There's a book from 1998 called "Who Moved My Cheese?" It features mice navigating a maze when their cheese supply disappears. At 94 pages, it's basically a children's book, and it sold 28 million copies anyway, because it touched something real: change arrives whether you're ready or not, and how you respond is the only part you control.

AI is the cheese being moved right now. Different maze, same instinct to stand in the empty spot where the cheese used to be and wait for someone to put it back.

The World Economic Forum's 2025 Future of Jobs Report projects that by 2030, 92 million roles will be displaced by AI and automation. That number gets repeated constantly. What gets less attention is that the same report projects 170 million new roles will be created, a net gain of 78 million jobs globally. Neither number tells the whole story, but together they tell a better one than most headlines do.

What "displaced" actually means

McKinsey's research puts it plainly: today's AI could theoretically automate about 57 percent of US work hours. That sounds alarming until you read what it actually measures. It's a measure of how much of what people do each day could be done differently, not a projection that 57 percent of jobs disappear. Most roles will change shape. Far fewer will vanish.

Say a bookkeeper spends twelve hours a week on data entry and reconciliation, and another eight on catching errors, chasing down mismatched invoices, and explaining the numbers to an owner who doesn't want to read a spreadsheet. AI can compress the twelve hours into two. It can't do the other eight, because that work depends on knowing which numbers look wrong before you check them and knowing how to explain a bad month to someone who's already stressed about it. The job doesn't vanish. It gets smaller in one place and more concentrated in another.

Same pattern, different job. Say a construction estimator builds thirty quotes a year, and roughly half the hours go into pulling numbers from supplier catalogs, past jobs, and standard assemblies before any actual pricing judgment happens. AI can do the pulling. It can't decide that a particular client always pads their timeline, or that a certain subdivision's soil conditions mean the excavation line item needs a buffer the spec sheet doesn't mention. That knowledge lives in the estimator's head, built up over years of being wrong and adjusting. No model has that yet, and it's not obvious one ever will, because that knowledge isn't data so much as judgment built from years of being wrong and correcting for it.

That's the pattern showing up across most office and administrative roles, in the trades and everywhere else. The mechanical, repeatable part shrinks. The part that requires judgment, context, or a relationship with another human doesn't, because AI still can't do that part.

Who's actually exposed

Real hiring data reflects this. Research tracking job postings from 2019 through early 2025 found postings for routine, repetitive roles dropped 13 percent. Postings requiring analytical, creative, or judgment-intensive work grew 20 percent. AI is changing what employers value, not simply shrinking the headcount.

The roles most exposed share a few traits: the work is largely digital, the inputs are structured and predictable, and success can be checked against a rule. Data entry, basic scheduling, first-draft writing, simple bookkeeping. The roles least exposed involve physical presence, ambiguous situations, or a client who needs to trust the person in front of them, not just the answer. A framer reading a tricky roofline. A service tech diagnosing a problem that doesn't match the manual. A project manager talking a nervous client through a delay. None of that shows up in a training dataset in a way that lets a model do it reliably.

Here in BC, the picture lines up. Statistics Canada data on why BC businesses are adopting AI shows 51 percent cite increasing task automation without reducing employment as a primary goal. Only 21 percent list replacing employees as a reason at all. Most small and mid-sized employers aren't running a layoff plan disguised as a technology rollout. They're trying to get a proposal out the door faster or stop losing a week to paperwork every time someone's on vacation.

The cheese has moved, but most of it is still in the maze.

"But this time feels different"

It's a fair worry, and it's worth taking seriously instead of waving off. Every major shift in how work gets done has produced this same fear, and workers who lived through those shifts weren't wrong to be uneasy. The printing press put scribes out of work. Mechanized farming emptied rural towns that had run on manual labour for generations. Manufacturing went through something similar decades ago when robotics and automated machinery showed up on the line. Some jobs on the floor did disappear. Others multiplied around programming, maintaining, and troubleshooting the equipment that replaced them, and those roles paid better than the ones they replaced.

What makes this shift feel different is speed and reach. Robotics changed manufacturing over a couple of decades and mostly stayed inside manufacturing. AI is moving into office work, customer service, and skilled trades administration inside a few years, all at once. The discomfort is legitimate. It doesn't mean the outcome is guaranteed to be worse, but it does mean there's less time to adjust than past generations had, which is exactly why waiting to see what happens is the riskier move, not the safer one.

There's an older, smaller example worth remembering too. When ATMs spread through North American banks decades ago, the common assumption was that bank tellers were finished. What actually happened is that ATMs made it cheaper for banks to open branches, banks opened more branches, and total teller employment held up for years, even as the job itself shifted away from counting cash toward sales and customer service. The lesson isn't that automation always works out fine, but that guessing the outcome from the headline alone gets you the wrong answer more often than not.

What actually helps

IBM's 2026 CEO study found that 41 percent of workforces had already been reskilled to perform their current roles more effectively with AI. The same study found 61 percent of employees say AI makes their work less routine and more strategic. The workers at greatest risk are the ones who decide AI has nothing to do with them.

A few things actually move the needle, and none of them require becoming a programmer:

None of that is complicated. Most of it is uncomfortable, because it means admitting the ground is moving before you're forced to.

What about apprentices and entry-level workers?

This is the version of the question I get most from parents and from young people themselves, and it deserves a straight answer instead of reassurance for its own sake. Entry-level office and administrative jobs, the ones that used to be a first step into a career, are genuinely the most exposed category. Data entry, basic scheduling, first-draft document work: those roles are shrinking, and pretending otherwise doesn't help anyone plan.

The trades sit in a different spot. An apprenticeship is supervised hours doing physical work under a certified journeyperson, not a desk job that can be replaced by software, and that's a structure AI has no path into. That doesn't mean apprentices should ignore the technology. The ones who learn early how AI touches the business side of their trade, quoting, scheduling, documentation, will move into running crews and eventually businesses faster than the ones who treat it as someone else's problem. Being early isn't a guarantee, but it's still the better bet.

What employers can do about it

Most of the advice about AI and jobs is aimed at individual workers. Business owners carry a separate responsibility, especially in a tight labour market where a good employee is hard to replace and expensive to lose.

Tell people what you're testing before they hear about it secondhand. Silence from ownership is what turns a reasonable new tool into a rumor about layoffs. Involve the people whose work is actually changing when you pick a tool, because the person doing the job every day usually knows faster than anyone in the office whether something will actually save time. And be honest about the difference between automating a task and eliminating a position. Those are different decisions with different consequences, and conflating them, even by accident, is how trust breaks.

If your job is mostly physical, you're not off the hook either

This one matters for trades and construction specifically, since it's the question I hear most from people in the field. The hands-on part of the work, the actual installing, framing, welding, and troubleshooting, is genuinely safe for a long while yet. Nobody's building a model that can run conduit through a finished wall.

The business around that work is a different story. Every trade sits inside a company that also estimates, proposes, schedules, invoices, and documents. That side of the business looks a lot more like the bookkeeper example above than like the job site, and it's exposed to the same shift. A tradesperson who assumes their whole career is safe because the physical work is safe is only looking at half the picture. The office half of a construction or trades business is exactly where the WEF's 92 million and 170 million numbers are going to show up first.

Manufacturing shops face a version of the same split. The people running machines, doing quality checks by hand, and troubleshooting a line that's jammed aren't going anywhere soon. The scheduling, the purchasing, the documentation trail that has to exist for every batch, the paperwork that follows a product from raw material to shipped order: that stack of work is where a manufacturer's exposure actually sits, and it's usually invisible to the people on the floor because it happens in an office they rarely walk into.

A few questions worth asking about your own role

Not a quiz, just a way to get past the vague dread and into something specific:

Most people can answer the first two without much trouble. It's the third and fourth that tend to go unasked, usually because the answer is uncomfortable either way.

In "Who Moved My Cheese?", the mice who went looking found new cheese. The ones who waited in the empty room did not, and the book doesn't pretend that waiting was a comfortable choice at the time. Fear kept them there long after the cheese was gone. It just wasn't a useful fear. The transition hurts either way. Moving through it still beats standing still in it.

If you're a business owner trying to figure out where AI fits into your operation and your team, or an employee wondering how to stay relevant, that's a conversation worth having before someone else decides it for you. The free AI Readiness Assessment is a reasonable place to start that conversation on your own terms.