We’ve all heard a lot about Artificial Intelligence (“AI”) and anyone doing an internet search for information now will probably have received a neat little summary of what they are looking for. But how much can you rely on it?
Principal Ruth Corkin describes her recent experiences with AI.
There are some limitations with AI. For a start, it can’t be wrong or return a nil answer. That does sound like a lot of tax advisors to be fair, but the advantage of a mark one human is that it can admit that it can’t find an answer or doesn’t know! AI can learn if you tell it that the answer is incorrect but won’t automatically change its view.
AI can’t deal easily with nuance, which in the tax world is the key to a number of arguments. There have been tax cases in the Tribunal where the taxpayer represented themself and based their case on AI generated case law. As Abraham Lincoln said, “A man who represents himself has a fool for a client” and in these cases it was definitely true!
In the most recent upper tier tribunal case of HMRC v Marc Gunnarsson, the taxpayer relied on three first tier tribunal cases to support his case. He then sent an amended and final skeleton argument that did not reference the decisions after HMRC pointed out the cases did not exist. In addition, the taxpayer admitted he used AI for his written submission.
Understandably, the Upper Tier Tribunal Chairman was unimpressed and referred to a High Court case decision on the use of AI which took a dim view of lawyers failing to check written arguments generated by AI. In Mr Gunnarsson’s case, the tribunal expressed the view that both legal representatives and unrepresented litigants are responsible for “the accuracy, both the reliability and credibility, of the information, both evidence and submissions, they present to the FTT (First Tier Tribunal) or UT (Upper Tier Tribunal)”. There was a further warning in the decision that the UT takes the matter of falsehood very seriously.
On the plus side, AI is very good at summarising lengthy cases or finding a particular phrase in a decision. As a very junior team member in the early 1990s, one of my tasks was to summarise every tribunal decision for the wider team. In those days, bundles of paper case decisions would arrive, and I would have to read through them and provide a summary. At the time, most of the cases were default surcharge reasonable excuse decisions, so there was a pattern and, to while away the time, I would challenge myself to guess the outcome. However, had AI been around, it would have made my job less of a burden and a lot less boring!
AI processes large volumes of data in a short period of time. It can put that data in a format that is readable. For example, it can draft a business case for staffing, attendance at a conference etc. or an itinerary for a trip (complete with GoogleMap references – Yes I have used it for that as well!)
Where I have found it useful with customs declarations and classification. Using Optical Character Recognition (“OCR”), AI tools can extract the data needed to complete fields on the declaration much faster than a human being can, cutting down on completion time and adding profit to the task.
By using natural language, it can interrogate the UK tariff and, ask relevant questions to reach the most relevant commodity code. Where it returns an incorrect answer (which is not often!), it will learn if you challenge it.
In my day job, AI is a useful tool, but I am naturally cynical and check the results. I can look at something in my field of expertise and query something that seems odd. Therefore, critical thinking and analytical human minds are still necessary, otherwise users can end up like Mr Gunnarsson with a fool for a client!
Please get in touch with Ruth (contact details below) with your own experiences of AI.
