Young: Access to justice delivered through responsible AI
The Solicitors Regulation Authority (SRA) has for the first time authorised a law firm whose offering is solely provided by an AI-powered litigation assistant.
The SRA said the authorisation of Garfield.law was a “landmark moment for legal services in this country”.
Garfield is initially supporting businesses recover unpaid debts of up to £10,000 in the small claims court, and eyeing small housing disrepair claims as the next area it could move into.
It can also be used, and white labelled, by law firms.
Garfield is the brainchild of Philip Young – a hugely experienced commercial litigator who co-founded City disputes boutique Cooke Young & Keidan – and quantum physicist Daniel Long.
Mr Young described the system as “access to justice delivered through responsible AI”. It is a hybrid AI/expert system built on “domain-specific expert knowledge, modelled to follow the Civil Procedure Rules with precision and transparency… unlike generic legal chatbots”.
Garfield charges very small sums for pre-action work – £2 for a ‘polite chaser’ and £7.50 for a letter before action – and once issued takes the relatively small fixed recoverable costs allowed for by the rules in such cases.
Mr Young explained that the system was designed in a similar way to how “a good law firm should operate, so the facts come from client but the knowledge of process, procedure and recommendations as to what to do come from the system”.
The user drags and drops the unpaid invoice – and contract, if there is one – into Garfield, which checks for issues such as limitation and advises whether they have a claim. It looks too at Companies House records to see if the debtor is solvent.
It then asks whether the user wants to send a chaser or a letter before action. The system checks that it has all the information it needs and, if so, asks the user to review the facts as set out in the letter before sending it. The letter will make it clear to defendants that Garfield is an AI solution.
Garfield will review the reply – asking the user for any information needed to assess the reasons given by the debtor – and draft a response. Ultimately, it can issue a claim, apply for judgment in default or respond to the defence and support the user through to trial, creating a trial bundle and skeleton argument.
It is not autonomous in that the user must approve each step.
Mr Young said they took a “multi-layered approach” to ensuring Garfield’s accuracy. “First of all, we’ve spent two years building and testing it. We’re confident in the results and we can track mathematically the quality of the results. There are real-time analytics built in.
“Secondly, because I’m your typical cautious lawyer, we’re going to have a staggered launch and I will track – and am tracking – live claims through the system and watching all the documents that have been generated to make sure they’re good.”
Once confident of this, he plans to move to a “risk-based sampling approach”. Law firm users would be able to track accuracy themselves.
Mr Young argued that Garfield was good for defendants too because they received properly presented claims in an area where there were many litigants in person.
It has received a warm reception from the senior judiciary, with Lord Justice Birss – the deputy head of civil justice – last year highlighting AI technology as demonstrated by Garfield as “absolutely at the core of what we can do for access to justice”.
Mr Young said users to date ranged from sole traders to big corporations, as well as both law firms and accountancy firms.
“It has been very interesting talking to law firms about Garfield because they have instantly spotted how they can leverage the platform to conduct more small debt claims at greater scale – saving the need to recruit additional staff – without having to increase their fixed cost base by additional salaries and overhead such as bigger offices, and potentially boost their margins in the process.
“Also, they totally get the friction that Garfield removes and they can see it potentially gives them a competitive advantage for clients who would prefer a human overseen solution rather than a pure AI solution.
“We’re agnostic – we don’t mind if users want to use us directly or via a law firm. Either is great.”
Mr Young praised the “very exhaustive” process he went through with the SRA over eight months to be authorised as a law firm. The regulator had a “forward-thinking attitude but wanted to make sure people were safe”.
It looked at “the product, the people involved and the risks” and it was continuing “to keep an eye on” Garfield.
The SRA said it had sought reassurance that there were appropriate processes in place to quality check work, keep client information confidential, and safeguard against conflicts of interests.
It has also highlighted the risk of AI ‘hallucinations’ – the system will not be able to propose relevant case law, which it described as “a high-risk area for large language model machine learning”.
Mr Young said the system was built in a “modular way”, meaning that other types of claims could be layered on top of its understanding of the small claims process, with low-value housing disrepair claims – where access to justice could be difficult to achieve because of the economics – a possible next area.
He predicted that Garfield could in time have a role in fast-track claims, but “when you get into higher tracks, people are going to be want human lawyers involved”. At the same time, it should be able to support and make lawyers handling such cases more efficient.
Garfield, which has angel investors, has around 15 people in its team, including four solicitors and a barrister.
SRA chief executive Paul Philip said: “The first regulatory approval of an AI-based law firm is a landmark moment for legal services in this country.
“With so many people and small businesses struggling to access legal services, we cannot afford to pull up the drawbridge on innovations that could have big public benefits. Responsible use of AI by law firms could improve legal services, while making them easier to access and more affordable.
“Yet trust and confidence in regulated legal services depends on the public knowing that high professional standards are being met. Any new law firm comes with potential risks, but the risks around an AI-driven law firm are novel.
“So we have worked closely with this firm to make sure it can meet our rules, and all the appropriate protections are in place.
“As this is likely to be the first of many AI-driven law firms, we will be monitoring progress of this new model closely, so we can both manage the risks and realise the benefits to consumers.”
Mr Young said the launch of ChatGPT had made him realise the potential for AI in the law and he had been inspired to target debt claims by problems his brother-in-law Andy King, a self-employed plumber, encountered.
Mr King said: “As a small business owner, I have had occasions where customers did not pay their debts. These usually had to be written off as it is too costly and time consuming to chase these debts through the courts. Many people know this so use it to get free labour and goods.”
Mr Young then went out and found Mr Long to bring his technical expertise to the project.
Leave a Comment