AI Career — choosing a direction while AI reshapes work

01 Who this is for

Two different roads, the same question

Two very different people end up here, and both deserve to be taken seriously. Their question is, in the end, the same one: “what do I do with the years ahead?”

Under 18

You are choosing now something you finish in six years

The pressure is enormous, and a lot of the advice you get is good advice for the job market of ten years ago. Nobody can tell you for certain what will be in demand when you graduate — we cannot either. But the right question is not “what is in demand today”, it is “what will still be hard to automate by then”.

We talk to you like a person, not like a child: we will not tell you what to choose. The choice is yours and you are the one who lives with it.

What you get: the criteria to judge any job on your own, including one that never appears on this page.

Mid-career

You worked for years at something that is now done automatically

It is hard, and the fear is legitimate — there is no point pretending otherwise. But nothing you have done so far was wasted. Your judgement, the people you know, the way you understand a field, the patience to see something through: all of that transfers. The tool changes, not the person.

You are not “too old” and you did not train for nothing. You simply have more things to weigh than someone who is eighteen.

What you get: we look first at what you already have, and only then at what you might still need. The second list is usually shorter than it looks.

Are you a parent reading this for your child? You are welcome here. One thing though: this page is written so you can read it together. A child is not helped by someone handing them the answer, but by someone giving them the criteria and letting them choose. The form has an option for parents.

02 The foundation

AI rests on something very physical. And it keeps growing.

A data centre does not run on enthusiasm. For a model to answer a question, somewhere there has to be power, cooling, water, cables, a building and people who fix things at three in the morning. The more the world goes digital, the bigger that foundation gets — not smaller.

What you see Apps, chat, screens, “it just happens”

Power, at industrial scale

Not “energy” in the abstract: lines, substations, transformers, panels, backup generators, and people licensed to touch them. Someone installs them, someone checks them, someone repairs them.
Cooling and water

Servers throw off heat. If cooling stops, computing stops too — not in a day, but very quickly. Chillers, pumps, pipes, ventilation, filters: all of it is installed, tuned and repaired by hand.
Cables and fibre

Between “the cloud” and you there are real routes: dug, pulled, terminated, spliced, and repaired after a digger goes through them. None of that happens remotely.
Buildings, roads, structures

Foundations, halls, doors, ventilation, an access road a transformer weighing several tonnes can actually be driven in on. Before there is a model, there is a building site.
People on shift

Someone answers at three in the morning when the alarm goes off. Someone climbs the ladder, someone checks what the sensor is showing, someone decides whether to shut something down. That is a job, not a cloud service.

The owner of this site puts it like this: every AI needs power — and not the kind you get from leaving two windows open. The joke has a serious core: a data centre consumes on an industrial scale, not the amount a few panels on a roof produce. We will not give you numbers, because the numbers going around on this topic are usually recycled from old studies. But the direction is plain to see: the need for power, for cooling and for people who keep them running does not shrink because AI showed up.

03 Criteria

How to recognise a job that is hard to automate

This part is worth more than the list of jobs that follows. With the criteria below you can judge any job, including one we never thought of. No job ticks all five, and none of them is safe forever — but the more boxes a job ticks, the harder it is to move onto a server.

The physical, unpredictable world

The work happens on site, where every situation is different: an old house, an installation done badly thirty years ago, a space you can barely fit into. A model can describe what should be done. It cannot reach in there.

Presence and human touch

Someone has to be there, physically, next to another person: to lift, to change, to hold, to talk. Presence does not travel through a screen.

Responsibility and a signature

Someone has to answer to the law for what they did. A signature has a name behind it and consequences attached. A program cannot be held accountable.

Trust between people

People choose the person, not the procedure. Trust is built over time, with a face and a name, and it cannot be copied from anywhere.

Improvising where there is no manual

The situation has never happened in exactly this way. You do not have the right part, it is night, it is raining, and it has to hold until morning. Here experience counts, not the statistical average.

Notice what is missing from the list: “it is hard” or “it takes years of school”. That is not what matters. What matters is how much of the job happens in the real world, with real people, under someone's responsibility.

04 The five jobs

Five jobs AI cannot do in your place

We will put it honestly: they are not “safe forever” — nothing is. They are hard to automate, and we write out exactly why, using the criteria above. For each one we also say which part does get automated, because a list that hides that is of no use to you.

01

Electrician

C1 physical world
C3 responsibility
C5 improvisation

The irony is the best argument on this page: everything AI does depends on electricity, and electricity is pulled, connected and repaired by people. Not only in data centres — in apartment blocks, in factories, in hospitals, in charging stations, in everyone's home.

Every installation is different, and a good part of the job is working out what the person before you did (C1). At the end you sign and you answer for it: a badly done electrical installation has a name on paper (C3). And when the power goes out in a factory at two in the morning, no procedure covers that exact situation (C5).

What does get automated: cable sizing, the quote, the drawing, the materials order, part of the diagnosis. The desk part gets shorter. The part where someone climbs the ladder does not.

02

Plumber, water, cooling and ventilation technician

C1 physical world
C4 trust
C5 improvisation

A data centre without cooling is not a data centre, it is a hot room that shuts down. Chillers, pumps, pipes, ventilation: someone installs them, tunes them and repairs them.

Beyond servers, this job has a second client that never goes away: everybody needs hot water, heating in winter and a pipe that does not leak. Every house is different (C1), breakdowns never repeat identically (C5), and a good plumber ends up recommended from person to person — that kind of trust cannot be downloaded (C4).

What does get automated: sizing, quotes, scheduling and remote monitoring — a sensor reports the pressure dropped before the client calls you. Diagnosis gets better. The repair stays manual.

03

Caring for people: nurse, carer, physiotherapist

C2 presence
C3 responsibility
C4 trust

Here the argument is not technical, it is human. Someone who is ill, frightened or old needs a person who is there: to lift them, change a dressing, listen to them, notice that today looks different from yesterday (C2).

It is also a job with responsibility — who administers, who writes it down, who calls the doctor (C3) — and one of trust: a family lets someone into their home and next to a person they love (C4). That need does not shrink as the world goes digital.

What does get automated: the paperwork, the charts, the scheduling, part of the monitoring and of preparing information for the doctor. Which is exactly the part that currently steals time from the patient.

04

Welder, fitter, maintenance technician

C1 physical world
C3 responsibility
C5 improvisation

The pipes, structures, halls and machines that hold up any industry — including the industry hosting AI — are welded, tightened, aligned and repaired by people. Often in awkward positions, on something old that no longer has a drawing or a manufacturer.

A good weld at height, in a forty centimetre gap, is not done remotely (C1). There are certifications, and a badly made joint has a name behind it (C3). And maintenance is, by definition, the situation that has never happened in exactly this way (C5).

What does get automated: series production. A robot welds the same part perfectly a thousand times, in a hall built around it. What does not get automated is the one-off repair, on site, on something old.

05

The person who teaches someone else: teacher, workshop instructor, trainer

C2 presence
C3 responsibility
C4 trust

Let us start with the uncomfortable part: the explanation itself — the theory, the examples, the exercises — is exactly what AI does well. We say it openly, otherwise there would be no reason to believe the rest.

What does not get automated is everything else, which is most of it: the person in front of the room who holds the attention of twenty-five children, spots the one who did not understand but nodded anyway, stops an argument and answers a parent (C2). Someone is legally responsible for minors inside a building (C3), and people learn from those they trust (C4).

What does get automated: lesson prep, materials, exercises, marking, the explanation given for the tenth time. What remains is the person in front of the room — and if you are mid-career, your experience so far is exactly what vocational training is missing today. You do not throw your trade away, you teach it.

But let us not stop here. Everybody needs…

Five jobs are not a complete list and should not be read as a recipe. They are simply the clearest examples of the criteria above. Look around with those in mind and you will see the real list is far longer, because it starts from needs that do not go digital:

water from the tap, in the morning
heat in the house, in winter
food on the table — someone grows it, moves it, cooks it
a roof that does not leak and a door that closes
a car that starts on Monday morning
power that comes back after it goes out
care, when you are ill or old
someone who arrives in time when it is an emergency
a person you can trust

These needs do not shrink as the world goes digital. They grow. And the jobs that cover them stay, even when their tools change. That is, in the end, the whole message of this page: AI does not take your place in the world. It changes your tools — and it leaves you room to choose.

05 What we do

Orientation, not promises

We are not a recruitment agency, we will not find you a job, and we will not sell you a course instead of an answer. We help you see clearly what you already have, what transfers, and which directions are realistic for you, in your situation.

You tell us where you are

Your age, what you have done so far, what you like and what you cannot stand, how much time you have and what is holding you back. Two or three lines are enough to start.
We look at what transfers

Almost nobody starts from zero. From what you have done so far there are things that move into a new trade: how you work with people, your knowledge of a field, your hands, your patience, a clear head under pressure.
We look at realistic directions together

Realistic means: in your town or within a distance you can actually travel, with the time and money you have now, with what the job truly requires — licences, practice and years of apprenticeship included.
We also tell you honestly what is not worth it

Including when the answer is “stay where you are and add one skill”. A full career change is a serious thing; it is not needed in every situation, and we will not recommend it just to have something to recommend.

The role of AI: sifting through options, comparing, preparing materials — a list of questions, a draft CV, a learning plan. The role of the human: the conversation and the decision. The decision stays yours; we do not take it from you and we do not automate it.

06 Limits

What we do not promise you

This list is more useful than a list of promises, especially in a field where a lot of fear gets sold. If any of the points below is a problem for you, better to find out now.

We do not guarantee you a job

We are not a placement agency and we have no way to promise anyone employment. We give you orientation, criteria and an honest conversation. The rest depends on you, on the market and on timing — and anyone promising otherwise is lying to you.

We do not give you numbers we cannot back up

No percentages of vanished jobs, no average salaries, no “demand is growing by so much”. This field is full of statistics recycled from old studies. If we do not have a source, we do not write the number.

We do not tell you your work was wasted

It was not. Not even if tomorrow that part gets done automatically ten thousand times faster. Experience is not erased along with the tool that produced it.

We do not speak badly of any trade

Not the desk ones, not the physical ones. There is no “lesser” job. There is only the fit between a person and the work, and that fit changes over time.

We do not sell you fear

“AI is stealing your job” is a good headline and bad advice. So is empty optimism, “just adapt and it will be fine”. We tell you what can be seen and what is not known.

We are not ready yet

The service is being built. For now we are collecting sign-ups: you write to us, we put you on the list and we let you know when we can talk properly. We are not asking you for money now and we are not promising an exact date.

07 New skills

If the direction you choose also needs new skills

It often turns out that you still need something you do not know yet — sometimes even the way AI is used as a tool inside the trade you are choosing.

The learning part, if needed, happens elsewhere

For applied AI skills there is learnai.ro, a school of applied AI with structured courses. We do not sell courses here: this page is where the direction is decided, and learning is the step after.

learnai.ro is the only external link on this site.

A course comes after the decision, not instead of it.
Not every direction needs an AI course. Many of the trades above ask for licences and practice, not an online course.
When we think you do not need it, we say so.

Visit learnai.ro

08 Contact

Write to us about where you are now

You do not need a CV and you do not have to know what you want yet. One sentence is enough: “I am 41, I worked fifteen years in X and now it is done automatically”, or “I am 17 and I have no idea what to choose”.

We put you on the list and reply by e-mail when we open the conversations. The service is being built, so we are not promising an exact date.
If what you write shows we can help, we will propose a conversation.
If we do not think we can help, we will tell you directly. That is more useful than a list of links.

Prefer ordinary e-mail? Write to us at contact@ai-cariera.ro.

Join the list

Fields marked with * are required. We do not ask for your phone number and we do not send a newsletter.

Full name

E-mail

This is where the answer goes. We do not use it for anything else.

Where are you starting from?

What you do now (optional)

Your trade, your job or your school. If you would rather not say, do not.

Where are you now and what worries you?

Two or three lines are enough. Write like to a person, not like at an interview.

The message goes straight to our e-mail. We only use it to reply to you. The data is processed by AIR BUSINESS SECURITY SOLUTIONS S.R.L., we keep the message for 12 months after the last exchange, and everything we do with it is written in Privacy.

AI is not your enemy. It sits on a very physical foundation — and people are the ones who build it.