effort, learning, and persistence. This mindset becomes even more critical in a world where technology evolves faster than any static career plan.
AI may automate tasks. But it cannot replace the human capacity to grow.
And perhaps this is the most interesting paradox of this technological era:
The more powerful artificial intelligence becomes,
the more valuable human intelligence becomes.
If you want to explore this idea further, you might find these resources useful:
- Mindset: The New Psychology of Success by Carol Dweck – a foundational book on growth mindset and learning.
- My book Morning Coffee with Yourself: Become Your Own Coach in 20 Days – a practical guide to building self-awareness, clarity, and personal direction through self-coaching.
Because in a rapidly changing world, one of the most valuable skills you can develop is the ability to coach yourself through change.
And that is a skill no technology can automate.