AI: Helping firm to decide on new cases
The biggest specialist personal injury firm in the country is working on ways to harness the power of technology related to ChatGPT to help it handle medical negligence claims.
Dan Taylor, head of integration at Fletchers, said that having successfully built and used two versions of an artificial intelligence (AI) tool to help lawyers make initial case decisions, work was beginning on a third version.
Mr Taylor said SIDSS (Structured Information Decision Support Systems) used by Fletchers had saved “tens of thousands of hours” in staff time previously spent deciding whether cases should continue to be investigated.
He said the first iteration of SIDSS was based on a “random forest” of decision trees, while the second introduced machine learning and neural networks.
Work on the third version would involve discovering “how we can harness the power of ChatGPT-type models and apply them to our data”.
Fletchers received £225,000 in public funding from Innovate UK to develop AI tools with Liverpool University through a knowledge transfer partnership launched in 2016.
SIDSS compares initial client details of medical negligence claims with the tens of thousands of previous cases in the law firm’s database to deliver a “confidence score”, expressed as a percentage, as to whether the firm should continue to investigate the case.
Every case, however large, is put through SIDSS, as long as clients give their consent. Rejected cases are sent to an “overturn team” to be checked to see whether they should still be investigated.
Mr Taylor said that the overturn teams agreed with SIDSS in an “overwhelming number of cases”.
Lawyers working on the case can also decide whether they want to ignore the confidence score. “This gives clients confidence that decisions are not being made just by AI,” Mr Taylor said.
Two further systems have been developed, but not yet implemented: MEDSS (Medical Evidence Decision Support Systems), to help lawyers make decisions on cases once the medical evidence had been received, and an AI tool to read and sort the medical evidence itself.
Mr Taylor said the problem with clinical negligence cases was that “complexity does not align with value”, and some relatively low-value cases could be complex.
By using AI, the firm wanted to close this “gap in the market” and provide access to justice for as many people as possible.
Before working with Liverpool University, the firm had spent a year working with IBM to see if software based on Watson could be adapted to support legal decisions.
“We could potentially have made it work for us, but it was not commercially viable because it was so expensive. We could have spent several millions, with no guarantee of achieving anything.”
Mr Taylor said the SIDSS technology could be applied to other personal injury cases, though no decisions had been made about this. The idea that the software might in time be licensed to other law firms was “not completely off the table”.
He added: “Early decisions are really important to people with medical negligence claims. ChatGPT-type models are an enormous opportunity to improve things even further.”
