The Weekly Brief — Issue 05

Hello,

Canada now has its clearest picture yet of how generative AI is being used at work—and the pattern is more limited, frequent and uneven than the headline alone suggests.

More than one in three Canadian workers used it on the job over the past year. Most users reach for it regularly, but only for a narrow share of their work. The survey tells us much less about whether employers have supplied the rules, training and oversight needed to manage that use.

Plus: an Inuit-led organization asks who should govern Indigenous language data, CIBC shows why “enterprise-wide” does not always mean fully deployed, and Canadian researchers train AI to predict where wildfires move next.

Here’s the week.

What happened

Statistics Canada has released the first results from TechStat, a new program measuring how Canadians use artificial intelligence at work.

In March, 35.9% of workers aged 15 to 69 said they had used generative AI as part of their main job or business during the previous 12 months. Counting other AI and automation technologies, the figure rises to 41.6%.

Use varies sharply by occupation. Three-quarters of workers in management reported using generative AI, along with 67.5% in natural and applied sciences. Use was much lower among workers in trades, transportation, natural resources and agriculture.

The more revealing finding is how people use the tools.

Nearly two-thirds of users—63.5%—apply generative AI to some, but not most, of their tasks. Another 24.9% use it for almost no tasks. Only 11.6% use it across most or nearly all of their work.

Within that limited set of tasks, use is frequent. Among users, 31.4% reach for generative AI daily and another 38.3% use it a few times a week. Together, that is nearly seven in ten users working with the technology at least several times a week.

Only 5% of non-users said workplace policy was the reason they had stayed away. That figure does not tell us whether people already using AI received employer approval, training or rules. It only tells us that formal policy was rarely the reason non-users gave for not adopting it.

Why it matters

Most of Canada’s workplace-AI conversation has focused on forecasts: which jobs may change, which tasks may disappear and how employers should prepare.

The new data shows that adoption is already underway. The immediate questions are now operational: which tools are approved, what information may be entered, where human review is mandatory and how workers learn to recognize output that is wrong.

The numbers also point to an uneven transition. Workers in management, science, finance and education are gaining regular experience with these systems, while many people in trades, transportation, agriculture and service work have little exposure.

Without deliberate access and training, AI could widen existing differences between occupations and workplaces.

AI adoption is no longer an executive forecast. It is already employee behaviour.

Your move

Before using generative AI for a work task, ask two questions.

What information am I putting in?

Remove anything personal, confidential, commercially sensitive or legally protected unless your organization has explicitly approved that system for that use.

Who is accountable for the output?

AI can produce a draft, summary or analysis. Responsibility for what gets sent, filed or decided still belongs to a person—you.

If your workplace has no written answer to either question, that gap is worth raising.

Indigenous languages are not simply AI training data

Heritage Lab, an Inuit-led organization working on AI, digital sovereignty and Indigenous language protection, has challenged a Université Laval–Anthropic initiative involving Quebec French and Indigenous languages.

Anthropic says the research will examine how large language models behave across varied cultural contexts and “low-resource” languages and dialects.

Heritage Lab argues that grouping Quebec French with Inuktitut, Innu, Naskapi, Cree, Anishinaabemowin and other Indigenous languages overlooks a fundamental distinction: these are the languages of distinct Nations with their own rights, governments and responsibilities.

The organization is not arguing against research. It is asking who helped design the project, who will own and govern the language data, what safeguards will prevent later unauthorized uses, whether communities can withdraw and who will receive the long-term benefits.

Why it matters

The material chosen to train an AI model shapes which dialects, meanings and ways of speaking it reproduces. Gaps or imbalances can later appear as errors, omissions or distorted representations in the model’s output.

Participation in a research project is not enough on its own. Indigenous communities also need meaningful authority over how their languages are collected, used and governed after the project begins.

What CIBC’s “enterprise-wide” AI actually means

CIBC announced CIBC AI 2.0, an internal agentic-AI workspace that can gather research, organize findings, prepare presentations, compare agreements and coordinate compliance work.

The phrase “enterprise-wide” needs context. CIBC says the new agentic platform is currently in pilot.

The bank’s earlier chat-based assistant is the system already available across the organization—to more than 50,000 employees, with an average of 20,000 using it each day.

A day earlier, TD introduced seven enterprise-wide Responsible AI Principles covering existing law, transparency and explainability, data use and privacy, fairness, quality and accountability, reliability, and security.

TD says the principles apply across the AI lifecycle and are being incorporated into employee education, assessments and ongoing monitoring.

Why it matters

The announcements describe two different stages of corporate AI adoption. CIBC is testing a more capable system on top of an established internal assistant. TD is publishing the framework it says will govern AI systems across the bank.

The useful question is not only what was announced. It is what has actually been deployed, to whom, for which tasks and under which controls.

AI could help predict where Canadian wildfires move next

University of Toronto researchers are developing an AI system that analyzes colour and infrared footage collected by firefighting aircraft.

The system identifies fire, smoke and terrain, maps the images onto three-dimensional landscapes and combines those measurements with weather and wind data. The goal is to forecast how a fire may grow over the following hours.

The project, called TankerVision, logged 50 fires last season and hopes to capture more than 100 this year.

The benchmark it is trying to improve is striking. Current wildfire-growth estimates still rely partly on data from 15 to 20 prescribed burns conducted in the 1980s, combined with current weather, forest conditions and decades of human judgment.

Why it matters

Better forecasts could help officials decide where to send crews, when to reposition equipment and whether communities need to evacuate.

It is also a useful example of AI supporting expert judgment rather than replacing it. The technology is intended to give wildfire professionals better and more current information—not to make high-stakes emergency decisions by itself.

Two Statistics Canada figures are now circulating.

One says 35.9% of Canadian workers used generative AI as part of their work. Another says 19.2% of Canadian businesses used AI to produce goods or deliver services.

Both can be accurate because they measure different things.

The worker survey counts individual people who used generative AI during their job or business. That could include someone using an AI tool for drafting or research even if their employer has not formally introduced AI across the organization.

The business survey asks whether the organization used any form of AI to produce goods or deliver services. It measures business-level adoption, not every instance of an employee using a tool.

The figures also cover different technologies, populations and survey periods. They should not be used to conclude that workers are adopting AI at a precise rate faster than their employers.

What they do suggest is that individual workplace use may be broader than formal business deployment.

Keep a record of every work task where you use generative AI this week. For each one, write down:

At the end of the week, look closely at the final item.

The goal is not to use AI more. It is to identify where the tool genuinely saves effort—and where it merely moves the work from doing to checking.

Do not include confidential, personal or protected workplace information in your log. Describe the type of information rather than copying the information itself.

A calm, practical briefing on the AI developments that matter in Canada—delivered every Tuesday.

That's the week. See you next Tuesday.

— Padge T.

AI Brief Canada is an editorially independent publication published by Groundshift Advisory Inc. and written by Padge T.

Groundshift Advisory helps Canadian organizations adopt AI responsibly. Sponsored, affiliated, or commercially connected content will always be clearly disclosed.

Issue disclosure: Nothing in this issue is sponsored, affiliated, paid, or commercially connected.

AI use: AI tools supported research, source discovery, outlining and drafting for this issue. All material was reviewed, edited and fact-checked by Padge T. before publication. Read our AI Use Disclosure.

Not affiliated with the Government of Canada.

Corrections or feedback: We welcome corrections and reader feedback. Email [email protected] or review our Editorial Policy.