Workday’s Sayan Chakraborty says that machine learning’s impact will be much like electricity—world-changing—and ultimately taken for granted. He and his team have played a key part in weaving ML into the very fabric of Workday’s underlying platform, which is critical to delivering compelling experiences and outcomes without customers even needing to realize it is there.
Earlier in his career, while at a number of Silicon Valley companies, he played a part in making the technology we rely on everyday—GPS, and wifi, for example—so ubiquitous that most of us take these revolutionary technologies for granted.
Chakraborty also co-founded and served as chief operating officer at GridCraft, a company that developed simple-to-use data analytics tools that Workday acquired in 2015. Now, as senior vice president of tools and technology at Workday, Chakraborty is responsible for the infrastructure on which our applications are built. In particular, he’s leading the charge to make sure that machine learning helps customers make faster, better decisions using all of Workday’s products.
Just to clear up a common point of confusion, what’s the difference between artificial intelligence (AI) and machine learning (ML)?
AI is rooted in the idea that we can make computers that think like humans. It’s an umbrella term, so in general when we use “AI,” we’re referring to possible ways of tackling problems that humans are typically good at solving and computers have traditionally not been. In every approach to AI is the central idea that the algorithms themselves adapt to improve outcomes, a process referred to as “learning,” in parallel with how humans learn.
Machine learning is a subdiscipline of AI that uses massive amounts of data to recognize patterns and to predict or infer insights or answers. So, data is at the heart of machine learning. Our goal at Workday is to use machine learning to help humans do more of what they’re best at, like strategic analysis and making judgments, and free people from tedious work as a result. Working with machines, humans can produce far better results than they can on their own.
You’ve said you don’t believe machine learning should be regarded as a standalone application. Why is that?
Machine learning is, at its best, a different way to conceptualize applications. If you think about Workday as the platform that businesses run on—their financials, human resources, all of their people, their data—then what we see is the opportunity to enhance almost everything and every interaction that our users have.
For example, machine learning powers Workday user interfaces to make Workday better suited to you and how you want to use it; selecting tasks our algorithms are pretty sure you will need right at your fingertips (and choosing differently if you are on your computer or on your phone). This personalization means we can deliver value with our machine learning to every single one of the 40 million users who touch Workday every day, and enhance their ability to get their jobs done.
On the product level,we’re giving our HCM customers the ability to understand the people assets they have, the skills they’ve developed as an organization, and being able to apply the right skills in the right places.
The same thing applies to the financial side—what is the most efficient use of your capital? Machine learning can help you understand both where you’re being inefficient and the opportunities for you to be more efficient, as well as alerting you to see, for example, accounts where you might need to keep a closer eye on. It’s all about insights that lead to better business outcomes.
And, when people get these insights, it’s just part of the overall Workday experience, correct?
Right. That’s an important point, and something we can say about the evolution of technology in general. Whenever you get technology, it often starts off as an app and it ends up as a base capability, right? You used to have dictation systems that you loaded onto your computer. Now, every system has dictation built into it.
That’s how we approach machine learning. The early adopter phase of technology, which a lot of our competitors are still in, is machine learning as an app; as a one-off experience. At Workday we see technology as most beneficial is when it becomes part of the platform; when you’re able to use it without even thinking about the fact that you’re using it.
What else should we be thinking about when it comes to machine learning?
Well, I think it’s important to understand our unique take on it, which is really around partnering with our customers. After all, the data we’re talking about isn’t ours, it’s our customers’ data. And so it’s not about inflicting technology on people, it’s about a partnership where we can help others get to a better outcome, be it for an enterprise or for an individual.
For good reason, you can’t say “data” these days without a lot of people thinking “data privacy,” right?
Yes, strong data privacy protections are critical to us. Our Privacy by Design principles are baked into all of our products and processes. I’d highly recommend reading some of what our Chief Privacy Officer Barbara Cosgrove has written.
And, I agree wholeheartedly with our CEO Aneel Bhusri’s position that every company using AI or ML needs to have the right people, process and technology to oversee ethics. Here at Workday, we have an ethics practice, a team that is pointed inward and outward, to make sure that when we process customer data or build data-based products, we are doing so in a manner that aligns with our Ethics Principles we published earlier this year.
This is important because privacy and ethics are not something that can be addressed by any one person. Ethics, bias, and privacy have to be understood and appreciated by the whole organization. Everyone has to sign up for it. People building and designing products, people supporting customers, our legal, privacy, security, and marketing teams: the entire company.
Our CEO and other Workday leaders have also said machine learning is going to be as disruptive as cloud computing. Why?
Because it’s going to fundamentally transform how people interact with software. As machine learning better understands what we want to do, what our role is, who we are, and how that’s changing, what you’re going to see is things coming to you as you need them, and going away when you don’t need them. I think that’s going to be transformative, and people aren’t even going to quite notice it even as it’s happening.
When people talk about machine learning making everything faster and better, hasn’t technology been having that impact on businesses for hundreds of years? Isn’t this just a continuation?
Well, yes and no. ML does build on existing technology trends—without cloud computing, big data, and the other advances we have seen, AI and ML would have remained an academic discipline. But also no, because there were always classes of problems we couldn’t solve with traditional approaches. Now, we have the capability to solve them. And some of these problems, at face value, may seem trivial. For example, I can now use machine learning to identify things in photos—machine learning can tell me whether a picture on the internet is of a cat or a dog, for instance. That was just not possible before, and that opens up a set of possibilities that is a truly fundamental change around certain problems that were previously considered intractable.
For example, every day we use machine learning so that employees can simply take a picture of the receipt they want to expense and the fields are automatically populated for them. This ability to sift through thousands, millions of images, and extract the needed information without involving a human, speeds up the business and frees up employees to do things that matter more. That opens up huge opportunities, and is just one small example of how machine learning is revolutionary, not evolutionary.