R&D Tax claims for AI companies: Is it time to rethink claims?

For many scale-up technology businesses, R&D tax relief remains one of the most significant sources of non-dilutive funding available. As companies grow, however, the challenge is no longer simply identifying innovation. It is ensuring claims remain robust, defensible and aligned with increasingly complex commercial activities.

For many AI companies, preparing an R&D tax claim is relatively straightforward in the early stages of growth. The business is typically focused on building a core technology platform, most engineering work is directed towards overcoming technological challenges, and there is often a close alignment between commercial objectives and technological advancement.

However, as AI / Software businesses scale, the nature of both the business and the R&D claim changes significantly.

Companies begin working with larger enterprise customers, delivering strategic partnerships, developing industry-specific solutions, operating internationally, and balancing core platform development against commercial delivery. At the same time, many are preparing claims under the UK’s Merged R&D Scheme for the first time.

Whilst much of the discussion around the Merged Scheme has focused on changes to tax relief rates, the bigger issue for many scale-up technology businesses is that the framework places greater emphasis on understanding who is undertaking the R&D, why it is being undertaken, and where qualifying R&D sits within increasingly complex customer engagements.

As a result, many of the assumptions that worked for an early-stage technology company no longer apply.

Does the nature of innovation change as companies scale?

Early-stage AI and Software companies are often building technology before substantial customer demand exists. In these circumstances, a large proportion of engineering effort may naturally sit within qualifying R&D.

As the business grows, however, engineering resources become divided between core platform innovation, customer-driven enhancements, enterprise integrations/deployment and commercial implementation projects.

The challenge therefore shifts from identifying R&D to correctly defining its boundaries.

Customer Partnerships Create New Challenges for R&D claims

One of the most significant changes for scale-up AI / Software companies is the increasing importance of customer-led innovation.

Many successful AI / Software companies work closely with enterprise customers to solve complex operational problems. These engagements frequently involve substantial engineering effort and may result in entirely new technical capabilities being developed.

Under the Merged Scheme, however, identifying qualifying R&D is only part of the analysis. Where development takes place within a customer engagement, it is also necessary to determine which party is entitled to claim for that R&D. This requires consideration of who initiated the R&D and whether, when entering into the contract, the customer intended or contemplated that R&D of that sort would be undertaken.

A project can be technologically sophisticated, resource intensive and strategically important to the customer, yet still contains relatively little qualifying R&D. Equally, the fact that development is customer-driven does not, by itself, determine which party is entitled to claim. A customer may specify a commercial outcome without intending or contemplating the R&D ultimately undertaken by the technology company to achieve it.

The strongest scale-up claims are typically those that distinguish clearly between customer delivery activities and the underlying technological challenges that required genuine R&D.

Are hybrid projects becoming increasingly more common?

In practice, most scale-up AI / Software projects are neither pure product development nor pure customer delivery.

Rather, they sit somewhere between the two.

A customer-led engagement may reveal challenges such as:

• Processing data volumes beyond existing platform capabilities.
• Achieving explainability levels not achievable using existing methodologies.
• Meeting extreme latency or scalability requirements.
• Integrating multiple AI systems in ways that are not readily deducible to competent professionals in the field.
• Extending agentic or autonomous systems beyond known capabilities.

In these situations, qualifying R&D often exists within a wider commercial project.

The challenge for claim preparation is identifying precisely which activities sought to resolve the technological uncertainties and which activities were simply concerned with deployment, implementation or commercial delivery.

This distinction becomes increasingly important as businesses grow.

The AI Baseline Continues to Move

As AI technologies rapidly mature, companies need to distinguish genuine advances in overall technological knowledge or capability from the application of established technologies and methodologies, including:

• Standard LLM integrations.
• Conventional Retrieval Augmented Generation (RAG).
• Off-the-shelf agent frameworks.
• Typical machine learning workflows.

This becomes increasingly important as businesses grow and claim values increase.

Activities that represented genuine advances several years ago may now be achievable through:

• Established foundation models.
• Off-the-shelf AI tooling.
• Commercial agent frameworks.
• Standard RAG solutions.
• Existing orchestration platforms.

As a result, AI companies need increasingly robust technical narratives, particularly around the technological baseline, supported where appropriate by dated evidence demonstrating the known capabilities and limitations of the relevant technology at the outset of the project.

Blanket statements of this nature are increasingly attracting HMRC scrutiny and may not provide sufficient evidence to support a claim.

A strong claim now needs to explain not simply what was developed, but why the desired outcome could not have been achieved using existing technologies, frameworks or established methodologies.

The focus must remain on the underlying technological advancement rather than the commercial functionality delivered.

Does the merged scheme make these issues more important?

Many of these challenges existed before the introduction of the Merged R&D Scheme (for accounting periods commencing from 01 April 2024). However, the Merged Scheme has made them more significant for growing technology businesses.

Historically, many software companies approached claims primarily by identifying qualifying staff costs and then building supporting narratives around those activities.

Increasingly, companies need to demonstrate a clearer understanding of:

• The precise boundaries of qualifying projects.
• Who initiated the R&D and, in customer engagements, whether the customer intended or contemplated that R&D of that sort would be undertaken.
• The relationship between customer contracts and R&D activities.
• Whether work was undertaken to resolve technological uncertainties.
• Whether activities formed part of wider commercial delivery programmes.
• The extent to which engineering effort related to platform advancement versus customer implementation.

Consequently, project boundary analysis has become one of the most important aspects of claim preparation for scale-up businesses.

Client and Subcontractor contracts and commercial arrangements are critical

As businesses mature, formal commercial agreements become increasingly common.

These can include:

• Enterprise customer contracts.
• Master Service Agreements.
• Statements of Work.
• Joint development agreements.
• Strategic technology partnerships.

Whilst technical evidence remains the cornerstone of any claim, contractual arrangements can also be critical in determining which party is entitled to claim R&D relief. Under the Merged Scheme, where R&D is undertaken as part of a customer engagement, it is necessary to consider whether the customer intended or contemplated that R&D of that sort would be undertaken when entering into the contract.

The written contract is important evidence, but it is not necessarily determinative. The wider surrounding circumstances must also be considered, including how the project arose, who identified the need for R&D, the parties’ respective decision-making roles, financial risk, autonomy over how the work was undertaken, IP arrangements and how the resulting technology will be exploited.

Increasingly HMRC will request to see these documents if they investigate a claim. A thorough R&D advisor should review these during the preparation of the claim to ensure they support the R&D claim being made.

We often review contractual arrangements alongside technical activities to help clients understand the potential R&D tax implications and ensure the commercial reality of arrangements is appropriately documented.

Operational Scale Creates Additional Complexity

As AI companies grow, they often experience significant increases in:

• Cloud computing expenditure.
• Data processing costs.
• Model training infrastructure.
• Annotation and labelling activities.
• International teams and specialist contractors.

Whilst these areas can remain important sources of qualifying expenditure, they also require increasingly sophisticated methodologies to distinguish R&D activities from production, operational or customer-serving activities.

This is particularly true where companies operate a mixture of:

• Experimental environments.
• Product development environments.
• Customer-facing production systems.

The larger a business becomes, the more important these distinctions become.

Conclusions

As AI / Software businesses scale, the challenge is perhaps less about identifying whether innovation exists. The challenge is more about demonstrating where qualifying R&D sits within increasingly complex customer engagements, commercial arrangements and delivery models. Successfully navigating this transition requires both deep technical understanding and a clear appreciation of the commercial environment in which those innovations are delivered.

For AI / Software companies, the transition from start-up to scale-up is often accompanied by two important changes: customer engagement becomes a central driver of development activity, and R&D claims become significantly more complex. Care must be taken with contractual wording as this could otherwise lead to a loss of R&D tax relief eligibility on customer-driven projects and/or related R&D expenditure.

Many reputable advisory firms are highly effective at helping early-stage technology businesses prepare their first R&D claim. However, the challenges faced by an early-stage start-up with minimal revenue are fundamentally different from those faced by a rapidly scaling company delivering complex solutions in partnership with enterprise customers and partners, in many cases outside of the UK.

At AAB, much of our experience lies in helping businesses navigate this transition precisely, combining a holistic team approach with PhD / MEng / MSc computer scientists to analyse the technology development and Chartered Tax Advisers / Chartered Accountants to review the commercial relationships for R&D claim eligibility.

If you have any queries about your R&D claim, please don’t hesitate to get in contact with Mark Graves or your usual AAB contact.