Read

Reliability 101: How to stress-test GenAI for legal and compliance

Watch the full chat with Ali Cook and Jake Brown.

 

How do you know your legal and compliance GenAI tool is reliable?

It’s a critical question for anyone looking to leverage GenAI today in a high-risk area like legal and compliance. If you’re going to offload high-volume FAQ work to self-service tools, they need to be dependable and keep risk to a minimum.

In this fireside chat, Josef Co-founder and COO Sam Flynn talks with in-house legaltech experts Strategy Lead Ali Cook and Head of Legal Technology Jake Brown. Together, they share insights from their work with Josef customers, and their reliability framework created in collaboration with Cornell and NYU.

They dive into questions like:

  • How do you know whether a tool is reliable? 
  • What does “reliability” mean anyway?
  • What challenges are legal and compliance teams up against with GenAI? 
  • What are their top tips?

Watch the full chat to find out, or jump straight to their tips below. 

Tip #1: Don’t sweat the small stuff

“Get comfortable with the fact that your tools may produce answers that aren’t necessarily phrased the way you would phrase them,” Jake says. 

“But that’s okay! As a former practicing attorney, I prefer my own phrasing over anyone else’s, but consider whether your tool’s phrasing actually introduces any risk.”

Sam adds, “It’s about bringing a healthy level of skepticism and practicality to your approach.”

“A GenAI tool isn’t going to include all of the elements that you or another subject matter expert would include, but think: 

  • Is the answer wrong
  • Does it introduce risk
  • Is it actually going to cause a problem

If so, of course, we need to address it and moderate the tool’s response immediately. If not, we might still want to address it—perhaps by capturing our tone of voice and our way of working—but it shouldn’t block us.”

“It’s your end users who are going to get value out of these tools, so you need to think about what matters to them.”
– Ali Cook, Strategy Lead, Josef

Tip #2: Remember who your tools are for: your users

Ali says, “I think in the flurry and hype of GenAI, people have forgotten to center end users and consider them early on.” 

There are lots of red-herrings in reliability testing, and by focusing on your end-users, you’re going to avoid a lot of them. 

“It’s your end users who are going to get value out of these tools, so you need to think about what matters to them. Consider: 

  • What do they actually need to know?
  • Is your tool’s answer actually helpful to them?
  • Does it truly answer their question?
  • Or, does it put them in a position where they might face a negative outcome?’”

Ali continues, “If you focus on what your end users really need when testing, you’ll be headed in the right direction.”

Tip #3: Bring a healthy dose of skepticism

That said, you do need to be cautious, particularly when testing a tool for the first time.

Talking about a recent assessment of a generic search tool that a customer was using alongside Josef Q, Sam shares: “On the surface, everything looked fine. Everything seemed hunky-dory.”

“The answers the tool was producing seemed accurate and beautifully formatted. It was referencing what appeared to be accurate documents. But, after digging in, we found full-on hallucinations.”

“When you’re using technology that hasn’t been built for high-risk settings like Josef has, you’re introducing a new level of risk, so be wary!”

Learn how Fnatic’s legal and HR teams use Josef Q

In 2023, Fnatic’s in-house legal team took their automation to the next level with Josef Q. 

Moving beyond document and workflow automation, they launched a suite of self-service tools to answer questions about contracts, handbooks, and a range of HR policies. They even connected them with Slack, ensuring business users can get answers right where they work and the legal team can track engagement in real-time.

Learn about Fnatic’s “game-changing” Legal and People Josef Q tools.

See Josef Q in action

Book a demo to see first-hand how easy it is for your whole team to get up and running with legal automation.

Thanks for requesting a demo!

We'll be in touch soon to arrange a time to speak.

Josef Q

Experience Josef Q,
our AI tool, in action

Book a demo to see the transformative power of AI to create digital tools that unlock corporate content.

Josef Q

Book a
Josef Q Demo

Josef No-code

See Josef's no-code
automation platform

Book a demo to see our no-code workflow automation platform in action and learn how it transforms your day-to-day.

Josef no-code

Book a Josef
no-code Demo

Thanks for requesting
a demo!

We'll be in touch soon to arrange a time to speak.

Get legal
innovation news

Subscribe to our newsletter to get regular news and
updates from the exciting world of legal tech.

Thanks for subscribing!