How AI is changing the workplace, now

Experts from Hudson’s Bay Company, Constellation Research and IBM Watson share their insights on AI

Is AI already changing the workplace? This was the conversation at BoxWorks 2017, as moderator Alan Lepofsky, a research analyst at Constellation Research Inc., entered into a dialogue with two leaders doing innovative work with artificial intelligence, machine learning and Cloud Content Management: Inhi Cho Suh is General Manager of Watson Work & Collaboration Solutions at IBM and Cloud Architect Peter Duhon, who is helping wrangle the content activity of various brands under the iconic Canadian retailer, Hudson’s Bay Company.

The three spoke about the AI already prevalent in our consumer lives, influencing our purchasing decisions with things like product recommendations and customer-service chatbots. And they weighed in on the near future of AI as we sit poised for its ubiquity.

“AR is already impacting your life. The greatest AI is like special effects in movies. It’s the ones that you don’t notice.”

— Alan Lepofsky, VP and Principal Analyst, Constellation Research

Why now for AI?

AI is a nascent business trend, and those already embarked upon the journey are early adopters. But the conversation about AI started way back in the ’60s and ’70s. So why is it suddenly gaining traction — and becoming a reality in the business world — now?

There are three colluding factors that are driving the current momentum behind AI:

• We have more available data than ever before. “The new data generation is impressive,” he says, “and that itself lends itself to new ways of solving problems.”
• We have vastly more availability of computing power and storage. People have been envisioning AI enhancement for decades, but now we have the ability to test it on live data at scale — data stored at a fraction of the cost, thanks to cloud storage.
• Consumer expectations have changed. We expect our digital and personal interactions to be seamless, yet personalized. Thanks to trailblazers like Amazon, when we visit a website, we now expert to be greeted personally and given prescient recommendations. As Duhon puts it, “Apple, Netflix and Google have gotten us conditioned to expect a seamless experience from beginning to end. Consumers expect brands to know who they are on contact.

These three factors are responsible for the current evolution in AI and machine learning. But we’re just getting started. The AI journey will continue to accelerate under pressure to innovate faster, thanks to the pace set by AI trailblazers in the consumer space.

“It’s these little incremental things that we’re all burdened with that we can free up our minds for more creative thoughts.”

— Inhi Cho Suh, General Manager of Watson Work & Collaboration Solutions, IBM

Valuable enterprise data, accessed securely in the cloud

At the very heart of AI is enterprise data. It’s your most valuable asset, one you need to both exploit and protect.

If your most differentiated data is your enterprise data, how do you use it to train machines without exposing it back to master data systems that can’t separate your data from that of your competitors? How to contribute to machine-learning efforts without exposing your sensitive data? Because if that happens, Suh says, “you essentially reduce your competitive edge faster than you built your enterprise.”

In her work at Watson Work & Collaboration Solutions, Suh and her team address the importance of supervising the data sets in an enterprise in such a way to give tight control parameters, without letting results train back to a master set. These are features the team is very careful to build into all AI efforts.

One of the interesting things that’s happening with AI early adopters is a vendor-agnosticism around which machine learning tools they’re adopting. They’re not beholden to a specific platform, and instead they’re selecting from best-in-breed services like IBM, Google and Microsoft, picking and choosing which tools to use for each project. By storing data in secure Cloud Content Management repositories such as Box, early AI adopters are able to apply various tools to meet different objectives.

The practical applications of AI in the enterprise

In a retail industry where product images are key to sales, Hudson’s Bay Company employees are used to spending hours a day manually downloading hundreds of images in order to manually rename them and then manually route them to their next destination. It’s a tedious process with enormous room for error. For years this has meant missing images, empty web pages, and lost opportunities to make sales.

But it doesn’t have to be this way. Using Box Skills and applied metadata, Peter Duhon’s team is creating a better way to route images in and across the company’s several brand banners, which include Lord & Taylor, Gilt, Saks, Saks Off 5th and The Bay. Hudson’s Bay Company has been able to reliably move image handling out of human hands and into the domain of machine learning. Now, thanks to AI’s ability to tag images automatically, at the press of a button, image files can be pulled from a photograph storage folder, renamed by the machine, and then pushed to their destination — all without human input. Duhon reports, “When we upload images, within seconds, we’re seeing metadata applied, so it’s frictionless.”

This shift in process is not just making image tagging quicker and more consistent. The business ramifications go deeper than that. With all photos and other content unified in the cloud, the disparate banners that make up Hudson’s Bay Company can now share assets quickly and securely, not just among themselves but with outside partners such as vendors and printers who need access to files in order to get marketing campaigns out the door. As Duhon explains, “Process automation will help HBC by getting product to the website and customer at a faster clip.”

“Process automation will help HBC by getting product to the website and customer at a faster clip.”

— Peter Duhon, Cloud Architect, Hudson’s Bay Company

Companies like Hudson’s Bay Company are developing applications for machine learning that directly enhance the lives of customers and employees. Today, AI is driven by practical business needs.

For those just getting started on the AI journey

So where to start? The biggest motivation often comes from the most challenging pain point. What’s the one thing you really feel pressed to solve?

The more specific you are about your objectives, the easier it is to train data around them. You can start simple, with existing open-source libraries, or a combination of the work that’s already been done in knowledge libraries of pre-classified images. And slowly, from there, move on to more advanced or proprietary solutions.

Duhon, who has a lot of experience introducing AI to a traditional retail brand, emphasizes: “The first thing to avoid is the fear factor.” He recommends bringing an AI expert in to talk to your team and identify quick wins. You have to have an open mindset, and a willingness to work with different vendors on specific tasks.

Suh backs this up: “You have to have a mindset that’s open to learning and change. We operate Watson in roughly 45 different countries around the world today, in 25 different industries — healthcare, financial services, law, medicine, retail. What’s been exciting is the ecosystem of the use cases and the partners that have been involved with it. It starts with the simple things.”

Above all else, know that it’s never too late to start on the AI journey. In fact, we are still just at the early stages.

The blueprint for starting your AI journey

• Start with the most critical pain point today. Identify it, be really thoughtful, and gather as many details as possible. Make some traction, and in due course, you’ll advance. Don’t wait for the most perfect skillsets and resources of funding.
• Do your best to form a digital SWAT team. Duhon describes this: “Find allies to partner with, and brainstorm different ideas.”
• Pick a use case in an area where you’re inundated with data. For instance, says, Suh, choose a challenge like “fraud-detection anomalies or machine-generated IoT data. Then think thoughtfully about your training around data ownership, because you’re going to be accountable later. The regulations around this are still very early.”