Executing AI for All: How to Get a Capability-Raising Result 

From: Harvey Naglie 

To: AI watchers  

Date: June 10, 2026  

Re: Executing AI for All: How to Get a Capability-Raising Result 

Last week, Ottawa launched AI for All, a five-year national artificial-intelligence strategy built on six pillars and three stated aims: Trust, opportunity, and sovereignty. 

AI will add close to $200 billion in economic growth, the government projects, create up to 250,000 AI-related jobs, and raise business AI adoption from roughly 12 percent to 60 percent by 2034. 

The diagnoses behind the strategy are right and its foundations are sound: Canada helped lead the world in AI research but is among the slowest to adopt it at scale. Whether it raises Canada's productive capability or settles into subsidy will be decided not by the announcement but by execution – by legislation, contracts, and programs still to be written. 

Government’s role in shaping the sector, beyond funding research, education, competition policy, and stable regulation, remains a legitimate debate. But the strategy has been launched and now the practical question is execution. 

One distinction should govern the strategy's economic instruments: Whether each builds durable capability or merely compensates for its absence. Capability-building instruments create durable assets – trained people, research depth, infrastructure, retained intellectual property, and the institutions and standards that let domestic firms compete. 

Compensatory instruments stand in for capability the economy does not yet have, by subsidizing demand, supplying capital, or brokering bespoke deals. Compensation is legitimate and can buy time, but it tends to leave a bill rather than an asset. The aim of execution should be to push each instrument toward capability. 

Several components already exist and should be protected as the strategy is built out. The talent measures – additional CIFAR AI Chairs and faster entry through the Global Talent Stream – build human capital directly; a National AI Literacy Initiative and AI access for post-secondary students broaden the skills base; and expanding the Canadian AI Safety Institute and modernizing privacy and online-safety law build the standards and trust a functioning market needs. 

These raise capability whether or not the projected gains materialize, which makes them the strategy’s most reliable component. Sovereign compute, including a public AI supercomputer under Canadian governance, belongs on the capability side as well, provided it is put to active use. And because AI hardware advances quickly, a publicly funded machine needs a dedicated refresh cycle and budget, or will soon be outpaced. 

It is in its demand- and capital-side supports that the strategy's character is still unsettled: Procurement as a “strategic anchor customer,” “scaling Canadian champions,” improved access to growth capital, and the bilateral partnerships under the Sovereign Technology Alliance. Each can raise capability sustainably or merely subsidize it, depending largely on how it is designed. This is where execution matters most, because it is where the difference between capability-building and ongoing subsidy will be decided. 

The design choices are concrete. Anchor procurement should be structured to develop a domestic supplier base and transfer know-how, rather than simply guarantee revenue to incumbents. 

Growth capital should be conditioned on what it builds – commercialization, follow-on investment, scaling in Canada – but the conditions must stay light enough not to deter the investment they are meant to attract. 

Tie firms down too tightly and you suppress the acquisitions and exits that recycle talent and capital back into the next venture. Partnerships should be used to acquire capability Canada lacks – skills, standards, market access – with explicit terms on what is transferred and retained and judged by the capability they deliver rather than the deal they represent. 

And “intellectual property protections” should aim at the value Canada keeps from the IP it funds, not at ownership as an objective in itself. Sometimes IP ownership secures that value; more often, licensing, equity, or revenue terms do. For a small economy, a place in global innovation networks tends to be worth more than legal title to the IP – so the test is value captured, not flags planted. 

Two execution disciplines would strengthen the whole. 

First, convert aspiration into commitment. The strategy makes its promises in open-ended verbs – it targets, supports, and advances; each should carry a measurable obligation and a milestone, so progress can be tracked rather than declared. 

Second, measure outcomes, not activity. The reported tallies against the headline numbers – $200 billion in growth, 250,000 jobs, 60-percent adoption by 2034 – can climb while the outcomes that matter, productivity, commercialization, and competitiveness, disappoint. 

The headline projections should be broken down into the assumptions behind them, checked against independent estimates, and paired with a small set of outcome measures – firms reaching scale, productivity gains, commercialized intellectual property – so the strategy is judged by the real economic gains it produces, not by whether those headline targets are hit. 

AI for All has the right diagnosis and real foundations. The work that determines its return is now in front of the people writing its rules. 

A single question can anchor each decision: Does this instrument build something Canada will still have when the funding ends? Designed to that standard, the strategy can turn Canada's research lead into durable capability. 

Harvey Naglie is a researcher and author, and spent 40 years in financial services, and most recently served as a Senior Policy Advisor with the Ontario Ministry of Finance.  

To send a comment or leave feedback, email us at blog@cdhowe.org.  

The views expressed here are those of the author. The C.D. Howe Institute does not take corporate positions on policy matters.