What jobs will AI actually create?

Seven categories are already visible in hiring data rather than speculation: AI-native developers (the new shape of the software role - the biggest category), AI workflow specialists inside ordinary companies, AI evaluators and model trainers (the gig-shaped entry layer), AI product managers who own AI quality in production, one-person AI company founders (a job invention, not a posting), AI trainers and educators, and - the largest hidden category - hybrids: existing professions with AI capability attached, where domain-plus-AI beats either alone. History says this list will keep growing: nobody in 1980 predicted "web designer," and the spreadsheet era created more accounting jobs than it destroyed.

Will AI create more jobs than it destroys?

The honest answer: nobody can promise the ratio, but history’s pattern and 2026’s data both point the same direction - transformation with net creation over time, concentrated pain in routine roles during the transition. The spreadsheet destroyed ~400,000 US clerk jobs and created ~600,000 better ones; ATMs preceded decades of teller growth. Today, entry-level routine postings are down sharply while AI-capable roles command documented premiums and can’t be filled fast enough. The catch that matters for YOUR planning: the new jobs don’t automatically go to the people who lost the old ones - they go to whoever prepared. The ratio is an economics debate; your position in it is a choice.

Which AI-created job is easiest to get into?

Depends on your starting point, so match honestly. Fastest income, lowest barrier: AI evaluation gigs (remote platforms hire now; treat as a bridge, not a destination). Best if you like your current field: the hybrid path - add building skills to your existing domain, often reachable through your current employer. Best long-term economics: AI-native developer - months of structured learning to a documented RM 6,000-9,000 starting band and the industry’s steepest curve. Most freedom: the one-person company route, if you can pair building skills with a validated niche problem. All four share one prerequisite: demonstrable ability over certificates.

Do the new AI jobs require a degree?

Mostly no - and this is one of the era’s genuinely good pieces of news. The categories above are so new that no degree pipeline exists for them, which forced employers into skills-first hiring: portfolios, demonstrated ability, work you can defend in an interview. Several Sigmaschool graduates hold developer roles with no degree at all, hired on exactly that basis. The exceptions are hybrids in credential-locked fields (health-tech clinical roles still need the clinical credential). For everyone else, the gate is capability - which is buildable in months - rather than paper, which takes years.

How do I prepare for jobs that don’t exist yet?

Build the layer all of them share. Every category on this list - and every one that will join it - sits on the same foundation: fluency in directing AI, judgement to evaluate its output, and enough building capability to ship something real. Those three transfer across every title the next decade invents, the way spreadsheet-fluency transferred across every finance role the 1990s created. Concretely: use AI daily on real work now, learn fundamentals with AI as tutor, and build a small portfolio of real things. The specific job titles will keep changing; the entrance exam won’t.

All posts

Careers · AI & your future · Opportunity

What jobs will AI create? Check the postings - it already did.

The internet has a thousand articles on what AI destroys and almost none mapping what it builds - because fear outperforms opportunity in every feed. Here’s the correction: seven job categories that barely existed three years ago, hiring right now, with real pay bands and honest entry paths. Not futurism - postings.

Deric Yee Updated 25 August 2026 9 min read

Why creation always lags the headlines

A structural truth about every automation wave: the destruction is legible immediately (a closed role has a name and a face) while the creation arrives unnamed - nobody in 1980 predicted “web designer,” “SEO specialist,” or “app developer,” yet those categories eventually dwarfed what the computer displaced. The spreadsheet era we’ve documented in AI is the new Excel ran the same asymmetry: 400,000 clerk jobs visibly lost, 600,000 better jobs quietly gained. So when your feed shows only the destruction side of 2026, it’s not lying - it’s just early, and biased toward what already has a name. The list below is what has acquired names so far. It will be longer next year.

Seven categories, already hiring

AI-native developer / AI engineer

Builds products powered by AI - agents, copilots, automation systems. The biggest and best-documented category: not a niche title but the new shape of the software developer itself.

The way in: The clearest path in: fundamentals + AI-product skills, months not years. Juniors start RM 6,000-9,000/month in Malaysia - the premium exists because supply trails demand.

AI workflow / automation specialist

The person inside a normal company who rebuilds its processes around AI - the operations, finance, or marketing professional who can automate what their department does. Every mid-size firm is quietly creating this role.

The way in: Domain knowledge you already have + light building skills. Often reached by promotion rather than application: automate your own team’s pain and the title follows.

AI evaluator / model trainer

Reviews and grades AI output to make models better - from general annotation up to specialist evaluation in law, medicine, finance, and code. The gig-shaped entry layer of the whole industry.

The way in: Lowest barrier on this list: platforms like Alignerr and OpenTrain AI hire remotely, today. Honest ceiling though - best used as a bridge that funds the climb to categories 1-2.

AI product manager / AI ops

Decides what AI features are worth building, owns their quality and cost in production, manages evaluation pipelines. Emerged because shipping AI is easy and shipping *good* AI is not.

The way in: Usually reached from adjacent roles (PM, ops, dev) plus demonstrated AI-product literacy - the person who understands both the business and what the models actually do.

One-person AI company founder

Not a job posting - a job invention: solo builders shipping real SaaS products with AI doing the heavy lifting. Documented solo founders now run six- and seven-figure products alone.

The way in: The most democratic category: no interview, no CV filter. Validate a niche problem, build with AI, charge. The barrier that used to gate this - needing an engineering team - is the thing AI removed.

AI trainer / educator

Teaches AI fluency to everyone else - kids, teams, executives. Demand grew in lockstep with every company’s and parent’s AI anxiety, and supply of people who can genuinely teach it lags badly.

The way in: Strong communicators with intermediate technical skills fit fast; former teachers hold a real edge. Ranges from tuition-centre work to corporate L&D.

The hybrid: [your field] + AI

The largest category hides in plain sight: not new titles but existing professions with AI capability attached - the accountant who builds reconciliation tools, the marketer who ships automations, the clinician in health-tech. Domain + AI beats either alone in every industry.

The way in: You’re halfway in already - the domain half is done. The AI half takes months, and it converts your existing career from AI’s target into AI’s operator.

The common thread (and the uncomfortable catch)

Look across all seven and one foundation repeats: every category runs on directing AI, judging its output, and shipping something real - the exact profile we map stage-by-stage in how to become an AI-native developer and whose interview-tested skills we itemise in the seven AI skills employers want. That’s the era’s genuinely exciting property: one learnable foundation opens seven doors at once, and whichever new titles 2028 invents will almost certainly open with the same key.

Now the catch, stated honestly because false comfort helps nobody: the new jobs don’t automatically go to the people the old jobs left.That’s the transition’s real cruelty - the displaced clerk of 1985 wasn’t handed an analyst seat; the seat went to whoever had learned the spreadsheet. The same sorting is running now: the categories above are being filled from the pool of people who prepared, while the displaced who didn’t prepare compete for the shrinking routine layer. The entire practical meaning of this article is to put you in the first pool while the premium for early movers still holds - the RM 6,000-9,000/month starting band for AI-capable juniors in Malaysia (the data) is what “early” currently pays.

And the first door costs nothing to try: the free trial - one signup, no card - puts you on the shared foundation tonight with real projects and a live instructor session. Category 3 can even fund the journey (the honest gig guide), category 7 means your current career is an asset rather than a liability, and category 5 - the one-person company - is documented, with real founders and real revenue, in our most-read article. The destruction got the headlines. The creation is taking applications.

FAQ

What jobs will AI actually create?
Seven categories are already visible in hiring data rather than speculation: AI-native developers (the new shape of the software role - the biggest category), AI workflow specialists inside ordinary companies, AI evaluators and model trainers (the gig-shaped entry layer), AI product managers who own AI quality in production, one-person AI company founders (a job invention, not a posting), AI trainers and educators, and - the largest hidden category - hybrids: existing professions with AI capability attached, where domain-plus-AI beats either alone. History says this list will keep growing: nobody in 1980 predicted "web designer," and the spreadsheet era created more accounting jobs than it destroyed.
Will AI create more jobs than it destroys?
The honest answer: nobody can promise the ratio, but history’s pattern and 2026’s data both point the same direction - transformation with net creation over time, concentrated pain in routine roles during the transition. The spreadsheet destroyed ~400,000 US clerk jobs and created ~600,000 better ones; ATMs preceded decades of teller growth. Today, entry-level routine postings are down sharply while AI-capable roles command documented premiums and can’t be filled fast enough. The catch that matters for YOUR planning: the new jobs don’t automatically go to the people who lost the old ones - they go to whoever prepared. The ratio is an economics debate; your position in it is a choice.
Which AI-created job is easiest to get into?
Depends on your starting point, so match honestly. Fastest income, lowest barrier: AI evaluation gigs (remote platforms hire now; treat as a bridge, not a destination). Best if you like your current field: the hybrid path - add building skills to your existing domain, often reachable through your current employer. Best long-term economics: AI-native developer - months of structured learning to a documented RM 6,000-9,000 starting band and the industry’s steepest curve. Most freedom: the one-person company route, if you can pair building skills with a validated niche problem. All four share one prerequisite: demonstrable ability over certificates.
Do the new AI jobs require a degree?
Mostly no - and this is one of the era’s genuinely good pieces of news. The categories above are so new that no degree pipeline exists for them, which forced employers into skills-first hiring: portfolios, demonstrated ability, work you can defend in an interview. Several Sigmaschool graduates hold developer roles with no degree at all, hired on exactly that basis. The exceptions are hybrids in credential-locked fields (health-tech clinical roles still need the clinical credential). For everyone else, the gate is capability - which is buildable in months - rather than paper, which takes years.
How do I prepare for jobs that don’t exist yet?
Build the layer all of them share. Every category on this list - and every one that will join it - sits on the same foundation: fluency in directing AI, judgement to evaluate its output, and enough building capability to ship something real. Those three transfer across every title the next decade invents, the way spreadsheet-fluency transferred across every finance role the 1990s created. Concretely: use AI daily on real work now, learn fundamentals with AI as tutor, and build a small portfolio of real things. The specific job titles will keep changing; the entrance exam won’t.

Back to all posts The other side: what AI replaces →

Seven doors. One key. Free first step.
The creation side is taking applications.

Every category on this list runs on the same learnable foundation - directing AI, judging output, shipping real things. The free trial starts building it tonight: one signup, no card, live instructor session included.

Start the free trial The full AI-native path