R&D tax relief •12 min read

AHK Recruitment Ltd v HMRC

MJ

Matthew Jones ACA CTA

Last reviewed September 2026 · Editorial standards

Tribunal
First-tier Tribunal (Tax Chamber)
Decided
20 May 2020
Judge
Judge David Bedenham and John Adrain

HMRC won. A recruitment company claimed R&D relief for work on a machine-learning system to assess candidates for job roles. Its adviser wrote the claim, its adviser gave the evidence, and nobody who had done the work appeared. The tribunal did not doubt anyone’s honesty; it simply had no evidence it could use, and said twice that it found the absence remarkable. It remains the clearest statement of what a competent professional must be able to do at a hearing.

What was at issue

AHK Recruitment Limited provided human resource services and systems to third parties, including recruitment (paragraph 5). For the periods ending 31 December 2014 and 31 December 2015 it claimed relief under section 1044 CTA 2009 and, on that basis, R&D tax credits (paragraph 6). The claim was suggested, and put together, by Optimal Compliance, a consultancy to the company (paragraph 4).

The costs claimed, as the company’s skeleton set them out, were in-house staff costs of £25,191 for 2014 and £27,004 for 2015, and external workers — software developers — of £141,201 and £84,332 (paragraph 57(9)); the decision gives no deduction or credit figure. Much of the expenditure related to one subcontractor, Evensys Technologies Ltd (paragraphs 27 and 31).

The project was described first as “Individual Behaviour Assessment” and later, in a revised report, as “Individual Behaviour Assessment Automation” (paragraphs 7 and 27). The question was whether the company had proved that it undertook R&D, and that the costs claimed related to it.

How the dispute got to a hearing

On 4 March 2016 an HMRC officer opened a check into the 2014 return under paragraph 24(1) of Schedule 18 to the Finance Act 1998, asking at once how the project qualified given that the Guidelines exclude the arts, humanities and social sciences (paragraph 8). On 13 April 2016 he asked the five questions that recur in every R&D enquiry since. He asked who decided the company had done qualifying R&D; who the competent professional was and what qualified them; and what advance was sought and in which field. He also asked what uncertainties were encountered during the project, and for a cost breakdown by heading (paragraph 10).

What followed was eighteen months of extensions. A letter naming the company’s competent professional arrived in August 2016 but without the project description, and further extensions ran through to the end of the year (paragraphs 11 to 19). At a meeting on 5 May 2017 it was agreed the competent professional would report by 9 June (paragraph 25); the revised report arrived on 4 September, with an apology for the delay (paragraph 26). A second check, into the 2015 return, opened on 3 October 2017 (paragraph 29).

Closure notices for both periods were issued on 1 October 2018, refusing the claims because “there has not been any evidence provided by the company to suggest that they have advanced the technological field of IT” (paragraph 38). HMRC recorded expressly that it accepted the claim had been made in good faith (paragraph 28). The notice of appeal was filed on 30 October 2018. The appeal was heard at Central City Tower, Birmingham on 29 November 2019 before Judge David Bedenham and John Adrain, and released 20 May 2020.

What each side argued

HMRC’s case had six limbs (paragraph 64). The company had proved neither an advance in science or technology nor a scientific or technological uncertainty. It had shown no “appreciable improvement…through scientific or technological changes”, appearing instead to have applied existing technology to its psychometric model. The person said to be the competent professional had explained nothing and had not been proved to be one. And the quantum was unproved, the subcontractor evidence being inadequate. The officer accepted that the company had sought to build an AI system making recruitment decisions to a human standard, but had seen no evidence of new code (paragraph 65).

The company’s case was technically specific. Its skeleton described an attempt to build a binary classifier with a convolutional neural network architecture, using deep learning and backpropagation, to predict a candidate’s suitability for a role. System architecture was said to be the hard part, because the company had nothing like the training data such networks normally need. The field was young, the first practical application dating from 2011 (paragraph 57).

The difficulty was who said it. Both witnesses came from Optimal Compliance. One had graduated in 2017 and joined the firm in September 2018, drawing his knowledge from documents and from people who had previously worked at the company. Asked why the reports sent to HMRC had never mentioned convolutional neural networks, he answered that it was “probably because my predecessors did not understand the project properly” (paragraph 60). He said the company’s competent professional was “competent through experience” (paragraph 60). The other had known the company since 2010 but was not contemporaneously involved; the competent professional had left, the founder was “time poor”, and the people who did the work were employed by the subcontractor (paragraph 61).

The provision the case turns on

Nothing in Part 13 defines R&D directly. Section 1041 CTA 2009 gives the term “the meaning given by section 1138 of CTA 2010”, and section 1138(2) CTA 2010 provides that “‘Research and development’ means activities that fall to be treated as research and development in accordance with generally accepted accounting practice.” Section 1006 ITA 2007 then lets the Treasury specify by regulations what is and is not R&D, and provides at subsection (4)(a) that those regulations “may … make provision by reference to guidelines issued by the Secretary of State”. That is how the Guidelines acquire force, and the parties agreed they applied (paragraph 49). The practical consequence is that the test a claim has to meet is not written in a tax Act at all. It is written in guidelines issued by a government department, and it is drafted around a person.

The Guidelines paragraph that decided the case is the definition of uncertainty, quoted by the tribunal at paragraph 52. “Scientific or technological uncertainty exists when knowledge of whether something is scientifically possible or technologically feasible, or how to achieve it in practice, is not readily available or deducible by a competent professional working in the field.”

The expenditure route was section 1052 CTA 2009, whose Condition B — in the version in force at 31 December 2015 — required that “the expenditure is attributable to relevant research and development undertaken by the company itself”. The officer relied on CIRD81960, the software page, for the proposition that digitising and coding data is not of itself likely to involve R&D (paragraph 30). The Guidelines themselves use the term competent professional without defining it. The working formulation came later, in Flame Tree Publishing. HMRC submitted there that to be accepted as one “an individual would need to be able to demonstrate appropriate qualifications, experience and up-to-date knowledge of the relevant scientific and technological principles involved” (Flame Tree at paragraph 66). The tribunal had “no hesitation in agreeing with HMRC” (Flame Tree at paragraph 68).

What the tribunal decided

The burden was on the company, and paragraph 69 set out what it had to discharge it with. It needed proof that the technology sought was not already readily available. It needed proof that the technology amounted to an advance and to more than “routine…copying or adaptation of an existing product [or] process”. And it needed proof that there were technological uncertainties a competent professional working in the field could not have readily resolved.

The answer on each was the same, and it was about who was speaking. Neither witness was contemporaneously involved and neither claimed expertise in the field. So “we do not attach weight to their assertions (for which they provided no supporting evidence – such as relevant extracts from, industry publications) as to what technology was and was not readily available during the relevant periods”. Nor was the company’s account of its competent professional enough: simply asserting that he was “competent through experience” and “providing a resume as brief as that contained in the 17 August 2016 letter was not sufficient” (both paragraph 70).

Paragraph 72 closes a loop many claims rely on. The company argued that because the uncertainties were not resolved, they cannot have been readily resolvable. That argument does not run, the tribunal held, “in circumstances where the Appellant has not proved there was a competent professional involved with the project”. Unresolved difficulty is not evidence of technological uncertainty unless someone qualified to judge the field says so.

Then the sentence the later decisions quote: “We find it remarkable that the Appellant did not provide evidence from someone that was contemporaneously involved in the project” (paragraph 73). The tribunal added “and/or from someone with relevant expertise who, having reviewed records of the project, might have been able to address the issues set out at paragraph 69 above” (paragraph 73). At the outset it had said what would have served instead. That was someone, even if not involved at the time, able to “state with authority what technology was readily available during the relevant period” and to say how what the company set out to build differed from it (paragraph 4).

The costs failed separately and for the same reason. The Evensys invoices described the work only as “Development”, “support”, “web development” and, once, “TDT Multi-lingual project”; the contract was unsigned with its statement of work never completed; and neither witness could say what Evensys had done (paragraphs 62 and 74). HMRC’s authorities, Gripple and BE Studio, gave no material assistance (paragraph 55). The appeal was dismissed (paragraph 76).

What it changes for a claim being prepared now

The scheme has changed; this part of the test has not. For accounting periods beginning on or after 1 April 2024 the current schemes apply — the merged scheme in Part 13 Chapter 1A, or Enhanced R&D Intensive Support in Chapter 2, with which one applies decided by the company’s R&D intensity. The definition of R&D reaches both by the same route, section 1041 CTA 2009 to section 1138 CTA 2010 to section 1006 ITA 2007, and the Guidelines definition of uncertainty still turns on what a competent professional working in the field could readily deduce. Nothing in what follows is affected by which scheme a claim is made under.

  • Identify the person before the narrative, and keep them reachable. The competent professional had left and the founder did not attend. Record who held technical responsibility for each project while the work runs, and take a signed account from them at the time.
  • Two or three lines of résumé will not do. Name the field, then the qualifications, experience and up-to-date knowledge of its principles — the three things the tribunal in Flame Tree Publishing said a competent professional has to be able to demonstrate.
  • Evidence the baseline from outside the company. The tribunal twice asked for industry publications or journal extracts. An assertion that nothing comparable existed is worth only what the person asserting it can be shown to know.
  • Describe subcontracted work in the contract and on the invoice. “Development” and “support” tie no cost to any project, and an unsigned contract with an empty statement of work ties nothing to anything.

A worked illustration. The additional information form asks, per project, for the field, the baseline, the uncertainties and how they were addressed — and gives the competent professional no field of their own. That does not make the evidence optional; it moves it into the file. What AHK needed was one page signed by its competent professional at the time: the field, the publicly available approaches with their limitations, and the question that could not be answered from them.

Where it sits against the other decisions

AHK is the earliest competent-professional decision and the one the others build on. Flame Tree Publishing adds the field-matching point — expertise in the company’s trade is not expertise in the field the advance sits in — and Hadee Engineering adds the treatment of a consultant’s report prepared after the event. Beer Express applies all three in 2026 and lands in the same place: the reports carried no weight because nobody with contemporaneous responsibility stood behind them.

It pairs with Strictly Money on attestations and with Tanglewood Care Services, where honest and experienced witnesses claimed no expertise in the field the advance was said to sit in.

One correction, because the passage is widely quoted against this decision. A statement of the burden of proof circulates as AHK Recruitment at paragraph 29 — “In order to satisfy the burden of proof, the Appellant would have needed to provide witnesses who could have testified to the facts necessary for me to conclude that the criteria set out in the Guidelines were satisfied” — and it is not in [2020] UKFTT 232 (TC). Paragraph 29 of this decision records the opening of the second check, into the 2015 return, on 3 October 2017; the words appear nowhere else in it; and the passage is written in the first person singular, where this two-member panel writes as “we” throughout. It enters the record as HMRC’s submission, recited by the tribunal in Flame Tree Publishing at paragraph 65, and Get Onbord at paragraph 79 and Beer Express at paragraph 91 reproduce that passage of Flame Tree rather than the source. Treat the words as a party’s formulation, not as a holding of this tribunal. Where else they may appear we do not say: Find Case Law’s full-text index does not reach First-tier decisions published only as PDFs before 2022, this one among them.

Has it been appealed?

No onward appeal has been reported, and no Upper Tribunal decision naming AHK Recruitment appears on the published record. Applications for permission to appeal are not consistently published, so the absence of a reported appeal is not proof that none was sought.

HMRC has not changed its published guidance as a result. Nothing in the CIRD manual’s dated change log is attributed to this decision and no Revenue and Customs Brief names it — unsurprising, since it applied the Guidelines rather than reinterpreting them.

The decision is First-tier Tribunal, so it binds only the parties and sets no precedent. Its weight is practical: it is the decision later tribunals reach for when the only person who can speak for the work is the adviser who wrote the claim. It sits with the rest of the R&D tax case-law register, one page per decision.

Sources

  • AHK Recruitment Ltd v HMRC — the decision, cited as [2020] UKFTT 0232 (TC), case number TC07718, appeal number TC/2018/06751, heard 29 November 2019, released 20 May 2020. Published on Find Case Law as a PDF under a permanent identifier, not the /ukftt/tc/2020/232 form. Every paragraph number above refers to it.
  • Section 1041 CTA 2009, section 1138 CTA 2010 and section 1006 ITA 2007 — the definitional chain behind the Guidelines; latest available revised versions, unchanged in the words quoted.
  • Section 1052 CTA 2009 — in-house direct R&D, as in force at 31 December 2015.
  • Guidelines on the meaning of R&D for tax purposes — the definitions of advance and of uncertainty the tribunal applied.
  • CIRD81960 — the software page the officer relied on.
  • Flame Tree Publishing Ltd v HMRC — [2024] UKFTT 349 (TC) at paragraphs 66 and 68, where the formulation of what a competent professional must be able to demonstrate was adopted, and at paragraph 65, where the burden-of-proof passage misattributed to this decision is recited.

CIRD81960 is indexed on our CIRD reference index, with HMRC’s own title for it, a line on what it says and the date HMRC last revised it.

First-tier Tribunal decisions bind only the parties to them and set no precedent. This entry describes the decision as it stood at the review date above, as general information rather than advice on your circumstances — see our terms. For an answer on your own facts, talk to us.