Meet Della AI – a new legal AI doc review and analysis platform that uses deep neural networks to identify relevant parts of contracts in response to your queries.
Some may say the market is already full of companies covering this area, but Artificial Lawyer’s view is that it is such early days still – for what is a growing market segment that has existed for only a few years – that there is every reason to consider new players.
As the London-based company’s CEO, Christophe Frèrebeau, told Artificial Lawyer: ‘We have a very different approach to the existing players, which is giving us an edge and we are getting traction – even with law firms who adopted older solutions, as well as with small to midsize law firms who used to be slow to adopt AI technologies.’
‘A lot of the clients we are talking to have been disappointed by existing AI options. Either due to the time required to train the system for new tasks, or even because of the learning and time investment required hasn’t delivered what was promised,’ he added.
He also underlined that most firms they encounter don’t have the spare time to train and engage with a new product. Therefore Della’s approach aims to address this pain point through ‘ease of use and flexibility’ as they put it. Of course, the proof of the pudding is in the eating.
Frèrebeau continued: ‘We have worked hard to make the solution as easy to use as possible and we have ensured lawyers can benefit from the AI, but always remain in control.’
All well and good. But, what does this mean in real terms? What does Della AI offer? The company explains it can identify the portions of text you are looking for in response to questions you set the system. This covers, at present:
- Due Diligence – Della is pre-trained on standard M&A and due diligence questions and can identify and flag unusual or onerous clauses automatically
- Contractual Review – Della extracts essential terms and can highlight deviations from your expectations
- Compliance – Ask Della to use your compliance playbook to analyse client contracts and spot potential issues
- Lease Analysis – Della can find standard lease provisions and metrics across a wide contract portfolio
- Libor Replacement – Della can identify out of date Screen Rate clauses for LIBOR repapering
- GDPR – Della can analyse Data Processing Agreement (DPAs) to help determine GDPR compliance
As you can see this is a broad range, covering scenarios ranging from what could be very high volumes to smaller sets of documents, such as summarising and keeping track of key contractual terms for in-house counsel.
And its approach to the task?
Frèrebeau explained: ‘In the past year NLP has taken huge steps forward with the development of BERT (an NLP model) and its successors. We test these state-of-the-art deep neural networks as they are released. We then adapt and improve them for answering questions about contracts and related documents.’
‘The technology we use is capable of exceeding human performance across a variety of NLP benchmarks. Legal questions are harder, but with sufficient review data we can exceed this level of performance for the relevant documents,’ he added.
All well and good again. But, is Della AI bringing something new to the party? Frèrebeau is adamant that they are.
‘The core technology we use is very different from traditional similarity detection algorithms,’ he explained.
‘We effectively train a neural network to understand legal grammar and semantics. By asking a lot of questions such as: ‘Is there a cap on xxx’, ‘Is there an exclusion to xxx?’, ‘Is a notice required for xxx?’, the system is learning to adapt to legal questions.’
And Artificial Lawyer especially noted this comment from Della AI’s CEO: ‘Our system also learns across languages and we have seen clients getting very good results on new sets of questions without any training.’
Of course, getting exact metrics here so we can say Della AI is as good as X company, or better than Y company, is tough. Over the years this site has asked legal AI review companies if they’d be happy to do a public bakeoff – the answer has always been no.
This leaves law firms, ALSPs and inhouse teams with the job of testing out applications and building their own test criteria. And this is not necessarily a bad idea. No matter what scores you may see it’s sensible to conduct your own A/B tests internally to decide what works best for you, and what suits your needs and work culture.
In short, if you’ve got the internal resources to test out a new doc review/analysis system then it’s probably worth doing so. The market doesn’t stand still, ‘new models’ come to market – maybe they don’t change the world entirely, but perhaps have some aspects that are genuine improvements.
Last question for the company: given the growing number of legal AI doc review and analysis companies, how do you see your product breaking through?
‘Della is easy to use, you simply ask your questions with your own words and review the results like you would review the work of interns. You can get started and get value immediately,’ Frèrebeau explained.
So, to conclude, is Della AI a radical departure from the competition? Well, there are a number of legal AI companies that allow you to pose questions to the system and get answers back from a doc stack. Other legal AI companies use neural networks, along with a host of other NLP approaches. And the core offering of review and analysis is not a massive departure either. But……! If we are very pragmatic about this, and focus on product performance, if Della AI can provide a smoother, faster, easier, perhaps even more economical result, then it has to be worth exploring.
After all, we are using this tech in the name of efficiency, if Della AI can move that dial forward in relation to doc review and analysis that has to be welcomed.