As AI investment accelerates across the North East thanks to the region’s AI Growth Zone vision, there’s growing confusion around what actually counts as R&D. Rebecca McColl, senior tax manager at Johnston Carmichael, looks at why investing in AI doesn’t automatically qualify for tax relief, while some tech teams could be overlooking genuinely innovative work that does.
We know that businesses across almost every sector are investing in AI, whether that’s automating processes, interrogating data, improving customer experience or developing entirely new products and services. Here in the North East, that conversation is only going to become louder as the region builds on its ambitions as an AI Growth Zone.
For those investing in that innovation, understanding where genuine R&D is taking place is really important from a technical perspective, but also because it can determine whether that investment could be eligible for valuable tax relief, and whether it is identified early enough to make a successful claim.
One of the biggest challenges I see is businesses assuming that because something is new or innovative for them, it automatically counts as an advance in science or technology, but the two are actually very different.
A company might invest in an AI-enabled platform that revolutionises the way they operate, but that does not automatically mean the investment is eligible for the tax relief.
Businesses can spend huge amounts on technology that is commercially significant without it meeting the criteria for qualifying R&D. At the same time, some of the most sophisticated firms I work with are doing really impressive work and barely recognise it as unusual. To their developers, solving difficult problems is just what they do every day. That is why we need to get much better at looking beyond the label.
‘Bespoke’ is not a magic word
If a business says it has commissioned a bespoke AI system, I firstly want to understand what was happening from a technical perspective. I want to know if there was an existing solution, what the developers did to overcome it, if the answer was readily available to a competent professional or if they actually had to work out something that wasn’t already known.
Taking an off-the-shelf AI product and configuring it around the needs of your company may be very clever and commercially valuable, but using current functionality in a different context does not mean it is R&D. The same applies to buying an AI tool and using it to automate administrative tasks. You might make a significant productivity gain, but you are still applying technology that somebody else has created.
Start with the industry baseline
The underlying R&D criteria have not changed, but claims are under more scrutiny. I have seen potentially strong projects look weak on paper simply because the right questions were never asked of the specialist team. A list of buzzwords about AI or LLMs is not enough – the claim needs to explain what the team was trying to achieve and why it involved genuine uncertainty.
A good starting point is the industry baseline – what could be accomplished using what was already there? There may not be an identical product in your own company or local market, but there will often be something comparable. The key is to identify where those solutions stopped short, then show what your team was trying to achieve beyond that and how they approached it.
For example, using an LLM does not make a project R&D but if a team is trying to make it work in a specialised environment where standard approaches cannot deliver what is needed, and has to overcome real difficulties in order for it to work, that becomes a very different discussion. The same applies in cybersecurity, where using authentication technology is one thing, but developing ways to significantly advance its speed or capability can involve much more complex work.
Innovation can be easy to overlook
There is a strange dynamic I have noticed when speaking to specialist teams. At one end, you have those that have undertaken fairly routine configuration or software development but describe it as groundbreaking because it has transformed their own company.
At the other, you have highly experienced software engineers designing sophisticated systems or solving complex issues who will tell you they were just problem-solving.
There is also a misconception that R&D has to mean something enormous or world-changing, which it doesn’t. Of course, there are businesses carrying out major scientific research, but particularly in software, it is often not the whole project that qualifies at all. The R&D can sit in small pockets of technically difficult work within a much larger commercial development. What matters is being able to clearly identify what the advance was and why getting it was challenging.
Don’t wait until somebody mentions R&D at year-end
For those that need to make a claim notification, including those claiming for the first time, HMRC requires that notification within six months of the end of the relevant period of account. Missing that deadline can mean losing the ability to make the claim.
We still speak to companies that have been carrying out potentially qualifying activity but approach the situation too late. R&D should form part of the wider tax planning, rather than something considered once the work is finished.
That means identifying potential R&D activity early and understanding where the challenges sit. Businesses should be able to explain what was already available, what they were trying to achieve beyond that and why the solution was not straightforward. It is also worth looking at the level of investment involved, because even where work meets the criteria, the qualifying spend needs to make a claim commercially worthwhile.
Importantly, the people closest to the work need to be part of that process. A finance director can tell you what a project cost, but the developer or engineer can explain what made it challenging. You need both sides of the story to properly assess the opportunity.
The North East opportunity
As the North East continues to build its AI capabilities, we should expect to see more experimentation, more investment in new products and finding new ways of solving problems.
Some of that work will involve exciting advances in technology and some of it will involve those using brilliant systems developed elsewhere to become faster and more productive. They are both worth celebrating, they are just not the same thing from an R&D tax perspective.
The reality is, businesses need to know whether their technical teams are trying to achieve something that couldn’t easily be achieved before. Strip away the AI terminology and buzzwords, and that is where the real R&D conversation begins.
