Reflecting on how HMRC went about tightening the reins of compliance in the R&D tax relief space, the good and the ugly.

Last month, I've been reflecting on how HMRC went about tightening the reins of compliance in the R&D tax relief space. The intention? Fair. The execution? Questionable.
The crackdown felt less like a structured, proportionate regulatory effort and more like reactive damage control. Uncertainty, delays, disruptions and loss of revenue followed. It was hard to tell whether HMRC's goal was to reform the system or reset it entirely; whatever it was, it was chaotic.
As poor compliance was what got us here, I wanted to sit down with Richard Edwards to talk about the state of compliance in our space and what could be done to elevate industry standards. Below is the summary of my posts from April, some of your insightful comments, and a bit more information that I found interesting.
Once upon a time, R&D claims were mostly driven by genuine technical expertise. Some consultants stayed well within the lines, others tested the edges, and some coloured way outside the borders.
Initially, making a bold move not to check many claims, HMRC watched passively from afar.
Until one day, like Sauron, it turned its gaze inwards.
After letting the market balloon with non-compliance, HMRC attempted to establish order in one massive, messy sweep or as Richard put it, "one great bun fight at the end of 20 years".
The best analogy I can come up with is being severely arachnophobic and finding a spider on your curtains; instead of overreacting and trying to find suitable means of disposal, you torch the place. Good riddance, Spidey...
🕷🔥
Or as Nigel Holmes put it recently, "sledgehammer to crack a nut".
After the Wild West era, what followed was a rise in aggressive enquiries into many R&D tax relief claims by HMRC. LinkedIn erupted with R&D consultants sharing notes and concerns about the inconsistencies in handling compliance across HMRC's different divisions. So much so that people were able to ascertain and categorise each division. Here's a simplified breakdown that Richard offered me, but echoed by many others:
If you'd like to read a more detailed breakdown of these divisions, here's Rufus Meakin's article, which goes into more detail.
Unsurprisingly, one of the most debated topics I touched on last month was how vocal we should be on LinkedIn about fraud within our industry.
"It's an unusual space to be so vocal," said Richard, "and I think part of it is the lack of regulation and standards within this space". Unlike other areas of tax, this industry has developed a loud, public-facing commentary culture. However, this has resulted in mixed feelings amongst fellow R&D professionals.
Some, like Derek Granger, cautioned that publicly calling out companies on LinkedIn could introduce unconscious bias or damage the reputation of companies that may not have had malicious intent, a point that ties in directly with Richard's remarks about the lack of defined standards. Others have commented that LinkedIn isn't the right channel for these sorts of discussions and that it fosters a negative environment on the platform.
On the other side of the debate, Rufus Meakin, Paul Rosser and Mark Evans https://www.gov.uk/government/consultations/enhancing-hmrcs-ability-to-tackle-tax-advisers-facilitating-non-compliance argue for a firmer approach. Rufus even compared ignoring it to "brushing the problems under the carpet". In their view, vocal accountability is necessary in a space where regulation is still playing catch-up.
This whole debate reminded me of a recent story from Japan. When the toll systems went down, the government didn't need to deploy extra police or other bullish tactics, instead, they opened the barriers and asked people to pay online later. And guess what, they did!
Can you see the same thing happening here?
This highlights another thing Rufus said, i.e. that it's "frustrating seeing firms get away with flagrant abuse costing taxpayers tens of millions of pounds", and as a fellow taxpayer, I'm not very chipper about this either. But this brings me onto my next question...
R&D tax relief is a remarkably intricate area. It's not solely accountants and tax professionals who engage in it; the field also attracts engineers, scientists, pharmacists, and other specialists who grasp the technical aspects of innovation just as well as the financial ones. This diversity is a significant asset, but I also question whether it contributes to the challenges of regulation.
Most would agree that HMRC should have acted sooner, not by slashing benefits and introducing inconsistency into the enquiry process, but by stepping in earlier to regulate the scheme properly. For years, it felt like HMRC sat on the sidelines, although now they seem to be moving in the right direction.
But wait, what is the right direction?
Should HMRC be the sole authority? Should advisors need formal qualifications? Or do we need to develop something entirely different?
Right now, there's an open consultation where HMRC is seeking feedback on how best to deal with tax advisors who facilitate non-compliance. Unfortunately, it closes on 7 May, giving little time for a response if you haven't yet done so. Many may have noted that the consultation period was only 6 weeks, instead of 12, and the poor timing around Easter and bank holidays. However, Chris Irwin has made effort to explain the reason behind this timeline.
Karen Evans FCCA also recently explored these questions in a thought-provoking article, breaking down several regulatory models being considered. Each comes with its potential strengths and trade-offs.
I won't go too deep into AI here, as I've discussed it at length before, but one angle we haven't explored much is what comes next, say, 5 years from now?
Richard made an interesting point. As more AI tools are used to prepare claims, HMRC may build systems that detect the use of AI in claim preparation to flag them for more attention to make "sure it's a fair representation of what the client's actually done". Alternatively, as AI-generated claims become harder to tell apart from human work, HMRC may try to cut through the noise by speaking directly to the professionals behind the claims.
It's a compelling thought. Although if both sides, namely consultants and HMRC, are using AI, we could easily end up in a loop where machines are interpreting machines, never fully grasping the actual context, or something completely different that we haven't realised yet.
We often say "AI is just a tool", just like Novel is a tool for preparing R&D tax relief claims. In both cases, the message is the same: these tools can make mistakes and are meant to be used by specialists who know how to spot and correct them. However, in the AI-loop described above, who is truly accountable when something goes wrong? It's easy to say - the claimant. But what if it's HMRC's AI that makes the mistake? Will there be a clear, transparent way to challenge decisions?
As we move further into this AI-integrated future, we need to think not just about functionality, but about traceability, aka building a breadcrumb trail, like Hansel and Gretel, that allows us to understand how decisions were made, and by whom (or what?).
The irony is that the solution to this is turning back to human interaction, which brings us full circle because we would need a large pool of highly competent HMRC inspectors to have meaningful conversations at scale. So we risk facing the same challenge, just from a different angle.
Perhaps the solution lies not in substituting one system for another, but in finding the right balance among speed, scrutiny, and trust. Regardless of whether a claim originates from AI or a human, someone must understand what's being claimed (its validity). And that's the part we can't afford or really should automate away.