What’s holding us back from AI?

09
Feb
2023

Nick Rich

Artificial Intelligence (AI) is expected to revolutionise the legal sector from chatbots disputing parking tickets, to contract analysis tools, and legal research. We’re about to see an industry first with an AI legal assistant defending a speeding case in a US court, where the AI will listen to the proceedings via a smartphone before advising the defendant on how they should answer through an earpiece. This represents a significant milestone and signifies that AI has already been accepted as a black letter law with its use set to become more commonplace over time. 

However, as of today the adoption of AI legal technology across the legal sector is piecemeal with its use largely limited to multinationals with high litigation workloads. A few years ago the Blickstein Group found that 66% of legal departments were not using the technology at all and even by the end of 2021 usage was light, with 80% of those questioned using technology assisted review in less than 30% of matters and more than half using it for just 10% of matters. 

The Law Society provided a snapshot of adoption at the end of 2021 when providing evidence to a parliamentary committee. It stated that “different segments of the legal market are at different stages of maturity” with AI used commonly in the business-to-business market but not the business to consumer or not-for-profit sectors. The most compelling use of AI was in legal analytics, project management, governance, and compliance and contract management with tools for collaboration, document management, IP management, and e-billing. 

Evolution of AI

Compared to digital transformation in other sectors such as FinTech, adoption remains slow which seems surprising given that AI promises to significantly reduce workloads. By providing critical insights before collection, AI-driven early case assessment can save downstream costs on hosting, processing, document review, and production, as well as associated staff time. In doing so, it enables legal teams to get to the facts faster and make comparisons with prior cases and outcomes. 

Within e-discovery, supervised machine learning AI which utilises predictive coding has rapidly become the most effective technology for increasing efficiency and accuracy in document review. Predictive coding sees the AI trained by a human operator who provides the system with plenty of examples and correct responses – seed sets – helping the AI to learn and then apply that learning to future cases. It’s been the dominant form of automation in document review for a decade but because it requires the AI to be fed information, the output was only ever as good as the data input.

That’s now changing, with AI becoming much more efficient and independent. Human-taught AI has been superseded by unsupervised machine learning AIs that don’t require seed sets at all. Self-taught AI utilise learning algorithms that operate in the background, observing how human attorneys review documents and learning the criteria that makes a document relevant to a matter.

This form of smart document review identifies hidden patterns in the data, similarities and anomalies before referring these to a human for investigation. Keywords, concept clusters and communication patterns are analysed in other legal processes in order to quickly and logically find facts and, because words and phrases are contextually connected and grouped at speed, the reviewer can drill down and expose other avenues for investigation hidden within millions of pieces of information. 

A tipping point

These AI tools won’t seek to replace the human reviewer but will instead augment and aid them, ploughing through mundane tasks that will then free up the reviewer to focus on more high-value cognitively demanding tasks. But AI still has some work to do in winning over hearts and minds. There’s still a widespread perception that its application is limited because its task focused and not as flexible as a human, even though it doesn’t tire and can get to the facts of a matter faster while making recommendations on how to improve the process. 

Thanks to recent advances, we’re now at a tipping point with AI as legal departments grapple with the decisions over when to commit to the technology and where to apply it. Ignore the adoption curve and they risk getting left behind, jump onboard too soon and they may find the process requires change management and is initially disruptive. But there is a middle ground that can allow legal professionals to adapt to the new normal of AI-assisted e-discovery.

Legal teams don’t need to adopt AI wholesale but can instead use it for specific tasks enabling staff to acclimatise to the technology. Document review provides an obvious starting point as it’s the most expensive part of the e-discovery process so that immediate gains can be achieved through automation of pre and post review processes. 

There are a number of options here on how the team could choose to implement AI. Its usage could be restricted to the prioritisation of documents for human review, for instance, or to reviewing incoming productions. Or the AI could be applied to the secondary review of lower-priority documents or as a form of quality control for human review. If, however, there’s still a lack of confidence over how to utilise the AI solution, the team could elect to use an external service provider who can then manage the workflows and train them on how to use it. 

In this way, the legal team can gradually adjust to new ways of working while capitalising on the benefits AI has to offer. It avoids the risk of becoming non-competitive and can use AI to more efficiently utilise its most precious resource – its human reviewers.