Gleb Tsipursky, PhD, is a behavioral scientist, CEO of Disaster Avoidance Experts in Columbus, Ohio, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026).

 Canada’s AI plans contain the right ambition. The AI for All strategy aims to create up to 90,000 AI-related jobs and work placement opportunities for young Canadians by 2031, expand AI literacy and speed adoption among small and medium-sized businesses.

But there is a missing layer between learning an AI tool and becoming valuable in an AI-enabled workplace: judgment.

Research Money recently highlighted Canada’s need to turn world-class AI research and talent into Canadian-made and adopted technologies. The publication’s Youth Voices coverage also points to a persistent question about how young Canadians build careers in a changing economy. The answer cannot be more prompt-writing courses alone.

We should treat early-career AI work as a judgment apprenticeship. A judgment apprenticeship is a deliberate, structured training method used by organizations and guided by mentors to tech junior workers how to think critically and make complex decisions now that AI handles routine entry-level tasks.

A judgment apprenticeship replaces passive learning with a conscious design for expertise.

That means new hires and students practise not only getting an answer from an AI system but deciding what to do with it.

Every important AI-assisted task should end with three questions: What did the system assume? What evidence did I verify? What would make me escalate this to a person with more expertise?

This matters because AI use is spreading faster than formal organizational systems. Statistics Canada reported that 22 percent of Canadian workers had used generative AI at work in the previous 12 months during its study period, while reported use rose from 17 percent in September 2024 to 30 percent in July 2025. That suggests many workers are learning AI in real time, sometimes before their employer has mature rules for it.

A judgment apprenticeship can be practical. In marketing, junior staff can compare AI-written audience claims against source data and record corrections. In finance, they can tag assumptions that require human sign-off. In software, they can test AI-generated code against failure cases instead of accepting a clean-looking output. In public service, they can practice identifying when a chatbot response touches rights, benefits, privacy, or another decision that deserves escalation.

Managers matter just as much. If leaders reward only speed, junior staff learn to hide hesitation and accept plausible output. If leaders ask for correction logs and brief after-action reviews, they make skepticism part of good performance.

The point is not to slow every task. It is to create friction where the cost of being confidently wrong is high.

This also gives policymakers better measures. Counting course completions and AI accounts tells us about access. It does not tell us whether people can spot a weak output, challenge it and improve a decision. Work-integrated learning programs should measure verified corrections, escalation quality and whether participants can explain why an AI recommendation was accepted or rejected.

Canada’s AI strategy explicitly says literacy should include recognizing bias, misinformation, privacy risks and unsafe uses. A judgment apprenticeship turns those ideas into workplace behavior. It connects education, employment, and adoption.

For young Canadians, this is also a better career proposition. Generic tool fluency ages quickly because interfaces change. Judgment is portable. A graduate who can verify evidence, identify uncertainty and know when to bring in another human can adapt as models improve. That is the skill employers will still need when today’s prompt tricks are obsolete.

Canada does not lack AI talent. It risks creating a gap between technical capability and everyday organizational judgment. Closing that gap would make the youth-jobs promise in AI for All more credible and help Canadian organizations adopt AI without turning every new worker into an unmonitored test case.

If we want the next generation to build and use Canadian AI, give them more than access. Give them structured practice in deciding when the machine is useful, when it is wrong, and when a human must remain responsible.

R$