Every time the conversation turns to AI and jobs, workers and business owners face a choice.
You can take the blue pill: believe your industry is “AI-proof,” assume your job will remain the same and wait for someone else to tell you when the future has arrived.
Or you can take the red pill and look directly at what is changing.
The red-pill reality is not that AI will take every job tomorrow. It is that AI is already changing the work inside the job.
Over a long enough period, I think the honest answer for many people is probably yes: jobs we recognize today will disappear or be redesigned completely.
The choice is not between panic and denial. It is whether you learn to work with the system before the system changes around you.
Glitches in the Matrix: the job market has not crashed
If you look at the Canadian data, the system has not crashed.
Statistics Canada examined employment through December 2025 and found no clear persistent decline across categories of work with different levels of AI exposure. Coding-intensive employment grew at a similar rate to employment overall.
Look closer, though, and there are glitches worth watching. Coding employment among workers under 30 stagnated while gains were concentrated among workers aged 30 to 49. Statistics Canada does not blame AI alone; demographic shifts, post-pandemic adjustments and other economic conditions may also be involved.
We cannot honestly say AI has caused a Canadian employment collapse. Its first effects may appear as fewer junior openings, higher output expectations and changing tasks long before they appear as the disappearance of an occupation. That is one reason leaders need to confront the institutional effects of AI, not merely count licences and pilots.
In March 2026, 35.9% of Canadian workers reported using generative AI at work. Nearly two-thirds of those users applied it to some, but not most, of their tasks. The job title may look familiar while the work inside it quietly changes.
The code of work can change in months
I have experienced how quickly the economics can move.
Salient AI started developing software to monitor Google rankings. Building the useful application conventionally could have cost roughly $10,000 to $20,000. Using AI, we found that we could develop what we needed in a couple of days.
That is my assessment of one project, not a universal software price. Production applications still require security, testing, maintenance and ownership.
But imagine spending $10,000 or $20,000 on software and discovering months later that a capable person using newer AI can reproduce much of its useful function in days. Nobody necessarily made a bad decision. The capability changed faster than the purchasing cycle. This is why organizations should examine the actual friction before deciding whether to buy more software or build a focused AI-assisted workflow.
I saw the same gap during a demonstration involving programmers at a college. They had spent weeks working on login authentication and a Stripe connection. In about two hours, I produced a version zero of the application that did almost everything needed for the demonstration.
That does not mean every production system can be built in two hours. Login and payments require serious review. It does show how well AI can handle software connections and API bridges. If it lacks context, you can often give it the manual and work through the problem with it.
A year earlier, I would not have expected the same result. Judging tomorrow's AI by what it cannot do today is becoming a dangerous career strategy.
You cannot simply unplug from the system
Some people respond by choosing work that appears less dependent on AI.
That may be a sensible short-term decision. I do not think it is a permanent escape route.
Robotics can carry AI from the screen into physical work. Painting, cleaning, laundry and gardening still present difficult problems involving movement, safety, reliability and cost. General-purpose robots cannot yet do all of that work dependably.
But “not today” does not mean “never.” Programming itself is the warning. Work that appeared difficult for AI a year ago can look very different now.
Nor can Canada simply unplug while everyone else continues. We need sensible rules and responsible AI boundaries, but development is international. If one country stops while others advance, it risks falling behind economically and scientifically. Governance and adoption have to move together.
Training in the construct: do not remove the first rung
The people most exposed to this early transition may be the ones still trying to enter the workforce.
Junior employees traditionally learn by doing routine work: preparing first drafts, reconciling spreadsheets, handling common questions and building basic software components. AI is very good at absorbing those tasks.
That creates an employer's dilemma. If one experienced employee using AI can do work that once required several people, does the company still hire the junior employee?
Used well, AI can help new employees learn faster. A study involving 5,179 customer-support agents found the largest productivity gains among less-experienced and lower-skilled workers. AI can become part of the training ground.
Used only to reduce payroll, it can remove that training ground. If every business automates the bottom rung of the ladder, where will the next generation of experienced employees come from? The same leadership question applies when a company is building an AI-first team from the beginning.
Employers should train their existing people before defaulting to replacement. Give them suitable tools, real workflows and enough capacity to become more capable. A limited experiment is useful, but a basic plan is not the same as serious AI integration.
We should also be honest. Few owners will sacrifice their livelihood or allow a company to fail simply to preserve every job unchanged. They may respect an employee, but when the choice becomes the survival of the business or one existing role, the business will usually come first. An employee who becomes an employer would likely face the same decision.
Responsible leadership is not a promise that no job will disappear. It means preparing people early, explaining what is changing, giving them a fair opportunity to adapt and preserving human judgement where safety, privacy, empathy or serious consequences matter.
Free your mind—and start practising
You do not need to be a technical expert to begin. You need to practise.
Choose one real task you understand well. Explain the result you need, provide the context, inspect what AI produces and correct it. Repeat the process until you learn where it helps, where it fails and where your judgement adds value.
One of the simplest changes is to stop relying exclusively on the keyboard and start talking to AI. Speaking makes it easier to provide context, explore uncertainty and turn the interaction into a working conversation instead of a search query.
For a business owner, the first move is equally practical. Choose one workflow with visible friction. Train the employee who understands it. Then measure:
- What took less time?
- What new work became possible?
- What still required human judgement?
- How should the employee's role develop if the improvement continues?
Do not automate the entire company overnight. Start with the smallest complete AI implementation that can produce useful evidence.
The red-pill answer is not hopeless
The evidence today points more strongly to job transformation than wholesale replacement. The International Labour Organization says transformation is the most likely near-term effect because most occupations still contain tasks requiring human input.
That is a snapshot, not a guarantee.
Over time, AI and robotics may perform much more of the work required for daily life. The destination could eventually resemble the Star Trek future people imagine: more abundance and more freedom to choose what we do. Getting there may also create enormous turmoil and force difficult questions about income, value and how productivity gains are shared. That larger economic question deserves its own article.
For now, workers need practice and a willingness to change how they contribute. Employers need to develop their people before replacing them and speak honestly when a role may not survive unchanged.
We may not be able to protect every job description. We can learn how the new system works—and make better decisions about how people move through it.
Written from Stephen Lau's experience and point of view, with AI-assisted research and editorial production. Reviewed by Salient AI before publication.