HMRC: Man meets machine
Take current tax legislation, add a little artificial intelligence of the sort underpinning digitisation, pursue it to its logical conclusion – and you could get a head-on collision as man meets machine, a tribunal judge suggested recently.
His comments are the more interesting coming alongside the KPMG report published this month, ‘Technology in Tax’.
KPMG’s report is a thought-provoking glimpse at the possible future in tax and technology.
Essentially, it’s about revolution generated by the internet and artificial intelligence (AI).
The techno-boffins call it the Third and Fourth Industrial Revolution – and that’s where we’re located now.
Where the Third Industrial Revolution is founded on digitisation, the Fourth goes further than that. Much much further.
With AI, “the speed, size and scope of change is such that it becomes qualitatively different,” the KPMG think tank tells us.
What would AI look like in the world of work?
Perhaps 30% of corporate audits carried out by AI by 2021?
The first AI machine on a corporate board of directors?
But in a tribunal case deciding the fate of taxpayer Thomas Richter, [Thomas Richter v HMRC TC05816] judge Richard Thomas recently pointed out an inconsistency when it comes to man meeting HMRC machine.
Judge Thomas made in his comments obiter in a case appealing late filing penalties following on from the Donaldson case.
They didn’t form part of his judicial decision – but judge Thomas made it clear where the battle lines are likely to be drawn up in the future.
Pointing out that the paragraph 5 and 6 assessments were made before Richter’s income tax self assessment return for 2010/11 was delivered, judge Thomas put his finger on a potential problem.
Paragraph 24 (2) (a) Schedule 55 stipulates that “HMRC is to determine the amount mentioned in sub-paragraph (1) to the best of HMRC’s information and belief.”
The six and 12 month penalties imposed by para 5 and 6, are the greater of £300 and “5% of any liability to tax which would have been shown in the return in question.”
So, when a penalty is determined before the return is submitted, this should logically involve some sort of review, consideration and estimate by HMRC of the likely tax liability on the return.
But HMRC hadn’t carried out a review.
Essentially because, as judge Thomas explained “not even the most sophisticated computers can (yet) form beliefs, and certainly not those operated by HMRC.”
Mechanically churning out £300 penalties isn’t enough. To satisfy the law, you also need to form a judgement decision. Would 5% of the tax liability be higher?
Musing on HMRC penalty generation procedure, outlined in HMRC’s Self-Assessment Manual at SAM 61240 judge Thomas pointed out that the process is automatic.
The SA computer trawls a database to find which tax returns have been issued but not received before the six month penalty trigger point.
The computer then issues £300 penalty notices.
“The computer is not programmed to interrogate any data it has about past liabilities,’ judge Thomas said.
“It does not seem to me that this practice involves an officer of HMRC examining information or, particularly, forming a belief about anything.”
These, judge Thomas indicated, were issues which he, or other tribunal judges were likely to return to in the future.
He put forward the view that the legislation in paragraph 24(2) demands action by a “flesh and blood human being.”
… and grammar.
The clue judge Thomas cited in paragraph 24 (2) was grammatical. The verb used is singular.
He pointed out “the use of the phrase ‘HMRC is.’ In other parts of Schedule 55 HMRC takes the plural.”
This meant that, on a point of grammar, HMRC’s automatic paragraph 5 and 6 penalty assessments might not be valid.
“In the absence of any indication that a human being formed a belief about the level of penalty appropriate in this case I find it difficult to see how the automatic paragraph 5 and 6 assessments can be valid.”
Artificial intelligence and tax
Judge Thomas’ comments sit interestingly alongside the future of the tax profession envisaged in the KPMG report.
The report anticipates a new world “likely to favour complex thinking rather than deep technical knowledge” – witness the prediction that “one can imagine a computer considering the optimal structure for an investment in the future.”
“Technical expertise… purely as knowledge will be diminished.”
So perhaps in the future a computer – instead of Judge Thomas - will trawl the legislation checking verbs singular and plural.
Current taxonomies - like the division between goods and services for instance - are heading for the melting pot.
KPMG asks: “When your lawn mower measures the height of your grass, lets itself out of the garage and mows your lawn while you’re at work, has the lawn mower manufacturer shifted into the realm of a providing a garden service?”
And like KPMG, judge Thomas’ remarks also imply that the road ahead is going to be bumpy.