Is Google moving into the legal tech space, in particular the contract analysis segment of the market? The short answer is: yes. But, Google sees this as an enterprise play where they will work with partners, such as Ironclad, and with others to be announced, rather than offer their services directly to the end customers.
As Vinod Valloppillil (pictured below), Head of Product, Google Cloud Language & Vision AI, told Artificial Lawyer: ‘Contract DocAI is one of our biggest investment areas. Where do enterprises get the biggest bang for their buck with AI? It’s in unstructured data [such as contracts] so we are committed long-term to this domain.’
But, what is Contract DocAI and how does it connect to the legal world?
Valloppillil explained that Google Cloud has a remit to focus on enterprise use cases, i.e. helping large corporates, especially around things that Google is a master at such as data analysis. Part of this is centred around the Document AI capability, which is generalised NLP, and then – which is the new bit that really enters the legal tech world – is the launch of Contract DocAI, designed specifically to enable the finding, extraction, and presentation of key data in legal and business documents.
It also has some new capabilities such as being able to use decision tree learning to infer information from other data in a contract, e.g. finding an expiry date that may not be mentioned, but the system can figure it out from other date references in the contract.
But, the key thing here, (see the second image), is how they are focusing on the core language and clauses of contracts that the legal world will be very familiar with, e.g. contract duration, governing law, and warranties.
As noted, in the CLM world they are working with Ironclad, but Valloppillil said that more partnerships will follow. In fact, their strategy is all about being the broad horizontal provider of constantly improving NLP software for contract analysis, while the legal tech companies – and others that want to enter this space – will be the subject matter experts with the Vertical engagement with the end clients and with their own wrap-around UI/UX and other capabilities.
I.e. Google supplies the contract language understanding core, the partners can build everything else they want around it, including additional training to be even more specific in its use cases.
Probably the next big question everyone in the legal AI world is asking now is: will this be better than what we offer? The answer is that without some scientifically managed ‘bake-offs’ it will be hard to prove at present. But, from Valloppillil’s perspective, Google’s Contract DocAI has a lot going for it and is going to keep improving because they are investing in it. They also have a couple of advantages……
As he explained: ‘We process the world’s largest corpus of language, the World Wide Web.’ And few would argue with that. They also, he said, are able to bring in entity understanding from other domains that also appear in contracts that legal teams may handle. For example, they’ve spent time working on mortgage documents, and that learning gets folded into the mix.
‘We have deep learning and language models from other domains. We take that knowledge from other areas, e.g. mortgage processing, and also from our general world knowledge. And all of that knowledge is very transferable to contract processing,’ he stated.
He then sums up: ‘We can take all that Google knows about the world and apply it to contracts.’
And that’s a fairly daunting thought for competitors. It also has more machine learning and language model experts on staff than probably all of the legal tech world combined. So, there’s that.
But, the good news for those wondering what will happen next, is that as Valloppillil highlighted, there is no plan for Google to go directly to the legal teams of companies, or the partners of law firms, it will instead work with tech companies that it wants to partner with, in what they simply see as the ‘contract vertical’. They don’t even really see the world in terms of ‘legal tech’.
However this works out, the reality is that if Google wants to do well in this area – it can. It has the resources, know-how, access to data, and all the rest, to really pioneer here. Moreover, when there are really big steps forward in NLP tech, no doubt Google will be first in line in terms of applying it directly into their offering. That will be hard to beat.
One last area that was very interesting, especially coming from such a tech-focused business as Google, is that Valloppillil very much believes in the need for a human in the loop approach when it comes to highly complex, high-risk contracts.
Despite the huge capabilities of this $150bn revenue business, Valloppillil explained that: ‘Humans in the loop is a core competency for an AI provider. They will be needed for the foreseeable future.’
Artificial Lawyer then asked him about accuracy levels. Given its tech prowess, will Google’s Contract DocAI ever get to the fabled perfect accuracy? The short answer is: no.
‘With human-made contracts AI can only be approximate,’ he noted, and added that with ‘less mission critical’ documents then accuracy challenges are sometimes not so major. But, for the complex matters, such as the contracts that Ironclad is handling ‘they can take our models and run our AI, but Ironclad also has a human in the loop’ who can check things and provide ‘a second level review’.
So, there you go, not even Google thinks we will be facing ‘the end of lawyers’ any time soon.
For readers in the legal sector what this all means is that you will still be working very closely with legal tech providers, there may just be ‘Google Cloud – Contract DocAI Inside’ for some of those products in the years to come.
For the legal tech companies, there is an interesting choice now: stay with their own adapted NLP models (which are usually based on well-known language models that are open to all), or try and build a partnership with Google?
Having Google Cloud provide the core components of your offering may reduce costs in the long run, it also then allows you to focus on really super-detailed and very specific use case training, and also to build great UI and UX that sets you apart from your peers.
Conversely, some companies may say that their core offering is their NLP capability – if you outsource most of that to another supplier, then what do they have left to offer customers? Perhaps the answer would be their specialised training that sets them apart from others.
For now, Ironclad and its CLM work is where the focus is, but, as Valloppillil said, there will be more partners to come. Watch this space, there are interesting times ahead for the contract analysis world.
(And thanks to Google for providing the slides about Contract DocAI.)