On May 5, 2026, Farhan Thawar & Shane Parrish submitted this brief to the House of Commons Standing Committee on Finance as part of pre-budget consultations for the 2026 federal budget. We are publishing it here so the policy ideas can contribute to the broader public conversation on how Canada can compete for founders, investors, and capital.

Introduction

Every few decades, a new technology arrives that reshapes what humans can do at work. Electrification in the 1920s. The personal computer in the 1980s. The internet in the 1990s. Artificial intelligence is the next entry on that list, and the evidence is already in: controlled workplace studies show productivity gains of 25 to 56 percent on knowledge-work tasks for people who use AI well.

The question facing Canada is not whether this technology will matter. It already does. The question is whether Canadians — individually and collectively — will be among the most capable users of it in the world, or whether we will spend the next decade watching other countries compound advantages we never built.

If people and businesses do not learn to compete using AI, they will quickly find themselves underemployed and out of business.

At present, we are losing the race. Canadian businesses report AI adoption at 12.2 percent.^[1] The OECD average is 20.2 percent.^[2] Canada trails the average of the club we belong to by roughly forty percent, and the gap is widening.

This submission proposes three tax measures to close that gap. None of them is novel in structure. Each one extends an existing, well-understood Canadian tax mechanism to AI adoption.

The Gap Is Wider Than It Looks

The headline number understates the problem. Two data points reveal the deeper issue.

First: among Canadian businesses not planning to adopt AI, 78.1 percent cite “not relevant to the goods or services they provide” as their primary reason.^[3] This is not a technology problem. It is a literacy problem. Businesses that have not seen AI applied to their work cannot recognize what it could do for them.

Second: among adopters, productivity gains are already substantial. Controlled studies show developers using GitHub Copilot complete coding tasks 55.8 percent faster.^[4] Forrester’s Total Economic Impact study of Microsoft 365 Copilot finds a three-year ROI of 132 to 353 percent for small and medium businesses, and 116 percent — with approximately US$20 million in net present value — for large enterprises.^[5]

The compounding effect is what matters. A country where 30 percent of workers use AI well generates more output, more learning, and more entrepreneurial velocity than a country where 10 percent of workers do — and the gap widens every year. This is not a subsidy question. It is a question about whether Canada builds a working population capable of thriving in the AI era.

Why Tax Policy Is the Right Instrument

Canada has a long tradition of using the tax code to accelerate strategic technology adoption. The architecture already exists. We are simply not using it for the most important technology of our era.

For individuals, existing precedents include:

• The Canada Training Credit, which reimburses up to half of eligible tuition expenses, is refundable, and accumulates at $250 per year up to a $5,000 lifetime limit for working-age Canadians.^[6]
• The Digital News Subscription Tax Credit (2020–2024), which offered a 15 percent non-refundable credit on qualifying journalism subscriptions up to $500 — a direct precedent for treating a specific type of subscription as worthy of tax relief.^[7]
• The tuition tax credit, which treats post-secondary learning as a core public good.

These programs reflect a durable principle: when a type of personal spending produces broad public value beyond the individual, the tax system should reduce its cost.

For businesses, existing precedents include:

• The Scientific Research and Experimental Development (SR&ED) program, which as of Budget 2025 delivers a 35 percent refundable investment tax credit to Canadian-controlled private corporations on qualifying expenditures up to $6 million annually (doubled from $3 million). The program now covers capital expenditures and certain Canadian public corporations.^[8]
• The United Kingdom’s 130 percent super-deduction (2021–2023), designed to accelerate post-pandemic business investment.^[9]
• The UK’s 186 percent R&D deduction for loss-making R&D-intensive SMEs under the Enhanced R&D Intensive Support scheme (since April 2024).^[10]
• Canadian film and video production tax credits, which can exceed 60 percent of eligible labour costs once federal and provincial credits are stacked.

Super-deductions and enhanced credits are proven instruments. The question is not whether Canada can design a generous tax incentive for a strategic technology. The question is whether AI is important enough to deserve one. It is.

Recommendation 1: The Canadian AI Subscription Deduction (CASD)

Create a 100 percent personal tax deduction of up to $3,000 per year for qualifying AI expenditures by Canadian residents.

Qualifying expenditures would include:

• Personal subscriptions to foundation-model services (ChatGPT, Claude, Copilot, Perplexity, Gemini, and successor products)
• AI-powered productivity tools with a primary AI function
• AI skills training, courses, and certifications
• Agent platforms and integration tools

The $3,000 annual limit reflects realistic usage: a premium AI subscription ($240–$600 per year), a second specialized tool ($300–$500 per year), and roughly $2,000 to $2,500 in AI-focused learning and tooling. Most claimants will not reach the cap. The design empowers serious users rather than subsidizing trivial ones.

The precedents are direct:

• The Canada Training Credit — built for exactly this kind of individual capability investment.
• The Digital News Subscription Tax Credit — established the principle that subscription-based public-interest content deserves tax treatment.
• The tuition tax credit — treats learning as a public good.

Estimated fiscal cost: $0.8–1.5 billion annually in the first three years, declining over time as competition drives subscription prices down. This estimate assumes 5 million claimants averaging $600 each at a blended 27 percent marginal rate. The public return — measured in human capital, entrepreneurial velocity, and productivity — is materially larger.

Recommendation 2: The Canadian AI Productivity Super-Deduction (CAPS)

Create a tiered business super-deduction for AI expenditures, structured to favour the small and medium-sized businesses that drive Canadian employment.

Tier 1 — Canadian-Controlled Private Corporations (CCPCs):

• 400 percent deduction on qualifying AI expenditures
• Up to $500,000 annually

Tier 2 — Larger Canadian Corporations:

• 200 percent deduction on qualifying AI expenditures
• Up to $5 million annually

Qualifying expenditures would include:

• Enterprise AI licenses and API spending
• AI integration consulting and implementation services
• Internal AI training and change management programs
• Custom AI tool development (excluding SR&ED-qualifying R&D, to prevent double-dipping)

The precedents:

• SR&ED’s 35 percent refundable credit for CCPCs on expenditures up to $6 million (Budget 2025)
• The UK 130 percent super-deduction (2021–2023)
• The UK 186 percent R&D super-deduction for SMEs
• Accelerated capital cost allowance classes for priority equipment
• Canadian film and video production tax credits, which can exceed 60 percent of qualifying labour costs when federal and provincial credits are stacked

The CCPC tilt is deliberate. Small and medium-sized Canadian businesses are the largest cohort citing “AI not relevant” as their reason for non-adoption. They are precisely the group for which a changed economic calculation is decisive. A 400 percent deduction on the first $500,000 means that a CCPC spending $100,000 on AI adoption effectively reduces taxable income by $400,000 of deduction, at a roughly 12.2 percent Ontario combined small-business rate, approximately $48,800 in gross tax savings on a $100,000 investment. For a business on the edge of adoption, that changes the decision from theoretical to obvious.

Estimated fiscal cost: $2–4 billion annually in the first three years. A full fiscal impact analysis — which this submission does not substitute for — would need to net productivity gains, tax base expansion, and talent retention against the direct revenue cost. International analogues (UK super-deduction, US R&D credit) consistently find payback within three to five years.

Recommendation 3: The AI Literacy Accountability Framework

Require Finance Canada to publish an annual AI literacy report measuring Canada’s standing against OECD peers on two metrics:

• Percentage of businesses using AI in the production of goods or services
• Per-capita business AI expenditure

Include a sunset mechanism: if Canada achieves a top-three OECD rank on both metrics for three consecutive years, both deductions taper by 25 percent per year until they reach baseline treatment of ordinary business and personal expenses.

The purpose is not to subsidize AI indefinitely. The purpose is to use the tax code to close a measurable gap and then retire the instrument once it has done its job. Programs designed without exit criteria tend to outlast their usefulness. This one will not.

Fiscal Context and Design Principles

Taken together, the combined first-year fiscal cost of Recommendations 1 and 2 is estimated at $2.8–5.5 billion. For reference, this is roughly equivalent to the annualized enhancement to SR&ED announced in Budget 2025, and smaller than Canada’s annual expenditure on accelerated capital cost allowance treatments for other priority sectors.

Three design principles underpin this submission:

First, extend what works. Every mechanism proposed is a structural sibling of an existing Canadian or allied-nation tax incentive. This is not an experiment.

Second, bias to the adopter. The goal is not to reward AI vendors or to direct capital to a specific domestic AI industry. The goal is to ensure that Canadians, as workers and firms, use AI at world-leading rates. Vendor neutrality is therefore a feature.

Third, build an exit plan. A permanent tax incentive tends to become an entitlement. A tax incentive with an objective exit criterion tends to deliver its intended effect and then retire.

Conclusion

The countries that will lead in the future will be those where the people use AI most effectively. Canada has all the inputs required to be on that list: a world-class AI research lineage, an educated workforce, a stable institutional environment, and a tax system flexible enough to accelerate strategic adoption when it chooses to.

What Canada currently lacks is a policy signal commensurate with the opportunity. A business sector reporting that AI is “not relevant” at 78 percent is not a technology problem. It is a national coordination problem.

The three recommendations in this submission require no new institutional architecture. They extend existing, proven Canadian tax mechanisms to the technology category, where adoption most directly determines future national capacity. They are designed to expire upon success.

Canada does not need to invent AI. Canada needs to ensure that Canadians use AI. That is a concrete, measurable, time-bound objective. It begins with the 2026 Budget.