Australian businesses spending money on software development might be eligible for a 43.5% tax offset on eligible R&D expenditure. Most don't claim it because they don't realize their work qualifies.Here's what actually qualifies under the R&D Tax Incentive for software projects, how much you can claim, and how to structure projects to maximize your benefit.Important: This isn't tax advice. Talk to an R&D tax consultant or accountant before making claims. This is a practical guide to understanding if your software development might qualify.
What the R&D Tax Incentive actually is
The R&D Tax Incentive is an Australian Government program that gives tax offsets for eligible research and development activities. For companies with turnover under $20M, you get a refundable 43.5% tax offset. Companies over $20M get non-refundable offsets at varying rates.What this means in practice: If you spend $100,000 on eligible R&D activities, you get $43,500 back (assuming you're under $20M turnover and your project qualifies). This is on top of the normal business expense deduction.The program exists to encourage innovation in Australian businesses. Software development can qualify, but not all software development counts.
What software development qualifies
The core test is whether your project involves experimental activities aimed at generating new knowledge or creating new or improved materials, products, devices, processes, or services.Translation: You're trying to build something where the technical approach isn't already known or readily available. You're experimenting to figure out how to solve a technical problem.
Examples that typically qualify
Custom AI integration solving novel technical challengesYou're building an AI system for document analysis in legal discovery, but existing OCR and NLP solutions don't work well with Australian legal document formats. You need to experiment with different approaches to achieve acceptable accuracy. The experimentation to solve this technical problem qualifies.Experimental software architectureYou're building a real-time data processing system that needs to handle 100,000 events per second with sub-50ms latency. The technical approach to achieve this isn't documented. You experiment with different architectures, databases, and processing frameworks. The experimental work qualifies.Novel algorithm developmentYou're creating an optimization algorithm for workforce scheduling that needs to handle constraints traditional algorithms can't solve efficiently. You experiment with different mathematical approaches and hybrid solutions. The algorithm development work qualifies.Innovative integration of existing technologiesYou're combining vector databases, LLMs, and business process automation in ways not documented in existing literature to create an AI operating system for professional services. The experimental integration work qualifies.
Examples that typically don't qualify
Standard web developmentBuilding a website with React, a PostgreSQL database, and standard authentication. This uses well-documented approaches. Doesn't qualify.Routine software customizationConfiguring Salesforce or HubSpot for a client's specific needs. No experimental activity. Doesn't qualify.Using existing libraries in standard waysConnecting your CRM to your accounting software using documented APIs. Known approach. Doesn't qualify.Cosmetic changes to existing softwareRedesigning the UI of your application without solving new technical problems. Doesn't qualify.The distinction: If you're following documentation or tutorials, it's not R&D. If you're experimenting because there is no documentation, it might be.
What counts as eligible expenditure
If your software development qualifies as R&D, here's what you can claim:Staff salaries: Wages paid to employees conducting R&D activities, proportional to time spent on R&D.Contractor costs: Payments to contractors for R&D work, at 100% of the cost.Cloud and infrastructure costs: AWS, Azure, OpenAI API costs, Supabase hosting - if directly used for R&D activities.Software licenses: Development tools, AI model access, databases used in R&D.Depreciation: Equipment used for R&D activities.What you can't claim: Marketing, general admin, routine maintenance, business-as-usual software development.
How to structure projects for R&D eligibility
If you're planning software development that might qualify, structure it properly from the start.
Document the technical uncertainty
Before starting, document what technical problem you're trying to solve and why existing approaches won't work. This becomes evidence that experimental activity was needed.Example: "We need to process legal contracts and extract specific clauses. Standard NLP models achieve only 60% accuracy on Australian contracts due to jurisdictional language variations. We will experiment with fine-tuning approaches and retrieval-augmented generation to achieve target 95% accuracy."
Separate R&D from routine development
Your project probably has R&D components and routine components. Track them separately.Example: Building an AI receptionist might include experimental work (training custom models, novel conversation state management) and routine work (user authentication, database CRUD operations). Only the experimental work qualifies.
Keep detailed records
AusIndustry (who administers the program) may audit your claim. You need evidence:
- Technical documents explaining what you were trying to achieve
- Development logs showing experimentation and iteration
- Test results demonstrating hypotheses tested
- Time tracking showing who worked on R&D activities
- Expenditure records for R&D costsDon't fabricate this after the fact. Track it during development.
Work with R&D tax consultants early
R&D tax consultants help structure your project documentation and maximize your claim. Engage them before starting the project, not after. They typically charge 20-25% of the successful claim amount.
Realistic claim amounts for software projects
Here's what Australian software development projects typically claim:Small AI integration project$50,000 in eligible expenditure (2 developers, 3 months part-time, cloud costs). Tax offset: $21,750.Medium custom software development$150,000 in eligible expenditure (3 developers, 6 months, contractor costs, infrastructure). Tax offset: $65,250.Large R&D initiative$500,000+ in eligible expenditure (dedicated team, 12+ months, significant experimental work). Tax offset: $217,500+.Remember, not all your software development costs qualify. Typically 30-60% of total project costs are eligible, depending on how experimental the work is.
How ThinkSwift structures projects for R&D eligibility
When clients ask about R&D tax claims, we approach it honestly: we don't design projects just to maximize claims, but we document qualifying work properly.What we do:At project scoping, we identify which technical challenges involve genuine uncertainty and experimental work. We document the technical hypothesis, why existing approaches won't work, and what we'll need to experiment with.During development, we separate experimental work from routine implementation in our project tracking. Our developers log time against specific technical challenges, not just general "development."We maintain technical documentation showing what we tried, what worked, what didn't, and why. This is good engineering practice anyway, but it's also evidence for R&D claims.We provide clients with detailed breakdowns of eligible vs non-eligible expenditure so their R&D consultants can prepare claims.What we don't do:We don't inflate claims by calling routine work "research." We don't backdate documentation. We don't promise specific claim amounts (that's between the client and their R&D consultant).
Common misconceptions
"Only large companies or tech startups can claim"
False. Any Australian company developing software with experimental components can claim, regardless of size or industry. Professional services firms, manufacturers, and small businesses all successfully claim.
"You need a lab or scientists"
False. Software R&D doesn't require physical labs. It requires genuine technical uncertainty and experimental activity. Many successful claims are from small development teams.
"Routine innovation counts as R&D"
False. Being innovative in your business approach doesn't qualify. The technical work must involve experimentation to overcome technical uncertainty. Building a better product with known techniques isn't R&D.
"Failed experiments don't count"
False. Failed experiments absolutely count, as long as they were legitimate attempts to solve technical problems. In fact, failed experiments are strong evidence of genuine R&D.
Recent changes affecting software R&D claims
The 2024-2025 program has increased scrutiny on software claims. AusIndustry is looking more carefully at whether claimed activities genuinely involve technical uncertainty.What this means: Document your technical challenges thoroughly. Be honest about what's experimental vs routine. Work with experienced R&D consultants who understand software development.The government wants to support genuine innovation, but they're cracking down on aggressive claims that stretch the definition.
The AI development opportunity
AI integration projects in 2025 often qualify because the technology is still experimental in many applications. You're combining techniques in ways not documented, dealing with accuracy and reliability challenges, and experimenting with novel approaches.Typical qualifying AI activities:
- Fine-tuning models for domain-specific tasks
- Developing novel RAG architectures
- Experimenting with prompt engineering for reliable outputs
- Building agentic systems with decision-making logic
- Creating hybrid AI systems combining multiple modelsNon-qualifying AI activities:
- Using ChatGPT API with standard prompts
- Implementing documented RAG tutorials
- Basic chatbot development with existing frameworks
How to get started
If you think your software development might qualify:Step 1: Document your technical challenges before starting work. What are you trying to achieve that isn't readily documented?Step 2: Engage an R&D tax consultant early in the project. They'll help structure documentation and identify eligible activities.Step 3: Track time and expenditure separately for experimental vs routine work during development.Step 4: Maintain technical documentation showing hypotheses, experiments, and results.Step 5: After project completion, work with your R&D consultant and accountant to prepare the claim.Timeline: Claims are lodged with your annual tax return, within 10 months of year-end. You can claim for the current year and amend up to 4 previous years if you missed eligible activities.
Finding R&D consultants and advisors
Look for consultants with software development expertise. Not all R&D consultants understand software projects well enough to identify qualifying activities.Questions to ask:
- Have you handled software R&D claims before?
- What percentage of your claims are software vs physical R&D?
- Can you help structure documentation during development, not just after?
- What's your success rate with software claims?Good consultants will be conservative about what qualifies and help you document properly. Bad ones will promise high claims and leave you exposed to audit.
What success looks like
A well-documented software R&D claim that survives audit has clear technical challenges documented at project outset, detailed records of experimental activities during development, honest separation of R&D from routine work, and claims that match the technical reality of what was built.The goal isn't to maximize the claim at all costs. It's to capture the legitimate R&D value in your software development work and get the tax benefit the program is designed to provide.
Developing custom software or AI systems that might qualify for R&D claims? We can structure projects with proper documentation from the start.
[Talk to us about R&D-eligible software development]
About ThinkSwift
We're a creative software agency in Melbourne building custom software and AI operating systems for Australian businesses. We work with clients' R&D consultants to ensure qualifying experimental work is properly documented. We don't inflate claims, but we don't leave money on the table either. Our project approach naturally generates the documentation needed for legitimate R&D claims.
