AI is the new Excel. History already ran this experiment.
If you’re scared about where AI is going, here’s the most useful thing we can tell you: your parents’ generation lived this exact movie. A machine arrived that did, in seconds, the knowledge work people built careers on. It killed hundreds of thousands of jobs - and created more, better ones, for the people who adapted. The machine was the spreadsheet. This essay is about what its era teaches about ours - with the real numbers from both.
Picture the accounting department of 1978. Rows of desks. On each one, a paper ledger - and a person whose actual job was arithmetic. Change one assumption in a financial plan and somebody spent the afternoon recalculating every dependent figure by hand, cell by cell, with a pencil and an eraser. Skilled, respectable, middle-class knowledge work.
Then, in 1979, a program called VisiCalc shipped for the Apple II - the first electronic spreadsheet. Change one number and every dependent cell recalculated instantly. Lotus 1-2-3 followed in 1983, Microsoft Excel in 1985, and within a few years the afternoon of pencil work was a keystroke. People in the profession said exactly what people say now: this machine does what I do, but faster, cheaper, and without lunch breaks. They were right. And here is precisely what happened next - the numbers, from US labour statistics, that make this the most instructive natural experiment we have:
~400,000
US bookkeeping & accounting-clerk jobs eliminated after 1980
The work of manually recalculating ledgers - hours of pencil-and-eraser arithmetic per change - collapsed into a keystroke. The people whose job WAS the recalculation lost that job.
~600,000
US accountant & analyst jobs ADDED in the same period
When modelling a scenario became free, businesses wanted a hundred scenarios. The tool destroyed the routine layer and expanded the judgement layer above it - a net gain, paying better, doing more interesting work.
Net effect of the “job-killing” machine: two hundred thousand morejobs in the field - better paid, doing analysis instead of arithmetic. The tool didn’t eliminate the profession. It eliminated the routine layer and expanded the judgement layer above it.
Why did jobs grow? Because of a pattern economists have documented across every automation wave: when doing something becomes dramatically cheaper, people want dramatically more of it. When one financial model took a week, a company ran one. When it took an hour, they wanted a hundred - every scenario, every quarter, every product line. The demand for analysis was never limited by appetite; it was limited by the cost of arithmetic. Kill the cost, unleash the appetite. (The same pattern repeated with ATMs, which everyone assumed would end bank tellers - teller employment rose for decades afterward as cheaper branches multiplied and the job shifted from counting cash to relationship work.)
But hold both truths, because the comfortable version of this story skips the second one: for the specific people whose job was the routine layer, the disruption was real.The ledger clerk of 1978 was not automatically the analyst of 1988. Some made that climb. Some didn’t. And the sorting mechanism between them was almost embarrassingly simple: some treated the new machine as a threat to wait out, and some sat down and learned it while it was still a differentiator.
2026: the same movie, wider screen
Now the present, with the same honesty. The disruption is not hypothetical - the 2026 data is blunt. Entry-level job postings are down roughly 15% year on year in the US and nearly 30% globally since early 2024, as employers shift routine work - exactly the tasks juniors used to learn on - onto AI. Goldman Sachs measured AI eliminating on the order of 16,000 net US jobs per monththis year, and in March 2026, for the first time, AI became the single most-cited reason in corporate layoff announcements. A Stanford analysis found employment for 22-25-year-olds in AI-exposed occupations fell 13% relative to their peers between 2022 and 2025 - the young feel this first, because the routine layer is where careers used to start. McKinsey’s long-standing estimate - up to 375 million workers worldwide needing to change occupational category by 2030 - no longer reads like a forecast. It reads like a schedule.
If that paragraph made your stomach drop: good. You’re paying attention, and fear that produces motion is worth more than comfort that produces waiting. But now put the spreadsheet frame over the same data and see what else is in it. The routine layer is collapsing - and the layer above it is expanding, on schedule. Employers report raising the bar for juniors, not closing the door: they want people who arrive already able to direct the tools. The premium for that fluency is measurable - in Malaysia, juniors who can demonstrably build with AI start at RM 6,000-9,000/month against the RM 3,500-6,500 standard, a gap our State of AI Hiring report documents in detail. New job categories that didn’t exist three years ago - agent builders, AI product engineers, evaluation specialists - are hiring faster than they can be filled. The experiment is running exactly as it ran in 1985. The only question the data can’t answer is the one your parents’ generation each answered individually: which side of the reallocation will you be on?
The three responses (and the only wrong one)
Everyone already knows AI is coming - knowing is not the bottleneck. What separates people is what the knowing produces. We’ve watched three responses across hundreds of students and thousands of conversations. Scared into motion: genuinely fine - fear is accurate here, and some of our best career-switchers arrived terrified; channelled, fear reads the risk honestly and acts early. Inspired into motion: also fine - the tools really are the biggest creative lever handed to individuals in decades; one person can now build what took teams, which is why our top-performing article is about solo founders shipping real AI products. Spectating:the only wrong answer - following the news, feeling vaguely anxious, doing nothing, waiting for clarity that will only arrive as hindsight. The spreadsheet-refusers weren’t fired in year one either. They were passed over in year three, and unremarkable by year ten - not because they were punished, but because the baseline moved and they didn’t.
Because that’s the quiet, load-bearing lesson of the Excel era: “proficient in Excel” stopped being a differentiator and became assumed - the way literacy is assumed. AI fluency is on the same conveyor belt, and you can already see it moving: employers screening for it, listings naming it, the premium being paid to early fluency. The window where adapting sets you apart - rather than merely keeping you eligible - is open now, and its whole nature is that it closes quietly.
What “adapting” actually means (three rungs)
Rung one - use it until you have judgement. Not novelty use; daily use on your real work, until you know with calibrated confidence what AI does brilliantly and where it fails while sounding certain. This alone puts you ahead of most colleagues, and it’s free.
Rung two - move up your own value chain.Audit your role the way the spreadsheet audited accounting: which of your tasks are the routine layer, and which are the judgement layer? Deliberately migrate your time and reputation to the second, because that’s where your role is going anyway - our guides on what stays human and future-proofing a career map this rung in detail, role by role.
Rung three - learn to build. The accountant who used Excel did well. The one who could build the modelseveryone else ran became indispensable. The AI-era equivalent - being able to build software and automations with AI, not just chat with it - is the highest rung, and here’s the part that would have sounded impossible in 1985: it now takes months, not a career change into the unknown.One of our graduates, Daniel, put it best in his (public, verifiable) review: “I got a Software Developer job before I ended the bootcamp... and no, I had no diploma or degree to begin with.” AI compresses the path - not the person walking it. That’s not a slogan to us; it’s the design principle behind everything we teach, and the reason we insist on fundamentals before AI, never instead of it.
Your parents adapted to their machine - most of them, in the end, at kitchen tables with manuals and evening courses, without anyone framing it as a heroic act. It became a Tuesday. This will too. The only decision that’s actually yours is whether you adapt while it’s an advantage or after it’s an expectation - and if you want to feel the difference this week instead of theorising about it, the free trial exists so the first step costs nothing but an evening.
FAQ
How is AI like Excel?
The parallel is structural. Spreadsheets automated the routine layer of knowledge work (manual recalculation) and expanded the judgement layer above it: after 1980 the US lost roughly 400,000 bookkeeping-clerk jobs but gained about 600,000 accountant and analyst jobs, because when modelling became cheap, demand for analysis exploded. AI is running the same experiment on a wider set of tasks: the routine layer (drafting, summarising, boilerplate code, basic analysis) is collapsing in cost, while the judgement layer - deciding, verifying, directing - is where the human value concentrates. And as with Excel, fluency is shifting from advantage to baseline: within a decade, "I use AI well" will be as unremarkable on a CV as "proficient in Excel" is today.
Will AI take my job?
The honest, three-part answer: AI will take over tasks inside almost every job; it will eliminate some jobs whose core IS those routine tasks (as spreadsheets did to ledger clerks); and it will reshape most others upward, concentrating value in judgement, relationships, and direction of the tools. The 2026 data shows real disruption - entry-level postings down sharply, AI now the most-cited reason in some months' layoff announcements - and real reallocation, with demand growing for people who can build and direct AI. Which side of the reallocation you land on is substantially a choice. We've written honest role-by-role answers for 17 professions if you want yours specifically.
What happened to people who refused to learn spreadsheets?
Nothing dramatic on day one - that's what makes the lesson easy to miss. The spreadsheet-refusers of the 1980s weren't fired in a wave; they were passed over gradually. Promotions went to the analyst who could model five scenarios by lunch. Job listings quietly added "spreadsheet proficiency required." Within roughly a decade, the skill stopped appearing in listings at all - not because it stopped mattering, but because it had become assumed, like literacy. The AI version of that clock is already running: employers increasingly screen for AI fluency, and the window where it distinguishes you (and commands a premium) is exactly the window worth acting in.
How do I actually adapt to AI - concretely?
Three moves, in order. First, use AI daily on your real work until you develop calibrated judgement - knowing what it does brilliantly and where it confidently fails (most people stop at novelty and never build this). Second, move your own role up the value chain: spend the hours AI frees on the judgement work - deciding, verifying, client-facing, cross-domain - that AI can't absorb. Third, and highest-leverage: learn to BUILD with AI, not just use it - the difference between the accountant who used Excel and the one who built the models everyone else used. That third move sounds technical but now takes months, not years, and it's the single biggest arbitrage of this transition.
Is it too late to adapt to AI - or too early to bother?
Neither, and both errors are common. Too late? No: most companies are still early in real adoption, the tools are only a few years old, and the fluency premium is measurably still being paid - in Malaysia, AI-capable juniors earn roughly RM 6,000-9,000/month against a RM 3,500-6,500 standard. Too early to bother? Also no: the entry-level squeeze is already in the data, and the Excel lesson is that baselines shift quietly and then all at once. The accurate framing: you are in the window where adapting is still a differentiator. That window is the opportunity - and it is not permanent.
Adapt while it’s an advantage. Not after it’s an expectation.
The window where AI fluency sets you apart is open now. The free trial - one signup, no card - puts you on rung one this evening, building real things on Sigmo with a live instructor session included. Your parents did this at a kitchen table with a thicker manual.
