Heather Fricke is the founder of Frick-E Energy™, where her work focuses on the Human-AI Understanding Gap: what happens to human meaning, context and judgment when machines interpret information before people or organizations act on it. She is the creator of Promptology™ and Human Algorithm™.

 Canada has set an ambitious AI adoption target. Its new AI for All strategy aims to move business adoption from roughly 12 percent to 60 percent by 2034, with particular attention to small and medium-sized enterprises.

Statistics Canada’s latest data already show acceleration: 19.2 percent of businesses reported using AI to produce goods or deliver services in the second quarter of 2026, up from 6.1 percent two years earlier.
That is meaningful progress. It also creates a measurement problem we are not talking about enough.
We are getting better at counting whether businesses use AI. We are still weak at measuring whether the AI actually understood the business, the human judgment behind the task, or the context required to make the right decision.
Those are not the same outcome.
Statistics Canada reports that among businesses already using AI: 36.6 percent use it for data analytics, 34.5 percent for text analytics, 28.2 percent for virtual agents or chatbots, and 13.7 percent for decision-making systems. That last category more than doubled from the
previous year.
The closer AI gets to decisions, the more expensive interpretation becomes.
A system can retrieve the correct policy, customer record, research document or operating data and still reconstruct the wrong meaning from it. The facts can be accurate while the decision is wrong because the hierarchy, exception, context or human boundary was never made explicit enough for the machine to preserve.
I call this the Human-AI Understanding Gap: the distance between what a person or organization means and what a machine can reliably reconstruct from the representations available to it.
The problem is easy to miss because modern AI is exceptionally fluent. When a system produces a polished answer, humans naturally give that fluency more credit than it deserves. We tend to inspect obvious hallucinations and factual mistakes. Interpretation failures are quieter. The answer may contain no invented fact at all.
Imagine an SME using an AI assistant to screen suppliers. The company has an unwritten rule that a certain certification matters only when the contract crosses a particular risk threshold. Employees know this because they have lived through the consequences. The documents
available to the AI mention the certification and the threshold separately, but never explain the relationship.
The model retrieves both facts correctly. It still recommends the wrong supplier.
That is not primarily a retrieval failure. It is not necessarily a model-capability failure either. It is a meaning-transfer failure.
Canada’s AI strategy correctly emphasizes practical, sector-specific adoption, trusted systems, skills and AI literacy. The next layer of that literacy should include the ability to test what the machine understood, not merely whether employees know how to operate the tool.
This matters especially for SMEs. Large organizations can build evaluation teams, governance processes and technical controls around AI deployment. A small manufacturer, professional-services firm, retailer or community organization is far more likely to put AI into an existing workflow and judge success by speed, convenience or output quality.
That creates a strange risk: the organization can become more efficient at repeating a misunderstanding.
The remedy is not to slow adoption or bury small businesses under another compliance vocabulary parade. Canada is right to want more businesses using AI. But adoption programs should distinguish at least three different questions:

• Did the system retrieve the right information?
• Did it interpret that information in the way the organization intended?
• Did the resulting action preserve the human decision boundaries that
  actually matter?

Those questions require different evidence.
My work through Frick-E Energy™ focuses on that interpretation layer. Promptology™ examines how meaning is exposed, tested and corrected across the human-AI gap. Human Algorithm™ names the human logic underneath the words: priorities, exceptions, lived context, evidence standards and decision rules that a machine cannot legitimately invent
for us.
The public-policy implication is larger than any one framework. AI measurement should begin separating usage from understanding.
Statistics Canada’s new TechStat initiative is explicitly designed to give decision-makers clearer evidence about how AI is affecting the economy, labour market and society. That creates an opportunity to ask a more mature generation of adoption questions.
Not only: Are firms using AI?
Also: Where is AI influencing decisions? How are organizations validating interpretation? What kinds of context failures are being detected? Where does human review remain necessary? Which sectors experience the largest gap between technically correct outputs and
operationally correct decisions?
Canada helped build the modern AI research ecosystem. Its next competitive advantage will not come from adoption volume alone.
It will come from learning how to use AI without confusing machine fluency with shared understanding.
If Canada wants 60 percent of businesses using AI by 2034, the real success metric should not be whether six in ten companies can say they deployed it.
It should be whether those companies became better at making decisions after they did.

Sources:
Statistics Canada, “Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026”:
https://www150.statcan.gc.ca/n1/pub/11-621-m/11-621-m2026010-eng.htm
Innovation, Science and Economic Development Canada, “Canada’s National Artificial Intelligence Strategy: AI for All”:
https://ised-isde.canada.ca/site/ised/en/canadas-national-artificial-intelligence-strategy-ai-all
Statistics Canada, “TechStat: Making sense of AI’s impact on Canada”: https://www.statcan.gc.ca/en/trust/ai/techstat-ai-impact-canada