Can Your Software or AI Project Qualify for SR&ED? What Canadian Companies Should Know cover image
    Artificial Intelligence · 6 min read

    Can Your Software or AI Project Qualify for SR&ED? What Canadian Companies Should Know

    October 1, 2026

    Can Your Software or AI Project Qualify for SR&ED? What Canadian Companies Should Know

    General information, not tax advice. Last reviewed October 1, 2026 against the Canada Revenue Agency's SR&ED guidance. Talk to a SR&ED or tax professional before filing a claim.

    Canada's Scientific Research and Experimental Development (SR&ED) program is one of the largest sources of support for companies doing technical R&D, and that can include software and AI work. But "we built software" isn't enough to qualify. Here is how to tell whether a project might be eligible, and how to run it so a claim holds up.

    What SR&ED is

    SR&ED is a federal tax incentive administered by the Canada Revenue Agency (CRA). Companies earn investment tax credits (ITCs) on eligible expenditures such as salaries and wages, materials, contracts and certain overhead.

    • Most Canadian-controlled private corporations (CCPCs) can earn a refundable ITC at the enhanced rate of 35% on qualified expenditures up to an annual limit. Under Budget 2025 measures (law since March 26, 2026), that limit is $6 million for tax years beginning after December 15, 2024, up from $3 million.
    • Other corporations generally earn a 15% non-refundable credit.
    • Provinces, including Québec, have their own R&D credits. Provincial credits and other government assistance, including NRC IRAP funding, reduce the amount you can claim federally, so plan them together.

    What makes work eligible

    CRA's guidelines describe two requirements:

    • The "why": the work aims at a scientific or technological advancement, and it faces scientific or technological uncertainty, meaning a problem you couldn't solve with standard practice or available knowledge.
    • The "how": a systematic investigation or search, by means of experiment or analysis: hypotheses, tests and analysis of the results.

    In software and AI, examples that can fit, depending on the details, include getting a model to reach accuracy or latency targets that no known approach meets on your data, or designing an architecture that handles a scale or reliability constraint existing patterns don't solve.

    What usually doesn't qualify on its own: routine development with established frameworks, integrating an API or model the documented way, solving a problem with a workaround based on available knowledge, or hiring experts to apply what they already know. The uncertainty has to be technological, not "we didn't know if customers would like it."

    How to run a project so the claim holds up

    1. Write down the uncertainty at the start: what you don't know how to do, and why standard approaches don't work.
    2. Record hypotheses and results as the work progresses: design notes, experiments, benchmarks and failed approaches. Failures are evidence too.
    3. Track time per person and per project.
    4. Separate R&D from routine work. Often only part of a project qualifies.
    5. Respect the deadline: your SR&ED reporting deadline is 12 months after your T2 return is due, roughly 18 months after your tax year ends, and CRA doesn't accept changes to a claim after it.

    Where a development partner fits

    If an external firm does part of the work, how the contract is structured affects who can claim which costs, so agree on it with your advisor before the project starts. At Algoseed Labs we run AI and software projects with this kind of documentation built in (discovery notes, experiment logs and sprint records), so your SR&ED advisor has what they need. We're not tax advisors, and we work alongside yours.

    Planning an AI or software R&D project? Talk to our Montréal team, or see our AI development and custom software services.

    Sources: Canada Revenue Agency, canada.ca: SR&ED tax incentive program; investment tax credit rates and expenditure limits; allowable expenditures; eligibility of work guidelines; reporting deadlines.

    All PostsArtificial Intelligence