Ironclad took the decision in 2017 to put on hold its own NLP development projects and then ‘wait until the technology caught up’ to the level they needed. Now, through their partnership with Google Cloud AI, the CLM company believes they have a game changing advantage.
As CEO and co-founder, Jason Boehmig, told Artificial Lawyer: ‘Cai, our co-founder and CTO, said ‘Trust me, just wait. Google will build something better than anything we can and then everyone [in the market] will have to start over.”
The NLP system is used by Ironclad to help clients with old contracts that need to have data extracted from them. For contracts generated within Ironclad these documents are already digital and so don’t require additional NLP data extraction.
This site mentioned a few well-known legal AI brand names and noted that some had been working on their NLP models for ten years. Could Google Cloud AI really be better?
Boehmig responded: ‘Much of what they have under the hood has been made generically now with Google Cloud AI.’
He added that language models from Google were really excelling for general language and they’d been able to build models for legalese very rapidly that were highly accurate.
He also noted that they are the only CLM company, as far as he knows, that is working closely with Google to use their AI technology.
And, in a blog post from last year, Cai GoGwilt, Ironclad’s CTO mentioned above, stated: ‘In 2017 we started experimenting with [NLP]. Unfortunately, the feature development was too slow. So, we put NLP on the back-burner and waited for the technology to catch up.
‘[During the pandemic] we discovered Google Cloud AutoML Natural Language. We started with AutoML’s Entity Extraction model. First, we uploaded a small, curated set of contracts and labeled three properties: entity name, signature date, and signer name. After a few hours of training, signature date had precision and recall rates surpassing 90%. This was the best result we’d ever achieved over three years of on-and-off experiments – and, incredibly, Google needed a relatively tiny data set to achieve it.’
More work followed, accuracy went far higher, and as Boehmig noted, they are now getting results that they are happy to accept with just the smallest human input.
And GoGwilt continues the story: ‘A few months later, we’re in alpha with a handful of customers. The feature works on an increasing number of key data properties with some accuracy rates exceeding 90%.
‘Yet even 90%+ isn’t good enough in the world of contracting. That’s why the feature also enables users to deliver the last mile of data accuracy themselves, aided by an intuitive data validation flow with human reviewers.‘
This new function is called Smart Import and it can be used by the CLM company’s clients to extract key data from contracts, which then can be used to keep tabs on contract values, parties, renewal dates and other business/law cross-over information.
Now comes the multi-billion dollar question: how much ‘human in the loop’ input is required, or just sensible to have?
‘Google Cloud AI is good enough,’ said Boehmig. ‘We will have humans in the loop to browse and check off, but we could do this automatically. We have 14 fields now (i.e. data fields) that we feel confident in.’
‘It’s game changing,’ he stated.
At this point Artificial Lawyer had to ask if Ironclad, although a CLM company, would consider eventually moving into M&A due diligence work? After all, if they felt so confident in their NLP capabilities now, then why not do this as well one day, (albeit that clause identification is a different use case to specific entity extraction)? Boehmig said that the focus remains on CLM and solving the key business problems around contracting.
He added that their recent purchase of PactSafe, a click wrap agreement acceptance system, shows where they are headed. They want to be present wherever a company might be contracting or engaged in using data embedded in contracts.
That said, as Boehmig noted, he’s been interested in NLP for a long time. In fact, when he was an associate at California-based Fenwick & West, before Ironclad began, he had experimented with IBM’s Watson. And even before that, as an analyst at Lehman Brothers in New York, where he worked for four years before the bank famously collapsed in 2008 during the financial crisis, he witnessed MIT coding experts sitting down with traders to better understand contracts.
So, one can see the relationship with Google as something of a coming of age for an idea that Boehmig has been thinking about for a long time. But, now that they have Google Cloud AI in their corner, they believe they are ready to truly wield NLP technology effectively and in turn will have a competitive advantage over other CLM companies.
It’s going to be really interesting to see how the clients react.