Can Machine Learning Qualify for R&D Tax Relief?

When it comes to machine learning projects, there is often a perception that these will automatically qualify for R&D tax relief because they are innovative. However, while certain projects and use of machine learning might qualify as R&D under some standards when it comes to R&D tax relief there are specific requirements that must be met. While machine learning certainly can qualify for R&D tax relief, the reality is that not all machine learning projects will meet the eligibility criteria.

When does machine learning not meet the criteria?

Here are a couple of examples of where the assumption that something innovative like machine learning is automatically R&D that will be eligible could be wrong.

• Machine learning is something that is new to your specific business and, as a result, there has been a lot of research into it so that your company can use it. The criteria for R&D tax credits is that the advance is being made in the overall field and is not limited to an individual company’s state of knowledge or capability. And, as a result, this would not qualify.
• Machine learning is new to the field or domain in which your business operates. This is a little more complex but, essentially, if the reason for this is that no one has thought to use machine learning in this field before it’s unlikely that the project would qualify for R&D tax relief. On the other hand, if the reason was because the right format of data wasn’t available and the business had developed new methods of data acquisition or pre-filtering then this could potentially be eligible.

Can machine learning qualify for R&D tax relief?

One of the major issues with machine learning is that there are very few situations where a brand-new algorithm is being developed. It is often more the case that proprietary company data is being fed into a commercially available third-party machine learning library. This would raise the question of where the field of technology is being advanced that would make the activity eligible for R&D tax credits. Fine-tuning algorithms is also a grey area – recent guidance noted that this would only count where a genuine and non-trivial improvement has been made in the underlying technology. This tends to be a complex field where there is no simple answer – but there are projects that have received R&D tax relief for machine learning technology so there are good reasons to make a claim. Crucial to this will be relying on more than just assertions that the innovation is genuine and meets the criteria for R&D tax relief. Detailed contemporaneous data will be necessary about the project being developed and all the steps involved in developing and analysing an algorithm will need to be documented in detail in order to increase the chances of success.

There is no doubt that machine learning can qualify for R&D tax relief – the right support could make this happen for your company.