Sometimes a policy is so fundamentally flawed that resurrecting it feels like the plot of a tired sitcom. HMRC’s current consultation on reforming the Advance Assuance/Clearance process for R&D tax relief is precisely such an episode. The original scheme failed. Everyone knows it. The attempt to breathe new life into it now feels less like strategic reform and more like watching someone try to resuscitate a donkey that was declared dead years ago.

Let us be blunt: the Advance Assurance process has never delivered meaningful certainty to claimants. It was sold as a way to give small companies comfort that their claims would be accepted, but in practice, was more about HMRC discouraging.
So here we are again, with HMRC consulting on how to repackage the same flawed process. The response from experienced agents has been consistently sceptical, if not openly cynical. Why? Because the problems that plagued the scheme the first time around have not been fixed. HMRC’s inconsistent interpretations, the lack of technical expertise in caseworkers, and the culture of volume compliance all persist. A newly dressed version of the same process will not magically fix these deeper issues.
If anything, the current consultation risks missing the point entirely. Advance Clearance is not a bad idea in theory. It is just the wrong idea for the current climate. What companies need is a smarter, scalable, and reliable way to assess risk and provide guidance without wasting time or resources.
The answer is not more paperwork. It is smarter systems. And that is where artificial intelligence deserves serious attention.
As an adviser who has worked on artificial intelligence R&D claims for over a decade, I have come to believe that we are now at an inflection point. Ai is no longer just an emerging field; it is becoming the infrastructure of the future. More importantly, it can do things today that neither agents nor HMRC can do efficiently or impartially at scale.
Imagine a system where, before an R&D claim is filed, a risk assessment ai model reads the Additional Information Form, scores it for compliance risk, and provides notes and suggestions to the claimant. This system could classify claims as low, medium, or high risk. Claimants would then have the choice to revise, abandon, or proceed with their claim, informed by an impartial, data-driven assessment.
Such a tool could serve both claimants and HMRC. It would streamline compliance, reduce the number of borderline or fraudulent claims, and free up human resource to focus on complex or high-risk cases. Unlike human reviewers, the AI would be consistent, objective, and scalable. And it would be fast. Damn, I sound like Elon Musk.
Some might game it. Ai can certainly be abused. But most people are fundamentally honest and want to get tax right. We must stop building systems that damage 99% of users to try and deal with a few bad apples.
HMRC must already know this is the future. The notion that a 21st century compliance regime should hinge on a labour-intensive assurance process that takes months to issue and offers no binding comfort is absurd. Rather than rebranding a broken system, we should be investing in tools that transform the process entirely.
Advance Assurance is not just unfit for purpose; it is a distraction from the work that actually needs to be done. If HMRC is serious about reducing error and fraud, it should be focusing its resources on better guidance, smarter targeting, and embracing technologies that can do the heavy lifting in compliance checks.
In short, drop the dead donkey. Let us not pretend that this latest consultation will breathe life into something fundamentally unsuited to the challenges we face. The solution is not a reheat. It is a redesign.
If you are navigating the complexities of R&D tax relief, or want to understand how future compliance might really work, we can help.
Christopher Toms MA MAAT
Compliance Director
RandDTax
Photo by Igor Omilaev on Unsplash