By Matthew Pascall, Legal Director
Estimated Reading Time – 5 minutes 17 seconds
Temple, along with many other businesses, is starting to explore ways in which AI can support the work we do. I have already started to receive marketing emails promising me software that can predict the outcome of a case. Am I soon going to be out of job? Is Temple going to start saying “Computer says No!”
No.
I am no IT expert, and I am not going to pretend I understand anything about AI. It follows that there may be many things in this article which an expert would challenge. If so, please do. An open and well-informed debate about the genuine and exciting opportunities AI can offer all of us that includes a careful reflection on its potential pitfalls is to be welcomed.
For a while now access to data from the Courts Service has enabled relatively sophisticated reporting that will show patterns in the way particular cases or types of cases are handled by particular judges at trial. This merely reflects the fact that some judges can be more conservative in their approach to some cases than other judges and, as a result, they can be seen as more claimant or defendant friendly. It is also becoming possible to see that, in overall terms, some types of case seem to be more successful at trial than others.
But that merely reflects the fact that, by their very nature, some cases are more difficult to prove than others. However, a judge seen as “conservative” or “defendant friendly” will often happily find in favour of a claimant. If a fair and reasonable application of the law to the facts leads to a judgment in favour of party “A,” party A wins, whether or not A is a claimant or defendant. The moral of the tale is, be careful when looking at trends and patterns.
If software was able to identify certain key features in particular cases and then look to see how other reported cases with broadly similar features were generally resolved at trial, that software might be able to give you a relatively reliable indication of the probability of success. If I have understood AI properly, it has the potential to carry out this analysis rapidly and with some degree of sophistication.
But why use artificial intelligence when you can have the real thing?
Trials involve the telling and interpretation of complicated stories. Evidence is often ambiguous. To finally resolve the matters in issue at a trial a judge must apply a set of relatively broadly based rules and often have to exercise a discretion. It is an intensely human process.
Humans are not always predictable and the judgments they make can be erratic.
- Underwriters who have studied the law and, in some cases, practiced it for many years will bring to the underwriting assessment of a case all that they have learned and, more importantly, experienced. That learning process and the acquisition of experience doesn’t stop when they take-up underwriting but becomes more intensive and richer given the wide variety of cases they underwrite on a daily basis.
- Underwriters are often well-placed to see trends emerge in the approach taken by the courts to certain types of case or particular judges. Statistical analysis may well help to act as a check or reference against which to measure impressions gleaned over a period of time, but the computer assisted analysis on its own is never good enough.
Nonetheless, we are not going to ignore AI or the contribution it may well make in the years ahead to support (but not replace) a well-informed and experienced critical human analysis of the merits of a particular case.
AI document summarisation
This is a tool we are already assessing, and it is impressive. In very simple terms, several hundred pages of pleadings can be up-loaded onto the relevant platform and within 5 to 10 minutes a detailed summary is produced.
There is the potential for Temple to invite those who currently complete proposal forms to simply up-load the relevant correspondence and pleadings to some new Temple software, confirm that the AI generated summary is correct and then complete a much shorter proposal form. The alternative is for us to carry out the AI summary process ourselves with the documents submitted to us as they currently are by solicitors and brokers. An accurate and concise summary is very useful for underwriters. The challenge is to ensure that we do have a complete and accurate summary.
Testing, testing…
In recent testing, we uploaded the pleadings in a professional negligence “lender claim” brought against a valuer by a bank alleging that a property had been carelessly overvalued.
Most of the particulars of negligence were included within the summary but it omitted a key and highly contentious allegation (common in these claims) that the valuer had wrongly advised that the relevant property was good security for the proposed loan. Having omitted that particular allegation, the valuer’s pleaded response was also committed. The pleaded allegation and response only took up a couple of lines in the Particulars of Claim and the Defence, but the omission was significant. Our concern was that the summarising software was not intelligent enough to know what could usefully and safely be omitted from a summary.
In Scottish procedure the pleadings close (as we would say in England, Wales and Northern Ireland) when the pursuer lodges and serves the closed record. The closed record is a summary of the pleadings in a single document which concisely but accurately matches against each factual allegation by the pursuer the defender’s response.
It is a hugely valuable document from an underwriter’s perspective because it encapsulates the respective cases of the parties. Once AI is able to produce for the cases we see from across the whole of the UK something akin to a closed record, it will have become an extremely useful tool.
AI also has the potential to enhance the quality and speed of actuarial analysis to provide good performance and loss ratio data. That in turn can feed into better and potentially more competitive premium pricing, particularly for delegated authority schemes.
The Temple Perspective
For the foreseeable future, those submitting cases to Temple will know that whilst we may well use AI to support and enhance what we do, it will not replace the critical analysis of the relevant risk by a well-qualified and experienced human.
If we decline to provide cover, that is, if it’s a no, it won’t because the computer says so!
Matthew Pascall
Legal Director – Head of Commercial
Matthew Pascall
Matthew was called to the Bar in 1984 and joined Guildford Chambers two years later. Spending more than 30 years in practice there, he was listed as a Legal 500 Tier One barrister.
He joined the commercial team at Temple Legal Protection as Senior Underwriting Manager in 2017.
Matthew was appointed to Temple’s Board in December 2022 as Legal Director and Head of Commercial.
His knowledge of the commercial legal sector and litigation practice is invaluable to the business and our clients, providing specialist experience to lead the commercial litigation insurance team.
Read articles by Matthew Pascall
