AI Transformation in Enterprise is Imperative

Ganesh Harinath, VP of Engineering, AI Platform & Applications, Verizon Media
Ganesh Harinath, VP of Engineering, AI Platform & Applications, Verizon Media

Ganesh Harinath, VP of Engineering, AI Platform & Applications, Verizon Media

The AI revolution is poised to be the next biggest innovative revolution after the internet and is already disrupting how we build, travel, secure and live. In some ways, what was once science fiction is now just becoming science, with AI already being democratized through software as service offerings. Companies like Salesforce and Oracle are enabling their AI prediction and forecasting capabilities for a host of customer scenarios.

Churn, customer experience, personalization and demand prediction are some areas where AI enablement has the potential to help improve efficiencies and increase revenue for enterprise, but, frankly, the applications are limitless. Without adoption, companies will risk losing competitive edge. In order for sectors of our economy to capture the power of AI and to implement it efficiently and responsibly, we need to be thinking now about how best to architect its application. To this end, AI transformation in enterprise is imperative.

Looking back to look ahead

Application of AI using computers started to evolve in the 1950’s and was experimented on the first commercial computer named Manchester Ferranti. Arthur Samuel built the first machine learning based checkers program in1952 but hardware requirements and the unavailability of distributed technology hindered its proliferation. Fast forward 60 years and the story is very different. AI can now be operationalized at scale in a cost effective manner.

Hadoop (big data technology developed at Yahoo) was one of the foundational elements to persist and process data cost effectively. Availability of open source technologies like Hadoop, evolution of big data technology and reduction in storage, compute RAM and networking costs helped trigger the Big Data revolution and served as an important precursor to today’s AI revolution.

  Churn, customer experience, personalization and demand prediction are some areas where AI enablement has the potential to help improve efficiencies and increase revenue for enterprise, but, frankly, the applications are limitless​ 

My understanding of the power of AI and the importance of its adoption is best exemplified by my firsthand experience building scaled platforms for AI capabilities. While I was at Zynga in 2011, I was responsible for operationalizing a scaled platform used to ingest 25 terabytes of data a day from dozens of data sources and that surfaced insights about the health of the infrastructure, game applications and security posture. The value of centralizing this data platform was astronomical. The scale issues to ingest tens of terabytes of data across multiple products were addressed in one central platform. In so doing, data products could be managed with a single technology interface, infrastructure efficiency was created and there was correlation across multiple data sources, avoiding redundancy of data sources in multiple platforms. There was huge cost savings and improved developer efficiencies by centralizing the platform and standardizing the technology across multiple AI product development initiatives.

This experience leads me to believe and build a central petabyte scale AI platform to launch multiple AI-based data products and solutions for Verizon in 2013. Dubbed Orion, the platform was operationalized in 2015, and it has been evolving since inception to help deliver several AI products and capabilities while injesting more than 75 billion records a day.

Basics of starting AI transformation in an enterprise.

AI is nascent enough that we’ve yet to realize hard and fast rules for deployment. As it develops, we should expect implementation best practices to evolve. However, there are some important things enterprise organizations can consider today when operationalizing AI-capable platforms.

Articulating and committing to a strategy is crucial. It’s expensive and inefficient to have multiple data and AI platforms distributed within the enterprise, so a centralized deployment tends to yield the most benefit. Embed in your company culture the value of the strategy and work to achieve buy-in from key stakeholder groups, including at the C-suite level. At least since the adoption of Hadoop, Yahoo engineering culture has been oriented toward centralizing data and AI capabilities as much as possible, and this is embraced by our engineering community as best practice. With Yahoo technology powering Verizon Media, we have an even bigger and growing platform for delighting Verizon customers.

You’ve set the foundation of the house with a strategy, now it’s time to frame it out. This starts with robust, horizontally scalable infrastructure, including ample compute, storage, RAM and networking capabilities. It’s important to build fault-tolerant distributed data storage capacity (e.g. Hadoop) that can ensure high-volume data flows with granular access. Operationalize frameworks like Kafka or NiFi to easily move data in a secure and controlled manner. Enable technology stacks like Spark help to process data. Enable the AI frameworks like Tensorflow, Torch, and Keras that can best serve your AI needs. Develop mechanisms that facilitate security, privacy governance and identity management. Finally, business capabilities can be enabled through secure API layer.

Teach your AI to be ethical before it’s too late. AI technology only knows what you give it and, without safeguards, it may extrapolate or infer in unintended ways. Investing in a set of ethical standards for AI will help to mitigate the risks. Controls should include technical standards, granular controls, enforcement and a governance framework. And have a philosophy: what is it that you hope to achieve for society when deploying AI?

AI is not a product, nor a platform, but it’s a new way of architecting modern day solutions and products with AI as the brain. With those stakes, make sure you have the right talent on the job. It’s not enough to build a framework for AI deployment. Having skilled technologists who understand the underpinnings of AI will help to maximize its effect on the organization and the stakeholders you’re working to serve.

Business groups should be encouraged within the enterprise to help build AI-powered products or solutions using a central AI platform. AI transformation is most likely still a multi-year journey, but once executed and, if done correctly, the results have the potential to be truly groundbreaking.

Read Also

Getting the Most out of Big Data

Getting the Most out of Big Data

Kris Rowley, Chief Data Officer, GSA
Big Data: Separating the Hype from Reality in Corporate Culture

Big Data: Separating the Hype from Reality in Corporate Culture
Brett MacLaren, VP, Enterprise Analytics, Sharp HealthCare
Maintaining Maximum Relevancy for Buyers and Sellers

Maintaining Maximum Relevancy for Buyers and Sellers

Zoher Karu, Vice President and Chief Data Officer, eBay
Building Levies to Manage Data Flood

Building Levies to Manage Data Flood
Adam Bowen, World Wild Lead of Innovation, Delphix