Key Considerations for AI Deployments

Video produced by Steve Nathans-Kelly

At Data Summit Connect 2020, Elliott Ning, cloud advisor, Google, discussed pre-requisites for AI deployments and best practices for implementing them.

Full videos of Data Summit Connect 2020 presentations are available at

"Nowadays companies do not just talk. They are doing AI. So this new strategy is transforming the business and impacting the industry," Ning said. "In a few months, by 2021, the majority of the enterprise applications may already have, or are going to have AI features to help users better understand their data. The industry progressing from traditional tools into intelligent systems in terms of autonomous capabilities in search and control environments, in warehouse support, and of course, in public roles."

Citing research, Ning said only one 10th of developers today can create custom ML models. This is the biggest sector that's slowing the overall AI implementation in most organizations.

The ability of people to utilize innovative AI tools, if done properly, can ensure that more people can benefit from this revolution. So AI is rapidly transforming from hype into reality and it's entering the mainstream, he said.

"We all understand that AI is critical. Let's start using it. But it's not as easy as it sounds," Ning said. "As you look to bring AI into your organization, it's important to prepare for the challenges you have."

Although most enterprises have identified themselves as entities evaluating or using AI, research shows less than 14%, of AI projects are actually deployed into large-scale production.

"So what things do you need for considering AI? First, you're going to need a very good problem to solve in order to bring value to your business. Second, you will need a team that really understands how to create meaningful insight from your data. Three, you better have a large amount of data, ideally in organized format, because no data, no AI. Clean your data, then AI," Ning said.

In order to simplify the AI process, comapnies need to consider key factors such as use cases that produce large amounts of data for training. We need good use cases.

"I personally believe you can find great value from the tools available on the market, particularly in the cloud. You may wonder, why in the cloud? Because you would get the latest innovative AI toolings while having the benefits of cloud computing, which include high scalability, flexibility, pay-as-you-go, dev ops, and more," Ning said.