Software Development is one of the fastest-evolving industries thanks to AI tools. This means that submitting a successful SR&ED claim for software development has become more complicated. What was difficult to solve in the past can be easily fixed by Claude or Gemini. The technical challenge of software development has diminished because of incredibly powerful AI tools. The tools that have evolved in the past two years are used by almost every developer. As IT claims make up a majority of SR&ED claims, how do IT companies continue to take advantage of the SR&ED program with AI being everywhere?
It is doubtful that the CRA will change the definition of SR&ED due to AI. After all, the SR&ED program is intended to promote innovation and create a competitive differentiation for Canadian companies. The good news is that software companies will most likely be able to submit an SR&ED claim that is as large, if not larger than they have in the past – even with AI taking some of the workload.
We have a number of blogs on what SR&ED is, how to maximize SR&ED claims and some best practices. This article is not going to discuss any of these topics, but it will focus on how to establish a strong SR&ED claim for software development companies.
AI is replacing some traditional tasks. As of today, it is rare to see AI replacing the entire development team within a company. Its greatest benefit is that it allows developers to be more efficient. This has the biggest impact on junior developers. AI coding is as good as, if not better than, junior developers and therefore these roles are being displaced fastest. However, from an SR&ED perspective, this is a double-edged sword. The bad news is that with fewer labour expenses, the SR&ED claim is lower. The good news is that much of what a junior developer codes does not qualify for SR&ED in the first place.
Systematic Investigation in the Age of AI
Consider a developer solving a performance and scalability issue. The nature of the issue may be understood by the developer or architect, but the solution may not be obvious. After searching Gemini and getting some information on how to solve the problem, the solution did not work. The developer uses Cursor to create a solution to the problem. The first iteration miserably fails. The developer debugs the code and then tries a different approach. This would be a great SR&ED project. Everything the developer is doing would qualify for SR&ED. The fact that they are more efficient by utilizing AI tools does not reduce the quality of the SR&ED claim. In fact, it may strengthen the SR&ED claim because the routine work that is usually done by more junior developers is now done by AI.
Fundamentally, the ability to claim SR&ED for software development has not changed with the evolution of AI. As the nature of software development moves from coding to architecture, there is most likely more SR&ED-eligible activities because companies will push the limits of technology, and there will be a huge number of challenges. The challenges can be in multiple areas. Keep these in mind during your daily work and note what challenges you or your team are facing. Contemporaneous documentation is a key factor in preparing a rock-solid SR&ED claim.
Software Development SR&ED Activities
There are countless projects a software company could do that would be SR&ED eligible. The key factors are technical uncertainty, systematic investigation, and adequate supporting evidence to justify the work. Here are a few examples of projects that would likely qualify for SR&ED.
- Algorithms and Data Processing:
- Developing new or more efficient algorithms (e.g., for search, optimization, or recommendation engines).
- Handling large-scale data sets (performance, latency, accuracy).
- Advancing machine learning models or training methods.
- System Architecture and Infrastructure
- Designing distributed systems that scale reliably.
- Improving fault tolerance, load balancing, or concurrency.
- Optimizing cloud or edge computing deployment.
- Performance and Scalability
- Reducing latency or increasing throughput beyond existing solutions.
- Memory management and resource optimization.
- Benchmarking and tuning for high-demand use cases.
- Security and Privacy
- Developing new encryption or anonymization methods.
- Securing APIs, authentication systems, or data pipelines.
- Meeting regulatory requirements through novel approaches (e.g., GDPR-compliant data processing).
- Interoperability and Integration
- Connecting disparate platforms, APIs, or hardware in ways not previously done.
- Building middleware to handle conflicting protocols or data formats.
- Ensuring compatibility across multiple environments.
- User Experience & Interface Technologies
- Creating new interaction models (e.g., voice, AR/VR, gesture recognition).
- Accessibility innovations for diverse user groups.
- Real-time rendering or visualization improvements.
- Quality, Reliability, and Testing
- Automating testing for complex, dynamic systems.
- Developing frameworks to simulate real-world edge cases.
- Advancing continuous deployment and integration pipelines.
- Domain-Specific Applications
- Healthcare: medical image analysis, clinical decision tools.
- Fintech: fraud detection, high-frequency trading platforms.
- Manufacturing: digital twins, predictive maintenance software.
- Cleantech: smart grid optimization, emissions monitoring.
By focusing on the technological uncertainties, software companies will continue to take full advantage of the SR&ED program.
Impacts of AI on SR&ED Claim Size
Won’t AI reduce labour expenses, thereby making SR&ED claims smaller? Like many answers regarding SR&ED, it depends.
Looking back at the start of computer programming, we have seen a continued evolution of tools. This is true for both hardware, middleware and software. I’m old enough to remember connecting my TV with a single Coaxial cable to the clicker box, which was connected to the TV with another single coaxial cable, and everything worked. Today, I have the choice of dozens of cables, versions of cables, and multiple configurations, making connecting a TV infinitely more complicated, while technology has “simplified” many tasks.
The fact that development languages have evolved, platforms are more capable, compute is more readily available has in no way simplified software development. The complexity of creating a unique offering is exactly what leads to great SR&ED claims.
Because AI will do much of the lower-level work, and the CRA will understand this, it is most likely that the CRA reviews will be simpler in the future. An SR&ED review (or audit) can be a taxing experience. However, because the CRA is now aware that a lot of the basic tasks may not have been supporting work for SR&ED are being automated, it is more probable that what will be claimed is eligible for SR&ED. I believe that SR&ED claims for software development will increase in size as greater efficiency and more complexity are introduced to the discipline.
Will there be more SR&ED projects claimed?
A key aspect of preparing an SR&ED claim is to determine the “level” at which to claim a project. If the level is overly granular, the CRA may not approve it, or there may be many SR&ED projects for a relatively small SR&ED claim. On the other hand, if there is just a single project for all technical uncertainties, there may not be enough information to satisfy the CRA, and therefore, it is more likely that the claim is pulled into a review. There is a delicate balance to determine when to create a new project.
Unfortunately, many SR&ED consultants tend to have fewer projects because of the cost of preparing each project. With the increase in efficiency, it is more likely that there will be more projects and more work for consultants to do. Ensure that your consulting partner has no issue with having multiple projects, no matter how many expenses are associated with each project. At GrowWise, we don’t care how many projects you have, because we’ve built the tech that maximizes SR&ED claim size regardless of the number of projects.
How the CRA Views AI in SR&ED Claims
The CRA cares about systematic investigation and overcoming technical uncertainty. They want you to have the supporting documentation to show that you did the work and that you have a logical method of tracking time. The CRA’s focus can apply to a biotech, software or manufacturing company. The tools the company uses are not particularly relevant for the purposes of SR&ED. Using AI is one of the tools that aid in development, but in no way does it diminish a strong SR&ED claim.
AI does not disqualify work, but claimants must show that human-led experimentation is still present. While today humans are in the loop, there may be a point in the future where AI is better at determining what the technical uncertainties are, diagnosing them and conducting systematic investigations to rectify them. AI may be able to learn and deploy solutions that reduce the need for human intervention. In 2025, it is likely that most development and experimentation is led by humans. From the CRA’s perspective, it is important that the SR&ED claimant demonstrates that people are leading the project and there is documentation supporting this.
SR&ED Documentation Best Practices in the AI Era
To effectively support an SR&ED claim, any claimant must be able to show contemporaneous documentation. With AI tools providing a great deal of assistance, it is even more important to effectively document SR&ED activities. There are many tools available to help manage projects and subsequently document SR&ED activities. Tools/software that make documentation easier include GitHub, Jira, Notion, ClickUp, etc. We suggest that you pick the project management or bug tracking tool that best suits your business needs and then customize it to easily document SR&ED. GrowWise can read most project management software to automatically extract SR&ED-eligible projects. The need for documentation will become even more important as AI proliferates in R&D.
Conclusion
The future is bright for software companies that claim SR&ED. Claim sizes are most likely going to increase as there will be more technical challenges as the world becomes more complex, and companies want to build moats around their businesses. Documentation will continue to be a cornerstone of the SR&ED program and will be even more important as AI takes over some of the basic tasks. As we learn more about how the CRA evolves in their treatment of AI, we will continue to update you so that you can maximize your SR&ED claims. We’ve put together a quick checklist for you to keep handy as you are developing your software so that you can get the most out of the SR&ED program.
SR&ED Claim Checklist for Software Development in 2025
1. Identify Eligible Projects
- Define the technological uncertainty you are addressing (e.g., performance, scalability, integration, security).
- Confirm that the work goes beyond routine development or configuration.
- Ensure there is a systematic investigation (hypothesize, test, analyze, iterate).
2. Track Project Activities
- Document the problem definition and why existing tools or methods were insufficient.
- Record failed experiments or iterations (not just the final solution).
- Capture how AI tools were used (e.g., to accelerate coding, generate alternatives, or test approaches).
- Distinguish between AI-generated work and developer-led problem solving.
3. Capture Supporting Evidence
- Maintain contemporaneous documentation (meeting notes, commit logs, Jira tickets, version control history).
- Save benchmark results or performance metrics from experiments.
- Archive test cases and debugging logs.
- Keep records of design discussions and architectural decisions.
4. Track Costs and Resources
- Log developer, architect and supporting hours tied to eligible SR&ED activities.
- Separate routine development from experimental work.
- Track cloud computing or specialized infrastructure costs where applicable (the CRA may allow this in 2025).
- Verify all contractor and consultant contributions are documented.
5. Organize for CRA Review
- Group work into projects at the right level of detail (not too broad, not too granular).
- Ensure documentation clearly links technical uncertainty → attempted solution → results.
- Prepare a summary of AI usage and explain how it enhanced, but did not replace, systematic investigation.
- Validate that supporting evidence aligns with CRA’s eligibility criteria.
