Machine learning is an approach where a system’s behaviour comes from patterns found in data, rather than from a rule a person writes out in advance. Canada’s Directive on Automated Decision-Making names it as one of six specific techniques an automated decision system may use — alongside rules-based systems, regression, predictive analytics, deep learning and neural networks — naming it separately from a plain, hand-written rule.
Statistics Canada’s Q2 2026 survey of AI use by Canadian businesses measured actual use directly: among Canadian businesses that used AI in the 12 months to Q2 2026, “Machine learning” was a reported application for 18.2%, essentially flat against 18.6% a year earlier — one of sixteen named application categories Statistics Canada tracks separately from “Machine or computer vision” (4.6%) and deep learning (13.1%). That is real, dated Canadian adoption data, not an estimate.
Canada’s Voluntary Code of Conduct on generative AI explains why the data-driven nature of the technique carries real regulatory weight. It attributes the “distinctly broad risk profile” of advanced generative systems to being “due to the broad scope of data on which they are trained, their wide range of potential uses, and the scale of their deployment” — in other words, precisely because their behaviour comes from data rather than from a rule someone can read and audit line by line.
A Canadian lender sorting mortgage-file intake by a written rule — missing a signature routes to manual review, nothing else does — is not doing machine learning: the behaviour is fixed until someone edits the rule. Replace that rule with a system trained on years of past files to learn which patterns actually predicted a problem, and the behaviour now comes from the training data. The lender can no longer point to a line of code and say “this is why it flagged that file” — which is exactly the accountability question the Directive’s own testing requirement is built to force someone to answer before the system goes live.
See also deep learning, neural network and AI model.
This is one term in a plain-English glossary on how AI actually works and where it fits in a Canadian business.
