What will jobs look like in 10 years? We’ve seen this movie before.
Most answers to this question are either sci-fi (nobody works, robots everywhere) or denial (nothing really changes). This one is built differently: five forecasts, each anchored to a trend already visible in hiring data or already completed by the last technology that did this to work - the spreadsheet.Forecasting from precedent isn’t exciting. It’s just more likely to be right.
The method: forecast from the last time this happened
Between roughly 1985 and 2005, a piece of software restructured white-collar work end to end. The spreadsheet eliminated about 400,000 US bookkeeping-clerk jobs - an entire routine layer - while accounting and analysis employment grew by roughly 600,000, because cheap calculation made judgement work more valuable, not less. Everyone in finance had to adapt; the ones who learned early spent a decade being the most employable people in their field; the title “accountant” survived while its daily contents were replaced almost entirely. We’ve told that story in full in AI is the new Excel, and it matters here for one reason: AI is running the same movie, faster, across every industry at once- and 2026’s data (routine postings down double digits, AI-capable juniors earning RM 6,000–9,000 against a traditional RM 3,500–6,500) shows we’re already well past the opening scene. So instead of imagining 2036, we can extrapolate it. Five forecasts, each with its anchor stated so you can check the reasoning.
Five forecasts for 2036
01
Most professionals will direct AI the way they use spreadsheets today - unremarkably.
Anchor: nobody lists "Excel" as a superpower anymore; it collapsed from competitive edge (1990s) to assumed baseline (2010s) in under two decades. AI literacy is tracing the same curve faster - job postings mentioning AI tools have already exploded across marketing, finance, and operations, not just tech. By 2036, "can direct AI" will appear on job ads about as often as "can use a computer" does now: never, because it is assumed.
02
The routine layer of white-collar work will be mostly gone - and its people will have moved, not vanished.
Anchor: the spreadsheet eliminated ~400,000 US bookkeeping clerk jobs while accounting employment GREW by ~600,000, because the routine layer died and the judgement layer expanded. 2026 is mid-repeat: entry-level routine postings down 15% US / 29% globally, while AI-capable roles carry documented premiums. Ten years on, "process this document / draft this report / reconcile this data" as a full-time human job description will read as anachronistic as "typist."
03
Judgement, accountability, and trust become the human job description.
Anchor: what survived every previous automation wave was never the task - it was the accountability for the task. The accountant survived the spreadsheet because someone must sign the audit; the doctor will survive diagnostic AI because someone must own the treatment decision. Work concentrates where a human must evaluate AI output, carry responsibility for it, and be trusted by other humans - which is why "can you judge what the machine produced" is already the interview question that matters in 2026.
04
Small teams and solo builders will do what departments did - and new job titles will keep appearing.
Anchor: solo founders already run six- and seven-figure software products with AI doing the heavy lifting - unthinkable in 2020, documented today. The leverage-per-person curve only steepens. Expect the 2036 economy to hold many more tiny companies, many more hybrid roles ("nurse + health-tech builder"), and titles that do not exist yet - the way "web designer" did not exist in 1985. History has never once produced a net-fewer-job-titles decade.
05
The gap between AI-fluent and AI-avoidant workers becomes the defining economic divide.
Anchor: it is already visible in one payroll line - RM 6,000-9,000/month for AI-capable Malaysian juniors vs RM 3,500-6,500 for traditional profiles. Stanford data shows employment for 22-25-year-olds in AI-exposed roles down 13% while AI-fluent peers command premiums. Divides like this compound: the fluent get the roles that build more fluency. In 10 years this is less a skills gap than a class line - and which side you are on is still, today, a choice.
What this means for you, at 25 or 35, in 2026
Read the five forecasts together and they compress into one sentence: the next decade automates tasks and promotes judgement - and it sorts people by when they started adapting. That last clause is the uncomfortable one, because forecast #5 compounds: the AI-fluent get the interesting roles, which build more fluency, which gets the next role. The clerk who learned Excel in 1988 and the one who refused until 1998 ended the decade in different economic classes despite starting in the same chair. Nothing about the fork requires you to work in tech - forecast #1 is about every profession - but everything about it rewards moving early.
And the preparation is unusually concrete, because the surviving layers are known: fluency (use AI daily on real work), fundamentals (you can only judge AI output in a domain you genuinely understand - the full argument for why shallow prompting fails is in vibe coding vs fundamentals), and shipping (leverage keeps flowing to people who build real things - the stage-by-stage path is in how to become an AI-native developer). Ten years is long enough that anyone can cross the divide from either side - and short enough that waiting five of them out is the one genuinely dangerous plan. The zero-cost way to start is the free trial - one signup, no card, real projects and a live instructor session - and if you want the fear itself examined rather than managed, scared AI will take my job is written for exactly that reader.
FAQ
What will jobs look like in 10 years because of AI?
The grounded forecast, anchored to visible trends rather than sci-fi: (1) directing AI becomes an unremarkable baseline skill, the way Excel did; (2) the routine white-collar layer - process, draft, reconcile - is mostly automated, with its people redistributed the way clerks became analysts; (3) human work concentrates in judgement, accountability, and trust: evaluating AI output and owning the consequences; (4) small teams and solo builders do what departments did, and new job titles keep appearing; (5) the AI-fluent vs AI-avoidant gap becomes the defining economic divide. The spreadsheet era ran this exact movie between 1985 and 2005; AI is running it faster and across more industries.
Will my job still exist in 10 years?
Reframe it, because "job" bundles two things with opposite fates. Your job’s routine layer - the predictable, repeatable portion - probably won’t exist in anything like its current form; that’s the layer every automation wave takes. Your job’s judgement layer - decisions, accountability, relationships, evaluating whether output is right - is what the role compresses into, and usually expands. The accountant of 2005 kept the title but did almost none of what the accountant of 1985 did daily. So the real question isn’t whether your title survives - it’s whether you’re positioned in the layer that does. For a role-by-role breakdown, our will-ai-replace hub covers specific professions.
Which skills will still matter in 10 years?
Bet on the layers that survived every previous wave: (1) judgement - evaluating whether output (AI’s or anyone’s) is actually correct, which requires genuine fundamentals in your domain; (2) building - the ability to make real things work end-to-end, because leverage keeps flowing to people who ship; (3) accountability and trust - being the person others rely on to own outcomes; (4) learning speed itself - the decade will invent tools nobody can pre-train you on. Note what’s NOT on the list: memorising syntax, tool-specific tricks, or any skill defined by a current product’s interface. Fundamentals age in decades; tools age in quarters.
Should I be scared about the future of work?
Concerned enough to act, not scared enough to freeze - and the difference matters because the two produce opposite behaviour. The genuinely scary version of the next decade belongs to people who wait it out: the divide forecast in this article compounds, and late entry costs more every year. The exciting version belongs to people who move early: entry premiums (RM 6,000-9,000 for AI-capable juniors), leverage no previous generation of builders had, and new categories opening faster than they can be filled. Our parents faced the same fork with the spreadsheet; the ones who learned it don’t describe the 1990s as scary. The fear is real; its correct output is a learning plan.
How do I prepare for the next 10 years of work?
Three moves, in order. First, get AI-fluent now, on real work - daily use, not tutorials - because forecast #1 says this becomes the assumed baseline and late adopters pay the compounding gap in forecast #5. Second, build fundamentals in a domain: judgement (forecast #3) is only possible when you understand things deeply enough to evaluate AI’s output - shallow prompting inherits none of the surviving layer. Third, learn to ship real things end-to-end, because leverage keeps concentrating in builders (forecast #4). That trio - fluency, fundamentals, shipping - is deliberately the entire structure of our programme, and the free trial (one signup, no card) is the zero-cost first step into all three.
The divide is forming. Pick your side early. Fluency, fundamentals, shipping - starting tonight, free.
Every forecast in this article rewards the same three skills, and all three are learnable in months. The free trial starts you on real projects with a live instructor - one signup, no card.
