A firm pioneering the use of artificial intelligence to assess the strength of civil cases says the system has reached a 91% success rate in spotting unmeritorious claims.

Noah Milton, AI project director for national firm Fletchers, said the claims-sorting system has made rapid progress since its inception almost a year ago.

Speaking to the Motor Accident Solicitors Society conference in Manchester, Milton revealed that testing has found that technology is better at identifying claims that have no merit than spotting claims that should definitely be pursued. This is when measured against the decisions that a human would have made based on the same information.

But with 70,000 cases already available for the system to learn from, the tool’s reliability is improving all the time. 

‘When it says to reject a case it is right 91% of the time,’ Milton told the Manchester audience. ‘We are not trying to make a final judgement but trying to reduce the burden on our people. ‘The tool is very good at identifying cases that are unlikely to be taken on. The impact is that we are able to spend less of our precious lawyers’ time on reviewing cases where we don’t have very much information to feed off but need to make a decision.’

Fletchers teamed up with the University of Liverpool last year on the AI development project. The partnership continued work on SIDDS (structured information decision support systems) which had the potential save tens of thousands of hours in staff time previously deciding whether serious clinical negligence cases should be pursued.

Milton was hired to head the development programme and a team of data scientists brought onto the team. Funding for the four-year multi-million-pound project was provided by Fletchers, which is owned by an affiliate of private equity investor Sun European Partners.

Milton told the MASS conference that the most exciting tool in development is generative learning to review medical records. This has the potential to sift through thousands of pages and spot key information that could be used to prove a case. The AI can take disordered, unstructured files prepared by different NHS trusts or providers and featuring doctors’ handwritten notes, and arrange that into dates and categories and provide a three-page timeline to help understand how a potential client has been treated.

Milton added: ‘Lots of clinical negligence cases hang on one page and you can find that very quickly. We are testing it with personal injury over the next few months – we are really excited to see where it evolves.’

Launching the project last year, Fletchers Group chief executive Peter Haden said the use of AI could be a significant step forward not just for the firm but the sector in general.

He said: ‘We have already shown that AI can be a force for good. We now intend to broaden the scope of AI to give many thousands more injured people the opportunity to benefit from the justice system, even for lower value cases where justice has been harder to access following government reforms and budget cuts.’