Many companies are finding meaningful ways to apply AI. One company in particular, Narrative Science, has been recognized as a company leading the revolution by using AI to translate large data sets into analysis via conversational text.
I sat down with the Founder and CEO Stuart Frankel to talk about the Chicago tech ecosystem and future of the AI industry.
Why did you choose to start your company in Chicago?
I am probably representative of most CEOs in that I didn't make a calculated decision as to whether I should start a tech company in Silicon Valley or Chicago.
I've lived here in Chicago for 20 years and what I've come to realize over time is that many Midwestern cities are incubating tech startups and creating environments that foster a healthy and rapidly growing startup ecosystem. I genuinely believe that good companies can be started anywhere, everywhere.
The big difference between cities comes down to a matter of concentration. It's a numbers game. What Silicon Valley has that many other cities do not is the sheer size of venture capital and talent density.
Have you talked to people who have left Chicago to start companies?
Yes, and they're usually younger or at least that's what I've seen within my network. It could be coincidence or an observation that younger people are predisposed to being able to uproot and go elsewhere more easily than others. There's a wanderlust that you have in your 20s.
How did you come to start Narrative Science?
We have developed an advanced natural language generation (NLG) platform that we call Quill, which is an intelligent system that turns data into natural language communication. That might be in the form of a document or part of some type of conversational interface, but it's expressing data analysis as language rather than as charts or graphs.
We started the company in 2010. Prior to that, the company was incubating at Northwestern University. I started the company with two computer science professors whose backgrounds are in AI. They had an interesting perspective, researching AI in its purest form for the entirety of their careers.
Like the Turing Machine from the 30's.
Yes, exactly. The professors I had been working with were researching how-to take data and turn it into intelligent narratives. They started with baseball data and built an AI that summarized games like an Associated Press reporter.
How did this tie into your background?
Previously I worked in ad infrastructure and technology at DoubleClick. I was with a startup that was acquired by DoubleClick in 2007, joined the senior team there and eventually we were acquired by Google in 2008. At the time, it was Google's largest acquisition ($3.2 billion).
Prior to that, my first job in tech was with Rolling Stone and I also practiced law for a while. I'm also a CPA and I worked at PwC for four years right out of college. I've done a bunch of different things and then found my way to tech about 20 years ago and haven't left.
Is Narrative Science the first company you've started solo?
It is. I've learned that as the founder, you can't appreciate the difference between being the first employee and the fifth employee until you start a company. You can't really articulate what that's like, unless you've been through it.
In 2009, I partnered with Northwestern professors to commercialize their idea. At that point, there was already a lot of data in the world and the market for business intelligence and analytics was just opening up, with many companies using dashboards and data visualizations.
We looked at this this trend and saw how exponential the growth of data was, and that it was going to accelerate even more and overwhelm us. At the time, we thought the available tools that allowed people to look at data to make decisions were not going to work.
We decided we can either teach the world to use yet another data analysis tool, application or we teach machines to communicate with us in our language, then we can really make data accessible and useful for people.
2010 was pretty absurd to start a company like Narrative Science. I don't know what you were doing in 2009 but you definitely weren't talking about artificial intelligence, you weren't talking about natural language generation. This was before Watson won Jeopardy. To further put this into context, in 2010, there were 16 artificial intelligence companies that received VC investments. In the last two years, there's been about 2,200 VC investments in AI, so we were early.
Was being early an advantage or a disadvantage?
I think it was a little bit of both. But like any emerging technology or any emerging technology market you have to find your spot and that can take a while. You have to kiss a lot of frogs, knock on a lot of doors because even if you have a breakthrough technology, it still has to solve a real business problem. There's the risk of creating an exciting piece of technology that people are intellectually interested in and getting a demonstration of, but are not really interested in buying. Over time and a lot of experience, we learned where the technology could be applied to solve real business problems.
Fast forward ten years, how will Narrative Science be using Artificial Intelligence?
We've always felt like we were building what ultimately will be a ubiquitous technology. Advanced NLG technology will continue to get widely adopted and integrated with all kinds of third-party software applications.
Over time, conversational interfaces will become more mainstream and become truly conversational, as opposed to the transactional nature that they have today. We believe we can play a significant role in powering these conversations.
Quill has domain knowledge, it has context, and it knows what it told you previously so it doesn't take much to extend that to some kind of conversation. Quill can generate responses that can be delivered through text or voice, and our business intelligence partners like Sisense, are integrating Quill with things like Slack messaging and Amazon Alexa.
So then Quill becomes a conversational agent for all products, not just data analytics. How much of that are you doing today?
It's a very small part of the business. There's intellectual interest and certainly curiosity. We're talking a lot about it with our customers, but the reality is that these conversational applications have not been widely adopted across the enterprise.
On the consumer side, these applications are much further along. The market will continue to move away from the traditional analytics experience where you're sitting at a computer and you're pulling a bunch of levers. There will be much more of a conversation. Asking your device or some application to give you just the information you need at the time you need it.
There is no way that we are going to live in a world where everybody has to be trained on the tools they use. I keep hearing that everyone needs to be data literate and I think that notion is absurd.
Applications like Quill can really step in and provide those skills and provide those activities that ensure that people don't have to become data literate and can play to their strengths and interests, helping them understand data and, most importantly, making decisions against it.
What's your biggest obstacle between where you are now and total ubiquity?
The market is still relatively small for all AI applications. For example, in 2016 the market was right around a billion dollars and its projected to grow about 60% a year for the next several years. By 2022 it's probably going to be a $30 or $40 billion dollar market. It's early from a market standpoint, and we have to manage that and invest at the right time to take advantage of the tailwinds that you get when the market opens up.
If you invest too little, you're not ready, if you invest too much you wind up spending all of your money before you get that lift and before you get to that market opportunity. We have actually done a very good job of striking the right balance.
Looking back, would you have done things differently and started the company somewhere other than Chicago?
The short answer is no.
I started the company with two professors at Northwestern and we have academic roots. A lot of our early employees actually came from Northwestern and I think that's an advantage--just in terms of how you get a company running quickly but also competitively. We have true domain experience that exists in very few places.
If we had started our company in Silicon Valley we probably would have raised $100 million dollars and we may not have been any further than where we are today. One of the benefits of starting a company in Chicago is that you get to stay under the radar a little bit. There isn't this need to broadcast to the world exactly what you're doing. With something new and emerging, it's pretty important that you focus on a few things in the beginning and you give yourself an opportunity to create something really right. I think you have fewer distractions when you're in Chicago than Silicon Valley. There are fewer events, incubators and VCs. With fewer distractions you can be more focused on building the business.
From a talent standpoint I can't imagine us getting better talent anywhere. The type of people that we've been able to attract on the engineering side has been incredible. People from the University of Illinois, the University of Chicago, Northwestern and Stanford. There are a lot of graduates from these schools that want to be in Chicago for family and lifestyle reasons.