Artificial intelligence (AI) and machine learning (ML), including Large Language Models (LLMs), are rapidly transforming industries – from manufacturing, construction, transport and energy to professional services such as health, education, law, and marketing. The range and sophistication of applications is exploding.
At Matthews Steer, we often receive enquiries from companies asking if their AI projects are eligible for the R&D Tax Incentive. While some projects qualify, many do not.
So, what would make an AI project eligible for the R&D tax incentive?
Understanding R&D Eligibility for AI Projects
Firstly, it’s important to note that simply using machine learning or LLMs does not automatically make your project eligible for the R&D Tax Incentive.
The key principles of R&D Tax Incentive eligibility have not changed, and still need to be satisfied. An eligible R&D claim must have activities that:
- Are experimental, with outcomes that cannot be known or determined in advance
- Can only be determined by applying a systematic progression of work
- Are based on principles of established science
- Follow a process from hypothesis to experiment, observation and evaluation, and lead to logical conclusions
- Are conducted for the purpose of generating new knowledge.
That all sounds quite scientific, and for good reason. The legislation is designed to support technically challenging R&D, so documentation and process are critical.
What the Department of Industry Science and Resources (DISR) and ATO Look For?
So, what does that mean in a practical sense, particularly in terms of what the Department of Industry Science and Resources (DISR) and the ATO want to see when it comes to AI?
DISR requires clear evidence that your AI project meets these criteria:
- Background research: Have you done your homework and conducted research into the state of the art or a literature review.
- Technical unknowns: Have you clearly articulated and recorded the knowledge gaps and technical risks of the project?
- Hypothesis and approach: How do you plan to overcome the knowledge gaps, unknowns and risks?
- Experimental design: How will you conduct the experiment? What technical parameters and methodology will you use?
- Novelty: Could a competent professional have predicted the outcome in advance?
- Evaluation: How will you assess and document the results?
DISR must see that each step has been systematically documented.
An R&D claim for an AI project is very similar and, in many cases, exactly the same as a software project.
Common Errors in AI R&D Claims
- Claiming the entire project instead of specific activities. A valid R&D project must be anchored by core activities and supplemented with supporting activities that enable the core activities. Only those things can be claimed.
- Failing to clearly identify specific knowledge gaps.
- Not documenting activities as you go. Including inadequate technical documentation, financial records and, importantly, timesheets or activity records for those involved in R&D activities.
- Treating routine development activities as experimentation. The functionality may be new and novel but if built on readily available, existing technology then it is not eligible
- Claiming activities conducted overseas.
Activities That Are Often Not Eligible
- Training an AI engine
- Prompt engineering
- Retrieval-Augmented Generation (RAG)
- Configuration and provisioning of infrastructure (may be eligible support activities)
- Using open-source or commercial models with minor changes
- Evaluating existing software or hardware solutions for your specific purposes
Activities That May Qualify as R&D:
- Developing novel ways to orchestrate disparate AI systems to produce a service that is responsive in real-time
- Developing novel ML algorithms or models
- Developing a new AI engine
- Adaptive robotics with reinforcement learning
- Designing AI-based fraud detection systems using unsupervised learning
If you’re exploring whether your AI project qualifies for the R&D Tax Incentive, our Innovation + Grants team can help. We provide expert guidance on identifying eligible activities, documenting R&D, and maximising your claim. Contact Matthews Steer’s Innovation + Grants team today to discuss your project.
About the Authors
Dr Richard Wraith – Partner and Head of Innovation + Grants at Matthews Steer, Richard has over 15 years’ experience helping Australian companies leverage the R&D Tax Incentive, with deep expertise in technology-driven innovation.
Dr Nicholas Kryatzis – Senior Innovation + Grants Consultant, Nick holds a PhD in Chemistry and is passionate about helping R&D-focused companies maximise the impact of their innovations from lab to market, including projects in emerging technologies such as AI and machine learning.