Advanced Operating Models
Dec 05, 2018

Our experiments with "deep learning"...

Raymond Kurzweil, the author and a futurist, proposed what came to be known as “Law of accelerating returns" in 1999. Simply put, it says that the rate of technological advancement is exponential against “intuitive, linear" view. We have now entered an era where this law is more than evident, and at the forefront of this revolution is artificial intelligence and deep learning. Gartner's Technology Hype Cycle has pegged “Deep Learning" at the “Peak of Inflated Expectations" two years in succession now – enough validation that there will be no business that will not be transformed by AI in the near future.

But, as is true with most things in life, there is hype and then there is reality, and the secret sauce lies in finding the boundary between the two. At my organization, we've also been experimenting with some of these emerging technologies, and I want to elaborate on one specific use case that has shown us the impact we can generate by leveraging AI and overlaying domain expertise on top of that. We experimented with a branch of communication science called Organizational Network Analysis (ONA) to understand communication patterns of a sample of our leaders.

We started with partnering with MIT researchers for their deep expertise in this field. They helped us understand the basics of ONA and how we convert raw communication metadata into meaningful metrics. We then put all of this into a machine learning model along our existing data and trends on performance and attrition. There were interesting patterns that emerged: we could figure out distinct differences between great and not-so-great performance based on specific communication traits. We could also create a persona for a successful leader within Genpact, which helps us integrate new leaders faster and more efficiently into the system. Our attrition analysis helped us understand lead indicators that would tell us who are the most disengaged/likely-to-quit employees.

As a last step, we took all of this intelligence to HR leaders and operating leaders, and it corroborated with their subjective judgment about their own employees. Interestingly, there was a lot of overlap between what the machine was suggesting and what operations/HR had assessed. This is both exciting and powerful, because a ubiquitous data source such as communication patterns can reveal so much about our people, without many people putting in time and effort. This is a big step toward “human – machine" partnership, in which intelligence from machines will be augmented with human wisdom to unlock the value hitherto unknown to us. This is also a great validation of our belief that tools alone cannot help us step into future – we need years of domain wealth and a curious mindset to explore the “art of the possible." The journey has just begun for us, and we are excited to be a part of the roller coaster ride ahead.

This blog was co-authored by Piyush Mehta, chief human resources officer, Genpact; Indira Sovakar, senior vice president, human resources, Genpact; and Praful Tickoo, assistant vice president, HR Analytics and Innovation - practice lead, Genpact. This article was first published in LinkedIn in September 2018.