Zuva Document AI is a set of APIs that developers can use to build their own AI applications to extract key information from contracts and other documents. I spoke to co-founder and CEO Noah Waisberg about the company’s development, use cases and wins post its launch in 2021, when Litera acquired Kira Systems, also co-founded by Waisberg. Zuva retained Kira’s technology as part of the spin out.
First things first, what does Zuva mean?
Kira means ‘ray of light’ in Sanskrit. We wanted to find something similar. Zuva means sun – which is why our logo and t-shirts are yellow – in Shona, which is a Zimbabwean language. We thought it was appropriate to choose a different language but the same meaning, because we’re doing the same thing in a different way.
How are you doing things differently to Kira?
We’re trying to be more aggressive. At Kira, when a company gets to a certain size you have a lot built up and you get more conservative in your decision making. We are trying to be less conservative.
So what does less conservative mean in practice?
With Kira and pretty much every legal tech company, you can get a lot of information, but you can’t go on the company’s website and find all the information you need about buying the product, including the price. That’s good in some ways because it forces people to talk to the salespeople. But for us, if you’re trying to provide a good buying experience, we say let people find everything out and decide if they want to talk to us. You can sign up online even with a Gmail account and there is tons of information that we wouldn’t have put out with Kira, including our prices.
Another thing is that we’re ready to try something and then iterate on it, as opposed to creating a programme or heavily preparing for something, now we just do it. It’s good for employees because we can move faster, although there will probably be times when things don’t work out. We’re trying to do smaller things and iterate on them as opposed to just big things when it comes to going to market. We would rather get a feature out that’s pretty good and then iterate on it.
Can you give some examples of the work you’re doing with Zuva?
At Kira people told us that our AI was the best, but then we realised that a lot of people used our AI with a different interface. Some people took Kira as the back end of another system. Kira was built to be a workflow tech but wasn’t perfectly suited to being the back end of another system. Before we spun out, we started separating the AI from the user interface. By the time we spun out, we already had a beta product that was in use by a Kira customer, so we were pretty sure there was a level of interest. What Zuva does now is sell the underlying technology in API form only, so there is no real user interface and you can only plug it in to another system.
Who is using the technology?
A couple of leading contract management vendors have built us in. Every system says it has AI in now and some vendors build it themselves but that is a path to mediocrity because it’s hard to build well. We spent over a decade building our technology and it has over 1300 built in models. We’re exploring with a leading document management system consultant how our technology can be used to improve search. If you want to improve search you can enhance the metadata on documents that people search and you can do that by getting people to add it or using technology. Using technology hasn’t worked well but our tech does work.
Heretik, which just got bought by Relativity, has also bought our technology.
We have a new feature coming out that provides multi-level classification and can automatically identify around 300 document types. It understands if it’s a credit agreement or employment agreement or a NDA and it auto-classifies it. You can teach it do more, but it will auto find those 300 data points. You can run it for historical agreements or if you are creating a new document, when you’re creating it, it gives you suggestions.
To find out more about Zuva visit the website at: https://zuva.ai/docai