Jobs being born right now, while everyone counts what is disappearing

Every report you read counts the jobs that will vanish. The opposite list is less dramatic and more useful, because you can act on it.

These are roles I see genuinely in demand, which did not exist in this form two years ago.

The roles, and what they actually mean#

Context engineer. Not "prompt engineer", a title that has already narrowed. The real role is broader: designing what the system sees before it answers. Which documents, in what order, chunked how, and what is excluded. This determines a system's quality more than the choice of model does.

Output evaluator. Someone who builds test sets and measures whether the system actually improved or merely seemed to. This has existed in labs for years and is now moving into ordinary companies.

AI governance lead. Someone who answers: what data goes in, who reviews, who is accountable. Acutely needed in regulated sectors.

Agent workflow designer. Someone who decides where machine work ends and human work begins, and designs the checkpoints.

Internal trainer. The role I see most in demand relative to supply in this region, and I speak from direct experience. Organisations buy tools then discover three people use them.

Arabic language data specialist. Someone who builds and evaluates Arabic data for training and testing. A genuine scarcity and a widening market.

The skill that unites them#

Notice the pattern: all of them require understanding a domain plus enough technical understanding, not the reverse.

Which is good news for anyone who thinks they are late. An accountant who understands models well is worth more than a model engineer who does not understand accounting, because the first knows what is worth building and the second only knows how to build it.

How to move into one#

None of this needs a new certificate. It needs evidence.

Build something small in your own field. Not a generic project. An accountant builds a tool that classifies invoices; a teacher builds one that generates assessment questions. A project a hiring manager understands is worth more than a technically impressive one they do not.

Document what failed. The thing that most distinguishes a candidate in this field is being able to say "I tried this and it did not work, for this reason". It proves you actually engaged with the problem.

Learn to measure before you learn to build. Someone who can prove their system improved with numbers gets ahead of someone who can build a prettier one.

The honest part#

Some of these titles will not survive. "Prompt engineer" appeared and nearly disappeared as a standalone title within two years, because the skill dissolved into other jobs.

Titles are temporary. The skill underneath is not: describing a business problem precisely enough to be solved automatically, and judging whether the solution works. That will not disappear, whatever the badge says.

In closing#

New jobs are not announced in the news. They appear in job adverts with confused titles, written by people who do not yet know what to call what they want.

Start today with one thing: open a job site and search "AI" within your own sector rather than technology generally. Read five complete adverts. You will learn more about the market than any report will tell you.

And if you want a structured route across, that is what the training programmes were designed for.

Common questions#

Which new roles has AI created?

Context engineer, output evaluator, governance lead, agent workflow designer, internal trainer, and Arabic language data specialist. All sit between the technology and the real work rather than in model building.

Do I need a technical qualification?

No. You need evidence: a small project in your own field solving a problem people there recognise, with a clear measure of success and documentation of what failed.

What is the most valuable combination of skills?

Understanding a specific domain plus enough technical understanding. An accountant who understands models is worth more than a model engineer who does not understand accounting, because the first knows what is worth building.

Will these job titles last?

Some will not. "Prompt engineer" appeared and nearly vanished as a standalone title within two years because the skill dissolved into other roles. Titles are temporary; the underlying skill is not.

How do I find out what my sector wants?

Search "AI" within your own sector rather than technology generally, and read five complete job adverts. Adverts with confused titles are usually the truest signal of genuinely new demand.