This is a companion blog to my earlier blog “Uncovering Hidden Value: Example Areas for Software R&D Tax Claims“.
When making an R&D tax relief claim for software development, identifying technological uncertainties is crucial. HMRC is not interested in routine software development—they want to see attempts to overcome technological uncertainties that could not be resolved with existing knowledge or readily available solutions.
But what does this look like in practice? Below are some real-world examples of technological uncertainties in software R&D, where companies have pushed the boundaries of what was previously possible.
1. Bridging the Gap Between Legacy and Modern Systems
Many businesses rely on legacy systems-built decades ago, yet modern applications require advanced AI and cloud computing integration. The challenge? Ensuring these outdated systems can effectively communicate with cutting-edge AI decision-making processes without breaking or causing errors.
Uncertainty: How can AI be integrated into legacy systems without disrupting their core functionalities?
2. Scaling Distributed Applications Without Bottlenecks
Cloud-based applications must handle unpredictable workloads, but traditional scalability methods often lead to database congestion or excessive latency.
Uncertainty: How can a distributed system scale dynamically without overwhelming databases or causing performance degradation?
3. Making AI Decisions Explainable and Ethical
AI models, particularly deep learning systems, often function as “black boxes,” meaning their decision-making process is unclear. Regulatory pressures (e.g., GDPR, EU AI Act) require businesses to demonstrate transparency in automated decisions.
Uncertainty: How can AI-driven decisions be audited and made interpretable for regulatory compliance?
4. Securing Data in Federated Learning Environments
Federated learning enables AI training across multiple devices while keeping data decentralised—a necessity for privacy compliance. However, this creates vulnerabilities where malicious actors could extract private information from AI model updates.
Uncertainty: How can federated learning models be trained without exposing sensitive data?
5. Enhancing Cybersecurity Against AI-Powered Attacks
Cyber threats evolve rapidly, and attackers are now leveraging AI to breach security systems. Traditional methods of cybersecurity are struggling to keep pace.
Uncertainty: How can security systems proactively defend against AI-enhanced cyber threats, including zero-day vulnerabilities?
6. Overcoming Performance Limitations in Blockchain Networks
Decentralised applications are gaining traction, but blockchain scalability and energy consumption remain major hurdles. Transactions on large blockchain networks often suffer from slow processing times and high fees.
Uncertainty: How can blockchain networks scale effectively while maintaining security and reducing energy consumption?
Why This Matters for R&D Tax Relief
Each of these examples represents a technological uncertainty—a key requirement for R&D tax relief. If your software project faced a challenge that couldn’t be resolved using standard knowledge or tools, it could qualify as R&D.
Remember, it’s not enough to simply use new technology; the real question is whether you had to develop or significantly improve a solution due to technological limitations.
If you’re unsure whether your project qualifies, get in touch. Our expert team can help assess your R&D claim and ensure you maximise your tax relief.
Christopher Toms MA MAAT
Compliance Director, RandDTax