Graph as the Foundation for Data Analytics and AI

This is an abbreviated version of a presentation by Rita Sallam during the Fall 2021 Graph + AI Fall Summit.

Rita Sallam, Distinguished VP and Fellow, Gartner

Watch the full session from Graph + AI Summit Fall

My goal for today is to challenge you to think differently about how you can innovate using data. Specifically, of course, by creating connections across what we know is increasingly complex data. As part of that, I’ll share with you why, at Gartner, we believe that graph technology does form the foundation of how you might achieve that with data analytics and AI. 

Data and analytics, and AI have never been more critical – it’s just harder to make sense of a world with ever-more interdependencies, connections, and unintended consequences. And we have to do that faster than ever before, but importantly, faster than our competitors. 

The global Covid-19 pandemic has forced most of you to respond in new ways to previously unimaginable demands to reimagine customer engagement, optimize resources, reinvent processes, rethink products and business models, and even, in some cases, your very purpose. And do that literally overnight. And what’s required to succeed in the face of this, just this unimaginable set of market shifts, is the ability to leverage increasing velocity and scale of analysis, to respond to this sort of never-ending change in the world. 

So, the question is, how will you respond? 

Three Scenarios

I’m going to begin by sharing three scenarios, and I’d like you to think about what these scenarios have in common.

Firstly, we know that supply chain demand planning has never been more important and unpredictable, right in the current environment. Imagine you’re a car manufacturer with many different models of cars, car parts, some of which are shared across models, and some of which are not. We know that car manufacturers face shifts in demand but also massive constraints on supplies such as computer chips and supply and shipping capacity. So what happens when demand for a product segment is suddenly surging in a particular geography. The manufacturer will no doubt need to know Will we have all the supply parts we need to meet this demand? What are the supplier risks or when demand drops drastically?

Now having surplus, how can we best use these spare parts? What impact does changing a feature of a car have on profitability? All critical questions that drive profitability and customer satisfaction. Okay, so that’s one.

Secondly, imagine that you are a city municipality, and you need to deliver food to elderly citizens who can’t leave their homes during the COVID crisis. How do you determine the best routes through the city that optimizes delivery speed and the use of transportation resources, while maintaining social distance between food couriers and other city residents along the road? Or along the routes? 

And case number three, which is actually my favorite, you’re trying to track the impact of climate change on penguins. And who doesn’t love penguins? To optimize innervate intervention strategies? How can you track distinct penguins, their movements, their shifts in migration, mating longevity patterns? And how does this all relate to weather temperature and other changes in their ecosystem?

What do all these things have in common? They are using graph technologies and AI to drive innovations.

Some Predictions

One of the things we do at Gartner is we make predictions about technology trends.

We believe graph technologies will be used in 80% of data and analytics innovations, facilitating rapid decision making across the enterprise. And that’s up from 10% in 2021. 

So why do we believe this is the case?

Well, first graphs enable agile data delivery by forming a platform that bridges data silos, captures metadata and data to extract abstract away from ETL processes. We know that knowledge graphs are a key component of that of data of a data fabric, exposing metadata, exposing business rules, enabling data scientists to quickly identify and use the data they need while preserving context, which is key.

The next thing is that graphs are relatable. They enable business collaboration because they represent data conceptually using objects and relationships in the same way that we think about the world. And a knowledge graph schema, or ontology, provides a unified vocabulary or lingua franca across technical and business teams, and business rules can be encoded in a knowledge graph to provide contextual information and sentence semantics.

From an analytics and AI perspective, graphs enable you to use algorithms semantics to derive new insights and patterns from existing data. Graph algorithms allow users to explore complex data in an intuitive way to ask questions that were really difficult to ask before if they could ever ask them before. Graphs also increase the predictive power of existing data current AI machine

Graph visualization allows for free-form exploration of graphs in order to speed up the investing investigative process and reveal big picture insights. And lastly, running supervised and unsupervised learning models and graph databases can also reduce training time.

And last, but not least, graphs are explainable, and are beginning to be used to enable explainability of deep learning models. Explainable AI combines knowledge graphs with graph-based machine learning to provide visibility into black-box models. 

So graph plus data and analytics and AI and cloud are making innovative use cases possible at scale across most industries.