Can AI Avoid The 'Trough Of Disillusionment' In Legal Tech?

Buyers have figured out that AI isn't a handful of magic beans. Now what?

What a difference six months makes in an industry. Well, maybe not the legal industry, where every quarter delivers exactly the same outlook as the quarter before, but the legal technology sector can really experience some big changes in the span of a few months. Back in August, the International Legal Technology Association convention required an unofficial side conference to discuss this “blockchain” thing that no one fully grasped yet. Today, I could generate a couple of million in few hours if I announced a startup focused on moving deposition defense to blockchain. That doesn’t even make sense and yet you feel that distinct tug at the back of your brain to just hand me $10 million, don’t you?

But the biggest difference over the last six months is the shift in discussions about artificial intelligence. In August, artificial intelligence talks were roughly evenly divided between people who described a sophisticated tool that takes input from attorney behavior and learns how to perform tasks quicker and more reliably than humans could and the people who think AI is magic.

Times have changed though, and the mood of the legal tech world seemed decidedly more circumspect when describing AI at this month’s Legalweek show. It all started before the show even kicked off, with an AI boot camp that rehashed some of the 60-year-old debate over artificial intelligence vs. intelligence augmentation. While there are serious distinctions between the two philosophies, it certainly felt like the resurgence of the IA terminology arose as a backlash to the excesses of AI sales rhetoric.

However, the need for an alternative lexicon to describe this branch of technology may be unnecessary. Walking through the halls at Legalweek, there was a palpable sense that the evangelists of bleeding-edge technology had recognized the era of science fiction had ended. Perhaps it made sense when buyers couldn’t wrap their heads around artificial intelligence as a concept, but the market is older and wiser now. AI is a reality to them. They see what is and is not possible today and they don’t need hyperbolic visions of the future, they need a discovery answer today.

Vendors seem more than comfortable shifting back to grounded conversations. My conversation with Dan Carmel of iManage covered a lot of ground, but on the AI point specifically, he said the goal for companies over the short-term will be navigating the downturn in hype over AI. For those unfamiliar with the Gartner Hype Cycle, new technologies ramp up in a predictable manner before falling into a well and then slowly clawing their way back up as disillusionment gives way to acceptance. The folks at iManage, leveraging their years of experience and goodwill in the knowledge management space, want to avoid the “trough of disillusionment,” as the Gartner model calls it, with AI that provides reliable, practical AI through their platform.

Staying ahead of the hype pushes sales teams to the sidelines and puts the ball back in the court of the hard-working developers and scientists. Chatting with Catalyst’s data scientist, he talked about the years of work that’s gone into building and refining their engine. He stressed that the most helpful tip for potential clients is understanding that purchasing from a vendor has to measured in relative terms. To understand what they’re getting, they can’t just accept market-polished statistics and they definitely can’t run side-by-side tests using dummy data like the convenient, but now ubiquitous Enron data. The strength of an engine can’t really be measured in an artificial test. How will it respond to your data? How will it function in your workflow? By its nature, building AI can’t be a static project — mastering an existing data set isn’t much use to anyone — and buyers should test products in a genuine setting to make the best decision.

That’s why AI vendors are working constantly to improve their offering.

Sponsored

Though one company that’s less concerned about building the best algorithm is Veritone. But they have a good reason for stepping outside the AI development arms race because they think clients should be able to take advantage of multiple engines simultaneously. Nikki Black discussed Veritone’s video tool that offers facial, object, and sentiment recognition which is, undeniably cool to see in action, but underneath that, Veritone offers clients is a system that runs searches using a number of different, fully vetted AI engines at once. All searches are different, and different engines adjust to new challenges differently. When I spoke with Veritone reps at Legalweek, they described a product that learns from other AI engines in its environment and delivers the best result for a client’s needs. In a sense, this makes Veritone artificial intelligence on a meta-level — they’ve functionally got AI learning from AI. It’s a cool concept for those that see ultimately embrace the wisdom of the (well-curated) crowd.

If the Gartner model is right, we’ve passed the point of no return when it comes to AI expectations. And yet it’s unclear how deep the trough of disillusionment will run. If Legalweek is any indication, major players in legal AI have already begun to manage expectations and have the realistic conversations with buyers that are supposed to come years into the hype cycle. Perhaps they will manage to stave off disillusionment with some well-timed realism.

Time will tell, but there’s a chance the robots have hacked the hype cycle.

Earlier: The Artificial Narrative Of Artificial Intelligence
Show Me Your Intelligence Augmentation
Running With The Robots At Legalweek


Sponsored

HeadshotJoe Patrice is an editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter if you’re interested in law, politics, and a healthy dose of college sports news.

CRM Banner