In this episode of One Quick Thing, Jerry Gupta discusses how his team at Swiss Re explores new technologies and business models. He also discusses his approach to working with Big Data and artificial intelligence. Gupta is the Senior Vice President and Digital Catalyst at at the Zurich-based reinsurance company.
Three takeaways from the conversation follow.
Don’t Be Seduced by Big Data
“Big Data projects are sort of seductive. And people tend to have unreasonable expectations,” Gupta said. “85 percent of models don’t even get deployed.”
When discussing the challenges of working with Big Data and AI, Gupta noted the lack of proper infrastructure at large organizations. “A lot of models that I’ve seen, which have a very high degree of accuracy don’t get deployed because you don’t have the right infrastructure…[or] you can’t explain it to regulators,” he said.
Another challenge is factoring for a large number of variables. Gupta pointed to a project he worked on in a previous role, where AI was deployed to improve customer satisfaction. “When you have too many variables…[in] an operational environment, it’s really difficult to deploy that solution,” he explained. “For example, if your job is to respond to customers, and you have to tweak 60 different things, it is really difficult. You’ll probably fail in that exercise.”
Best Practices for Working with Big Data and AI
When asked for best practices, Gupta suggested to focus less on the data model itself but more on the need the model seeks to fulfill. “What is the objective? What is the decision you need to make?” He asked.
Gupta also emphasized the importance of setting a path to deploying new Big Data models. According to Gupta, that involves communicating the value proposition to get buy-in, building the right capabilities to manage the data, and navigating regulations.
Before building a new model, Gupta recommended that teams first look at already existing APIs (application programming interfaces) to see if already existing interfaces will solve the issue. “[If you can,] use an API. Save yourself a lot of problems,” Gupta said. “And then if you really want to build something, see if something already exists that you can repurpose, and if none of those are [available] then you start from scratch.”
Proving the Value of Intangible Assets
When it comes to innovation and new ventures, many assets — like the value of networks or connections — may seem difficult to measure. “For folks like us who are on the early-stage business side, tangible KPIs actually are a detraction as opposed to a big help,” Gupta said.
To prove value, Gupta instead focuses on milestone-based metrics, including the number of projects the team is working on and how many projects have been killed. He also advocated for looking at new ventures at a portfolio level.
“Don’t look at projects in silo but or as individuals, but group them in a portfolio and then we track it by portfolio. Say, ‘This is the portfolio we’re working on. We’ve got five projects on it. This is how it’s tracking,'” Gupta said. “Nine out of those 10 projects are going to fail, but that one project is going to have some benefits.”
