Handling HMRC R&D Investigations: Transparency and AI

Over the last three years, there has been a sharp rise in HMRC R&D investigations.

With tighter scrutiny in place for R&D claims, it’s now essential to clearly demonstrate how your work advances technology. As AI adoption grows, more companies are submitting claims involving AI. That’s positive. But using advanced technology isn’t enough. To succeed, your project must advance it.

Frequently, HMRC will reject such claims on the basis that the project is “using advanced technology, rather than advancing technology”.

R&D investigations: Why are applied AI claims often investigated?

HMRC frequently rejects applied AI claims on the basis that the project:

• Only creates a commercial advantage for the company, not the wider field of technology, or
• Is “using advanced technology, rather than advancing technology”.

Outside of Silicon Valley, there are very few companies who are undertaking fundamental AI research i.e developing their own LLMs. Therefore, the vast majority of AI projects involve integrating with commercial tools provided by one of the large LLM providers. As such, it can be challenging for companies undertaking such AI based projects to evidence how they have made an advance in technology, rather than just using a commercial “black box”.

HMRC ai r&d claims challenges: The importance of framing the technological advance

When submitting AI-based R&D claims, and especially when responding to an HMRC R&D investigation, it is vital to not only describe the work that was undertaken, but to analyse this work within the overall field of AI, to demonstrate the specific advance in knowledge that has been achieved. Just describing the benefits of the AI solution, even if they are considerable in the specific use case, will not be sufficient.  go beyond describing what was built.

The work must be analysed in the context of the wider AI landscape. HMRC needs to understand:

• What technological uncertainty existed
• Why existing solutions could not resolve it
• How the project sought to overcome that challenge
• What advance in knowledge or capability was achieved

Simply outlining the benefits of the AI solution, even if they are significant for the business, will not be enough.

A real-world example: when benefits weren’t enough

We were recently engaged by a client that had developed an AI chatbot designed to make customer service interactions more realistic. The technology was genuinely world-leading. Despite this, their R&D claim had been rejected by HMRC.

The claim had been submitted through a generalist accountant. The client was confident their technology would qualify.

Following an HMRC R&D investigation, there were four rounds of correspondence over a 15-month period. Ultimately, HMRC fully rejected the claim and began considering penalties for an incorrect submission.

Throughout the process, the company and their accountant focused on explaining the benefits of the AI within its application. What was missing was a clear explanation of the underlying technological advance.

This was made harder by the fact that the HMRC inspector had experience using AI tools. As a result, they did not initially appreciate the subtle but important technical differences between this solution and familiar commercial products.

Finding the real advance “under the hood”

Once appointed, our technical specialists reviewed the project in detail. We quickly identified a significant technological advance, one that had not been communicated to HMRC. This advance related to the computational resources required by the AI i.e. an “under the hood advance” rather than a benefit that the end user / client would see.  This was a fundamental, low-level technical improvement rather than a visible feature for end users.

Because this detail had never been shared, HMRC had assessed the project purely on surface-level features and benefits, comparing it to other chatbots already on the market.

Our client had not realised, and had not been advised, that this level of technical depth was essential to support an R&D claim.

How we were able to turn the claim around

We were appointed on the final day available to submit an appeal. We immediately wrote to HMRC requesting a one-week extension to gather further technical evidence.

A meeting then took place with HMRC, the client, and our team. During this discussion, we clearly explained the underlying technological advance and how it differed from existing solutions.

HMRC acknowledged how difficult it can be to assess fast-moving technologies like AI. They also noted that having an adviser with deep technical knowledge, alongside a strong understanding of the R&D scheme, made the process far more effective.

The claim was approved the very next day.

Can applied AI claims succeed?

It is absolutely possible to make successful R&D Tax Relief claims in applied AI. At AAB, we have submitted and defended AI-based claims across a wide range of sectors, including:

• Construction
• Education
• Financial Services
• Healthcare
• Insurance
• Legal Services
• Oil & Gas
• Retail
• Supply Chain Management

The key is getting the technical description right.

Any claim or HMRC R&D investigation response, that focuses only on features and benefits, without clearly explaining the technological advance, is likely to be rejected as “using advanced technology”.

How can AAB help?

At AAB we understand both the technical detail and HMRC’s expectations. That combination matters. It’s often the difference between a rejected claim and a successful outcome.

If you have any queries about submitting a successful R&D claim, or need support with an R&D investigation, please do not hesitate to get in contact with Mark Graves, or your usual AAB contact.