Research and Development (R&D) tax relief remains one of the most valuable incentives available to companies investing in innovation. For businesses creating new and exciting ideas, it can provide a substantial cash benefit or tax reduction, helping to fund further growth and technological progress.
However, before any numbers are discussed or calculations made, every potential claimant must first clear a critical hurdle: does the work actually qualify as R&D?
It does not matter how much a company has spent, how technically impressive the outcome may be, or how commercially successful the product has or could become. Unless the activity meets HMRC’s definition of qualifying R&D, the expenditure simply will not obtain relief.
Is there an advance in science or technology?
The starting point for any R&D claim should be a simple but fundamental question:
“Is there an advance in science or technology?”
This question often determines whether an idea moves from a good project to a genuine R&D activity. HMRC’s definition requires that a company seeks to achieve an advance in the overall field of science or technology, not merely an improvement to its own knowledge, processes, or systems.
This distinction is crucial. Many claims come unstuck because, while a company’s project might be clever, complex, or novel, the technical advance only benefits the company itself. It does not push the boundaries of what is possible for the wider field.
For example, building an internal platform that automates tasks or integrates existing tools in a new way may be innovative within your business, but unless it resolves a scientific or technological uncertainty that others in the field could not easily overcome, it is unlikely to meet HMRC’s threshold for R&D.
Opportunity and misunderstanding
This issue has become particularly relevant with the rapid rise of Artificial Intelligence (AI). AI is everywhere, from chatbots and data analysis to design tools and automated software testing. It has quickly become part of the everyday toolkit for businesses of all sizes and sectors.
Unsurprisingly, we are now seeing a surge of enquiries from clients who are using AI in new ways and believe they may have carried out qualifying R&D. There is no doubt that many of these projects are creative, technically impressive, and commercially viable. However, the question remains: do they truly involve R&D?
Using AI is not automatically R&D
A common misunderstanding arises from the assumption that simply using AI to produce a new or improved product automatically qualifies as R&D. In reality, using AI is not the same as developing AI.
Many businesses using AI today are not creating new machine learning models or advancing the underlying algorithms that drive the technology. Instead, they are using existing tools developed by others, often from large technology companies, and applying them in innovative ways.
For example, a business might use OpenAI’s GPT models, Google’s Vertex AI, or Amazon’s SageMaker to generate reports, automate customer service, or predict maintenance needs. These uses can certainly transform a company’s efficiency or output, but they do not represent an advance in the field of computer science or AI itself.
In other words:
• The AI tool was created by someone else.
• Using that tool to do something new or time-saving alone does not count as R&D.
• To qualify, you would need to be working at the level of the underlying technology. For example, improving the model’s architecture, developing a new algorithm, or solving a genuine technical uncertainty about how the AI operates.
Building on existing work — Know where the line lies
That is not to say that valid R&D cannot involve AI at all. It absolutely can. There are legitimate R&D projects in which businesses take an existing AI model and extend or modify it in ways that solve genuine technical uncertainties.
For instance, if a company develops a new training approach that allows a model to operate in a previously unachievable way, such as interpreting complex non-linguistic data or improving bias control beyond current capabilities, that could qualify.
The key is being extremely clear about where someone else’s work ends and where yours begins. If your team is experimenting with new methods, architectures, or data processes to overcome genuine unknowns, you may well have an R&D claim. If you are simply prompting, configuring, or deploying someone else’s system for a particular use case, that would be product development, not R&D.
When does R&D stop?
Another important point is knowing when the R&D actually stops. R&D continues only until the underlying scientific or technological uncertainty is resolved. Once you have overcome that uncertainty – that is, once you know how to achieve the desired result – any further development, refinement, or commercialisation usually falls into product development rather than R&D.
For example, once a technical challenge has been solved, improving the interface, testing for users, or integrating the system into other platforms does not necessarily qualify. It might be complex work, but the scientific or technological uncertainty is already resolved.
Increased HMRC scrutiny
In recent years, HMRC has significantly increased its scrutiny of R&D tax credit claims. This is especially true in areas where definitions can become blurred such as software, data analysis, and now AI.
We are seeing HMRC request more detailed technical justifications and supporting evidence than ever before. They want to see clear documentation of the uncertainties faced, the hypotheses tested, and the methods used to overcome them. Simply stating that something was “innovative” or “complex” is not enough, with such stock phrases actively dismissed by HMRC.
Submitting a claim that does not meet the qualifying criteria can lead to rejected claims, clawbacks, penalties, and potential reputational damage. It is therefore vital to be cautious and ensure that claims are well-founded before submission.
Seek professional advice early
If you are unsure whether your project qualifies, it is always best to seek professional advice before making a claim. AI is an incredible tool one that is transforming how we work and what we can achieve. We must remember, however, that the work to create AI was someone else’s R&D. Using it in new and exciting ways does not automatically make your project a qualifying R&D activity.
At Hillier Hopkins, we regularly review potential R&D projects and help clients assess whether their work meets the definition of a qualifying activity. We continue to support innovative companies in making robust, defensible R&D claims, and as AI becomes more integrated into everyday business, we are encouraging clients to take a cautious, but well-informed approach, giving you greater confidence in your claim.
