AI Startup Turns Data into Narrative — and Compliance Reports

Register now

What started off as a Northwestern University software project to automatically generate baseball game recaps has transformed into a company that codifies stories for publishers like Forbes and personal financial management (PFM) providers like Personal Capital. Banks can use the technology to generate compliance reports and personal finance updates.

"Ten years ago we might not have had a business," says Stuart Frankel, chief executive at Narrative Science. "The technology timing is important." The market would not have been ready at that time, nor was the amount and quality of data sufficient. Today, "there's an inevitable ubiquity to what we're doing," he says. "Every data set can automate a narrative."

The broad idea of Quill, the name of Narrative Science's artificial intelligence engine, is to add context to a company's numbers through a computer-written narrative, all the while freeing up an enterprise's staff from template-driven work. "It's not Mad Libs," Frankel says. "The data impacts the story. …It's content for an audience of one."

Narrative Science sells to a variety of industries, but identifies the financial services category as one of its core markets. The startup's algorithms, according to Stuart, can offer financial firms an array of applications, including automatically generating text for internal documents, such as compliance reports, as well as writing personalized portfolio summaries for end customers.

At Personal Capital, for example, the engine will generate personalized stories for mass affluent users, who link an average of 16 accounts to the service. Through the addition of Narrative Science's algorithm, Personal Capital will offer more personalized insights and advice to its customers, such as recaps of how a user performed relative to certain markets as well as peer groups, in the coming months.

Brevity, however, remains key. "We're not writing a novella," says Jay Shah, COO of Personal Capital. "We're just saying what's important."

The engine behind the stories Narrative Science produces for customers of customers like Personal Capital comes with a malleable writing voice and can write tight prose such as a headline or a caption for a pie chart, for example. To date, Quill can only write in English.

The need to translate data into meaning is on the minds of most business execs. "We are very impressed with Narrative Science," says Shah. "We think they are onto a big idea and into a big space."

This is literally true. The three-year-old company is getting ready to move into a larger office space for the fourth time this spring. The firm employs about 40 people and is looking to hire eight more.

As Narrative Science grows in size, the culture naturally evolves with the addition of personnel. "You can't have the same culture with 40 people as you did with five people," says Frankel. "It doesn't make sense. The culture will change again when you get to 100 people. …Instead of fighting it, let the culture evolve."

For now, there are no restrictions on vacation time. "It's not about hours, it's about quality of work," Frankel says.

The Narrative Science team, as Frankel describes it, is ambitious, with many boasting advanced degrees in fields like engineering, neurobiology and computer sciences and talent coming from Northwestern and University of Chicago.

The process for improving the product never ends for the team. "You don't sprinkle innovation like salt and pepper on something existing," says Frankel. "It's the kind of thing where we don't know if we made the right decision until we're down the road. …We've gotten as far as we've gotten because of making little decisions."

Roger Lee of Battery Ventures invested in the startup, dazzled by the software's capabilities. "It felt like magic," says Lee. "The product, on a simplistic level, helps any human interacting with data generate a story.

"Narrative Science has grown up for sure and turned into an exceptional business," Lee says. "The most tangible thing is the amount of customers and variety of use cases in place with the product."

For reprint and licensing requests for this article, click here.