Sponsored content: Kevin Perkins, head of legal defence at law firm Carpenters Group, discusses how the industry should respond to the wave of AI-generated information that customers are now creating
At the recent Insurance Times Fraud Charter meeting, sponsored by Carpenters Group, there was plenty of discussion about artificial intelligence (AI) and its application in facilitating fraud – deepfakes, synthesised voices, fabricated documents and the other obviously scary stuff.
However, the part of the conversation that particularly landed with me was much more mundane and probably much more relevant on a day-to-day basis – the amount of resource that organisations are now wasting having to read, unpick and respond to extraordinarily lengthy AI-generated complaints, claims and dispute documents.
Whether it comes in as a customer complaint, a pre-action letter, or even something filed at court, we are seeing more and more of what I can only describe as “confident nonsense”.
Historically, the more hopeless complaints, letters and pleadings tended to announce themselves quite early. They were often rambling, repetitive, inconsistent, or so badly put together that a sensible triage process could deal with them fairly quickly.
AI changes all of that.
It will rarely tell a determined user that they have misunderstood the point. Instead, it takes the misunderstanding and dresses it up. A weak argument suddenly arrives with headings, bullet points, citations, careful legal language and a level of confidence that it simply has not earned.
That is the problem. Not that the argument is better, but that it now takes longer to work out that it is still twaddle.
You cannot simply ignore these documents, because they look more professional than they are and organisations rightly have to give them proper consideration.
People therefore end up spending inordinate amounts of time on them, because each one has to be read to work out whether there is a real point buried somewhere in the noise.
Inventing solutions
As has been well documented, AI can invent authorities, dates, rules or quotations that sound just plausible enough to require checking. That means time spent proving that something does not exist, which is about as productive as it sounds.
You cannot always send a stock response. If the complaint or letter is long, technical and superficially credible, someone has to produce a careful answer to something that may have taken minutes to generate.
We are told AI is making access to justice easier. There is patently some truth in that. But it is also making it easier to produce large volumes of persuasive rubbish. Someone who might previously have sent one angry email, or perhaps would not have raised a complaint at all, can now generate a suite of complaints, legal arguments and follow-up requests before lunch.
The problem is that organisations cannot simply assume a complaint, claim or letter is nonsense just because it looks like it may have been generated by AI. These issues have to be treated as genuine until they have been properly reviewed. That is entirely right, but it is also where the resource issue bites.
The risk is not just that organisations waste time on the hopeless cases. It is that genuine issues get prejudiced because the same people are stuck sifting through the time wasters. The more resource that is spent dismantling polished nonsense, the less resource there is to identify, investigate and resolve the genuine and worthwhile complaints, claims and disputes.
That is likely to be the next real challenge. Not stopping people from using AI, because that plainly is not going to happen, but working out how to separate genuine complaints, claims and dispute correspondence from AI-assisted noise quickly, fairly and consistently.
The organisations that manage that will save time and be better placed to deal properly with the issues that actually deserve their attention.
