Enterprises looking to get their AI efforts off the ground face three core challenges, Genpact Chief Digital Officer Sanjay Srivastava has argued in a new video.

A digital professionals services firm with a global reach and over 70,000 employees, Genpact are perhaps best-known for their work in implementing digital transformation within one-fifth of Fortune Global 500 enterprises. As Chief Digital Officer, Srivastava works at the frontline of Genpact’s digital platform Cora, which provides enterprise-strength digital, analytics, and AI capabilities in a single modular service.

“We’ve been applying artificial intelligence in our clients’ environments for the last few years now, and the key learnings for us have been three very critical, practical considerations across every business environment,” says Srivastava. These include explainable AI, low data-density environments, and what he coins the ‘search for richer knowledge graphs’.

Explainable AI

Much has been made of explainable AI, which loosely refers to the integration of transparency mechanisms within AI systems that enables users to see in detail the reasons why an AI arrives at a certain decision. A lack of such transparency, commonly known as the ‘black box’ problem, can often produce the desired outcome but in an opaque fashion. It’s one of the key factors leading to bias in algorithmic decisionmaking, and a major obstacle to public trust in AI. For enterprises moving forward with AI today, explainability is key.

“Explainable AI is about answering the question ‘why?’,” Srivastava says in the video. “Every single time artificial intelligence makes a decision, we have to be able to trace that decision right back to the source and be able to present it in an explainable, traceable fashion.”

Low density data environments

Data, data, data. It’s what is fuelling the technological revolution underway today, and is growing exponentially as new software and hardware generates ever greater datasets. In fact, by 2017, humanity generated more data than in the entirety of human history put together. It’s part of the reason AI has been able to come on in leaps and bounds in the past 5 years.

What happens to AI when you don’t have that much data? One of the key challenges facing enterprises is access to datasets of the appropriate size and scale to make machine learning work. Genpact works closely with organisations to design and develop AI-powered software which can thrive in low-density data environments. “The whole idea of applying AI in low-density data applications has to do with the fundamental challenge of making AI work when you don’t have lots and lots of information,” Srivastava explains in the video. “When you work in enterprise environments, you don’t have access to large-scale datasets, and you have to make artificial intelligence work in low-data density application areas.”

Richer knowledge graphs

Gaining simple but clear insights can be a huge challenge. Diving deeper, however, is a whole different story. When analysing a whole industry, there are countless variables and inputs to keep track of, which makes it harder and harder for enterprises to arrive at strategically valuable insights.

“The depth of knowledge graphs currently available in the industry is a large consideration around applying the right AI capabilities to the right enterprise problems,” says Srivastava. “The challenge there is, simple questions, simple lookups, based on intent analysis are very easy, but when you come down to the nuances of a specific industry, then being able to apply a richer knowledge graph and embed all of that knowledge in a way that drives better applications of AI is really critical.”

Genpact is a premium sponsor of The AI Summit London, June 13-14. Catch up with Sanjay and the team at their booth or on-stage at their keynote or panel discussion. Find out more about how you can attend