In a Packed Room at EDI 2019, There Are Signals That AI is Maturing

In a Packed Room at EDI 2019, There Are Signals That AI is Maturing

The new PwC UK Law Firms Survey has just published a landmark statistic: 100% of top UK law firms are using or piloting AI. It’s the first time we’re seeing the figure of “100%” in the context of AI adoption. “AI has started to mature," the study concludes. It’s no wonder we packed the room last month at EDI. The audience was there to sort all the generalized excitement around AI and seek practical ways to bring it in.

Our panel was titled: “Using AI to Protect and Find in the Age of Sensitive Information.” Moderated by our Co-founder and COO, Omar Haroun, we were joined by Josh Kreamer (AstraZeneca), Joel-Henry Mansfield (Nationwide), Dayna Munsch (U.S. Bank), Joanne Sum-Ping (3M), and Shelli Toletti of Tesla.

EDI Blog Post body

There has been a perception that outside counsel are behind in-house counsel on the AI adoption curve. But as we’ve written, top law firms are already tightening their embrace of AI, now that their clients—and in particular, their Fortune 200 clients—have experienced the results of AI for themselves. As one of our panelists said: “We used to ask our law firms: ‘Is this defensible?’ Today we ask: ‘How can you make this defensible?’”

At Text IQ, we’re working on unsupervised machine learning that can overcome the inherent limitations of technologies like TAR and predictive coding, where humans manually create a seed set that is then used to make predictions about a much larger set. Even among ediscovery vendors who do use artificial intelligence, many of their models use supervised learning trained on small seed sets.

As AI professor and Text IQ advisor John Paisley wrote, these supervised techniques carry questions and challenges around bias and accuracy. At our panel, we heard from litigators and heads of ediscovery who have used unsupervised machine learning in their workflows to reduce the bias and the inaccuracy that supervised techniques have traditionally exposed in document review. 

We also heard about the underlying similarities between using AI to find privileged, confidential, and reputation-damaging in the context of litigation; and using AI to find personal information like PII and PHI in the context of privacy. One audience member fancied: “Will the Head of Ediscovery end up being the Chief Privacy Officer’s new best friend?” 

These similarities are why our core mission at Text IQ—building AI for Sensitive Information—will never change, even as we expand into new areas like Privacy, Security, and Compliance.

Finally, we heard panelists contextualize AI adoption within the ease of its implementation. We recalled the early days of cloud implementations, five years ago, when highly regulated companies were reluctant to move data to the cloud. Now that’s become the norm. Soon, a panelist stated, AI implementations will be the new norm; only, unlike cloud implementations, an AI project can be completed in weeks, not years.