How to make a successful R&D tax relief claim for AI projects

What are your qualifying AI development costs?

An AI project doesn’t just involve writing code or training models – it also requires people, infrastructure, and resources. Qualifying costs under HMRC AI R&D guidelines are inclusive of a wide range of costs directly linked to eligible R&D tax credits for AI. Understanding what you can include is key to maximising your claim.

Which staff expenses qualify for AI development costs?

People are usually the biggest cost in an AI R&D project. You can claim R&D tax credits for AI for the salaries, employer National Insurance contributions, and pension costs of staff directly involved in qualifying activities. This typically covers roles such as:

• Data scientists – designing and testing models
• AI engineers and developers – building algorithms and systems
• Project managers – coordinating the R&D process, where their work relates to technical uncertainty

Subcontractor costs can also be included in R&D tax credits for AI, though they are treated differently depending on whether you use the SME scheme or the RDEC scheme. It’s important to track these costs carefully to ensure compliance.

Related Reading: Claiming R&D Tax Relief for AI Projects: A Complete Guide to What Qualifies.

Which software, data, and cloud computing costs qualify for R&D tax relief?

AI projects rely heavily on specialist tools and infrastructure. You may meet AI R&D tax credit eligibility to claim for:

• Cloud computing services – a proportion of qualifying AI development costs for services like AWS, Azure or Google Cloud, particularly where GPU power is needed for model training
• Software licences – tools used directly for the R&D process, such as coding environments, AI frameworks, or specialist data visualisation packages
• Data sets – purchased or licensed datasets that are essential for training or testing cybersecurity models may qualify as cybersecurity R&D

These costs must be linked directly to R&D activity to qualify for R&D tax credits for AI, not general business use.

Which consumable items and R&D materials qualify for tax relief?

Finally, you can also include certain consumables – items used up during the R&D process. In AI projects, this often means:

• Electricity – particularly relevant where large-scale model training requires intensive computational power
• Raw data acquisition – where data is gathered specifically for training and testing new AI models
• Other consumables – for example, storage media or robotics that are used during experimentation

The key point is that these resources must be consumed in the process of carrying out the qualifying R&D activity.

Even in the early stages, you can claim R&D tax credits for developing MVPs and prototypes.

Step-by-step: how do you make a successful AI R&D claim?

Making a strong claim for R&D tax credits for AI projects requires doing genuine R&D, keeping enough AI project documentation for HMRC to make a case and presenting this clearly to HMRC within deadline.

Ready to turn your AI innovation into a significant tax benefit? Our team has helped businesses like yours secure over £83 million in R&D tax relief. Let us handle the complexity so you can focus on what you do best. Start your risk-free claim with Alexander Clifford today.

Comprehensive documentation is key to meeting HMRC AI R&D guidelines

The foundation of any successful claim is thorough documentation. HMRC AI R&D guidelines outline that you need to show not only what you did, but why it was R&D. This means keeping:

• Technical reports – explaining the challenges, uncertainties, and approaches your team tested
• Financial records – detailing qualifying AI development costs such as staff costs, subcontractor fees, software licences, cloud usage, and consumables linked to the project
• Evidence of uncertainty – showing that the outcome wasn’t known in advance and couldn’t easily be solved using standard methods

The more clearly you can demonstrate these points when claiming for R&D tax relief for AI, the stronger your case will be.

Meeting HMRC AI R&D guidelines – our AI R&D tax credit checklist

To make the process of building a claim for R&D tax credits for AI easier, here’s a quick self-assessment checklist. If you can tick these off, you’re likely on the right track for a valid claim:

1. Do you have evidence of a technological or scientific advance?
2. Can you show the uncertainties your team faced and how you tried to resolve them?
3. Have you kept records of experiments, iterations, or failures that demonstrate the R&D process?
4. Do you have detailed cost breakdowns of the people, tools, and resources directly involved?
5. Is your project focused on solving problems that go beyond routine development or application?

Have a question about your AI project? Our specialists are here to help.

FAQs: your most pressing questions about AI R&D tax relief for data science answered

 

What AI project documentation for HMRC do I need to support my AI R&D claim?

HMRC expects to see clear evidence of your work when claiming R&D tax credits for AI. This includes technical reports, records of experiments or iterations, and financial documentation showing the qualifying AI development costs directly linked to the project.

Can I claim for staff who only spend a portion of their time on the R&D project?

Yes. You can apportion staff qualifying AI development costs based on the percentage of time they spend on R&D activities when claiming R&D tax credits for AI. Even if someone only spends a few hours on innovative AI work, the relevant portion of their salary, pension contributions, and employer NICs can still be included in your claim.

How is a ‘competent professional’ defined in the context of AI?

A competent professional may be consulted to identify what qualifies for R&D tax relief (AI). This will be someone in the field of AI with the relevant skills, qualifications, and experience to judge whether your AI project involved scientific or technological uncertainty. This could be a data scientist, machine learning engineer, or AI researcher with experience in the field.