Chatbots have been at the forefront of efforts to automate legal services and internal knowledge repositories. But building and deploying these services can come with their fair share of welcome, and unwelcome, surprises. At the "Optimizing the Role of Chatbots in Legal" session at Legalweek New York, those from all corners of the legal industry discussed what they learned from launching their own internal chatbots. Here are some highlights from the talk:

There's a Reason It's Not Called a Talkbot

Law firm staff may be taking chatbots too literally. Flyn Flesher, knowledge management counsel supervisor at Ogletree, Deakins, Nash, Smoak & Stewart, noted that when his firm launched an internal chatbot, anyone in the firm was able to ask it questions via email or through a dedicated website. What he soon found out, however, was that many preferred to, well, online chat with the chatbot.

"[What was] surprising to me was the extent to which people used the web-based version of the chatbot rather than sending emails to it. … The email [version] was down for a week and no one noticed expect for me."

Jim McKenna, CIO at Fenwick & West, had a similar experience when his firm launched its internal voice-enabled chatbot called Fenni in 2018. "A surprise for us was that we initially thought people would want to use their voice to interact with it, [but] what we found is people would type out their questions."

Constructing a Chatbot Isn't Hard

Building a program to answer and retain queries from internal staff can seem like a tall order. But those legal teams that have done it in the past know firsthand it is anything but. "First surprise was how easy it was to program it … it maybe takes one month to make a chatbot with very limited features, and everyone can do it now," said Vitaly Vinogradov, CEO of Amulex, a legal services provider in Russia.

Ogletree Deakins's Flesher noted that he started working on a proof of concept for his firm's chatbot in early January and had a working prototype by the end of March. "Essentially very little has changed with the underlying code from [March] … and that was 30 hours of coding from someone who has not done serious coding in 15 years."

There's a People (and Content) Problem

To be sure, creating a chatbot is not without its fair share of challenges, which oftentimes stems more from people and content challenges than technical ones. "The easy part of it was the technology, the hard part of it was getting everyone to agree this is the single source of truth" and keeping that source updated, said McKenna, of Fenwick & West.

Indeed, even though Flesher noted how relatively easy and quick it was to develop a chatbot, he added that filling it with useful content was a whole other affair. "Gathering the questions and answers, that's a multiyear ongoing process."

Francine French, lead e-discovery Manager GSK, also ran into the same challenge with her company's development of an internal chatbot. "The hardest [thing is] keeping up with the questions and answers and new feedback."