Applying Artificial Intelligence and Machine Learning more widely in business is not a matter of if, but when. There is more and more data that can benefit from advanced processing, as well as an increasing complexity in the problems we need to solve that call for algorithm-augmented intelligence.
Gauthier Vasseur—Executive Director at the Fisher Center for Business Analytics, Berkeley Haas School of Business—moderated this panel discussion with practitioners to learn more about how AI and Machine Learning are shaping the present and future of work.
Meet the Panelists
Sarah Aerni, Senior Director Machine Learning + Engineering at Salesforce
Dan Feld, Head of Global Enterprise Business, Hardware Partnerships at Google
Robert Brown, Vice President, Center for the Future of Work at Cognizant
Introduction From Moderator Gauthier Vasseur
“It's high time we stop and think about how we can make AI and Machine Learning work for us. How can we apply it to do better business or create a better world?
As AI becomes smarter, stronger and faster, what is the value that we can all bring to the table?
“This expert seminar series empowers you to understand what AI and ML is about and how it works in the real world. Where do we fit into this as human beings? As AI becomes smarter, stronger and faster, what is the value that we can all bring to the table? We'll finish by taking a look at what the future holds because a lot is going to evolve and we need to find our bearing.”
What is your definition of AI or Machine Learning?
Robert Brown: There are two ways to think about AI. The first is a lot of people talk about ANI—narrow intelligence. That's very specific applications of AI to business processes, to customer experience and so forth. That's the preponderance of what we're seeing today.
The second is general intelligence—some of the new cutting-edge applications of AI engines like GPT-3.
And then there's superintelligence, which is AI “Terminators” running amok in our popular imagination. But I think the action is with Machine Learning.
Sarah Aerni: I run a few of the automated Machine Learning teams at Salesforce that deliver AI into the hands of our customers so that they can leverage it in the way that best helps them do their jobs in the context of our customer relationship–management platform. When I think of AI, it's really about having machines think like humans, teach them how to make things easier for us. We can then have the human level-up what he or she should be doing that's new and different.
In the area of Machine Learning, we also think of deep learning where it's the application of Machine Learning to structured data, image data, text and voice. In every context where we have access to data, how can we augment that experience and allow the human—who's doing his or her job day to day—be superhuman in that experience?
Dan Feld: At Google, I focus on our mobility platform—both with the Pixel phones and the Android Enterprise platform—and how large enterprises are utilizing it. AI for me is a business opportunity to scale. Everything that we've done with databases and with data-driven systems in businesses were fragments of systems, fragments of values.
You needed so much understanding of the processes and of the technology to make sense of it. Even if we're successful in doing so, the precious time in focusing on our customers is being moved into focusing on the fragments of technology. AI gives us the opportunity to scale all of it and to connect the dots and have a much more coherent process that can adhere to the business process.
That's what I think is exciting today: the availability of solutions and the ability of business leaders to be part of making these decisions.
Are there any concrete projects or applications you see today where you see value?
Dan: I think that vertical-specific solutions that can help us solve problems or get insights into processes are the name of the game today. If I would like to use Machine Learning to identify if there is a certain object in a picture, I can build complex Machine-Learning systems or even deep learning. I can train algorithms. I can do a lot of things and get the perfect result for what I need. Or I can find out that today in the Cloud business, there are specific high-level services that can give me either exactly what I want or 80 percent of what I need.
It's just one example of the availability of specific solutions and the need for business leaders to understand the meaning of the problem, the various types of solutions, and to make sure that we, as a company, are focused on using the right solutions with the right priorities. That's what I think is exciting today: the availability of solutions and the ability of business leaders to be part of making these decisions.
Leaders need to get their hands into a bit of technology today. It's not a matter of whether you've got to do it or not. It's a matter of when you have to do it, and when is probably now.
Robert: One of the most fascinating examples that I've seen in the past several years is the use of AI facial-recognition software, which we think societally, at least in the West, is a little bit spooky. But The Nature Conservancy actually transposed facial-recognition software and called it FishFace. It's used on commercial fishing fleets to look at how many endangered species did they actually pull up in the fishing net? FishFace allows them to do that.
And then for anybody who has followed along the development of AlphaGo, the transposition of that technology to AlphaFold helps scientists get to massive-scale protein folding as an on-ramp to solving pre-onset dementia and Alzheimer's—that is an absolutely super-cool news story.
Sarah: At Salesforce, we are putting AI in the hands of our customers to use it in the context that makes sense for them. And I think what's fascinating about that is it's beyond what I could really imagine. For example, Einstein Recommendation Builder, which is the ability to recommend things to things. How do you actually allow a customer to recommend products to a sale? We have other examples where it's used around optimizing a service interaction.
For me, it’s impossible to think about how we could address this at scale by hiring data scientists in every organization, building up the infrastructure required, or even finding the right tools that exist out there and leveraging them. But our customers are showing us that if they're given the right tools, they are actually able to apply it in the right domain.
Audience question: Is AI overhyped or underhyped, or just right at the level it should be today?
Sarah: When you look across the industry, AI is maybe under-leveraged: We don't have enough access to the tools needed in every context that's required. I'm always hopeful that we will start shifting the conversation toward the value. How much value have we gotten out of AI in the particular context that we need it?
Robert: Our team just completed a data set of 4,000 interviews globally with people from the C-Suite. And it was quite interesting that 70 percent had some level of AI implementation in their business processes. The minority had deep full production, some were further along in their journey and some were just beginning. We asked them, what are the top benefits that you're getting out of these initiatives for Artificial Intelligence? And the top two for 2020 were operational efficiency and decision-making.
We asked them, what benefits will be for 2023? They maintain that it will still be operational efficiency and decision-making, but almost a 10-percentage point more intensity in terms of the value. How do we use AI as our co-pilot? As humans, how do we get to good decision-making?
I think everyone can benefit from looking at how AI is applied and thinking about how it could be used to drive impact—start from there.
Audience question: How do you make good usage of Machine Learning and AI if you don't have access to high-end computing? Is that still a barrier to begin working with AI?
Dan: There are tools that are available to all of us that can be used with basic concepts of data—of understanding not only what answers we want from the data, but also what questions to ask. These are the insights that we need to ask ourselves. How do we do this? Which tools are doing this?
Robert: I think you hit the nail on the head, Dan. I would argue it's access to data. Is the data clean, believable, real time? What are the sources where that data is coming from? So when we think about skills for the future of work, it's electrical engineering, computer science and data science. And there's precious few data scientists today. And that also asks, how are you building the modern data stack for tomorrow?
So I think there's a certain generation of practitioners who are thinking about next-generation data lakes. But again, if the data is not clean, believable, real time or integrated, the data lake is like a data cesspool. You have to refine that data. And then people who are maybe of an older school are thinking about relational database management systems.
Sarah: There's been an evolution where we focus heavily on the technology and what needs to be in place. And as a data scientist, I’m thinking about what's required to move it forward.
We need to have these pipelines in place. We need to have applications. We need to be thinking about why we're doing this. How do we measure the impact that it's having on the business? We're trying to improve efficiency. How do we measure efficiency and prove that implementing a model and putting it into production is appropriate?
I think everyone can benefit from looking at how AI is applied and thinking about how it could be used to drive impact—start from there. How will I measure the impact? How will I then implement it in a way that you actually measure that change? That can prepare you for when the technology is available to you.
Audience question: How do you start today to do some AI and ML and at least apply that to your business?
Sarah: It depends on where you're tackling it from. In my session, I talk a lot about what I believe are the foundations of putting AI into production. I talk about the mindset that I think we all came from: Centralized data, then bringing in a person who understands data and then magically we will have AI in production.
But that’s disconnected from the reality of what it takes in terms of a platform that allows experimentation, a way to iterate, a way to put it into production. Having a model on your laptop is far removed from having it in production where it's consumed and driving an outcome.
There's more around what I think some of the foundations are. Definitely bringing everyone together to the same table and thinking through it.
Robert: In my seminar, my guest is one of our clients from PepsiCo, Martha Roos. And Martha talks about the application of technologies to build the perfect Cheeto. One of the lessons learned that Martha talks about is to be prepared to fail fast. What have we learned, and how can we apply it to the next thing? And then once you get that success, how do you scale it out into production? And whether you're building the perfect Cheeto or optimizing communications or building things like cyber defense for your organization, that's the place to start.
Dan: In my session, one of the things we talk about is, “How do you think about this? How do you apply it in real life? Sometimes you just need a spreadsheet and that's it—we don't need to push it where it doesn't belong. But when you do need AI, do not be ashamed to understand it and implement it. Most implementations today are narrow AI: You identify where you really need it and apply a very focused solution. You learn from there. You see what the result is in production, how it benefits you and scale from there.
When you put the art and the science of the job together—people and machines—really amazing outcomes can occur.
How do you see AI and Machine Learning shaping up in tomorrow's world?
Robert: We need to put people back into the equation. We are really good at the art of the job: judgment, ethics, being able to answer, “What's the right thing to do in this context?” Machines are really good at the science of the job.
When you put the art and the science of the job together—people and machines—really amazing outcomes can occur. In our work at the Center, we've undertaken a number of projects looking at what we call “21 Jobs of the Future”: In 2030, what does the job description for a quantum Machine Learning analyst look like? Something very technology-heavy. But further down a pyramid of sorts, there's a lot of focus now on purpose in work.
We just published a piece called "21 Places of the Future." And one of the places we talk about is Silicon Wadi, which is shorthand for Tel Aviv in Israel, because one of the jobs of the future is cybersecurity, which is going to be critical in this world suffused with AI.
We're moving from a services economy to an experience economy. You're seeing the experience economy emerging in places like Wellington, New Zealand; in Dundee, Scotland, that are far outside of the orbit of the typical mindset that future work happens here in Silicon Valley.
Sarah: At Salesforce, it's the number of ways that we see AI being leveraged. As we move forward, it will inevitably require us to work with the data that we have and think through that business experience.
It's about putting those tools into the hands of the business experts so that they can move us forward in that experience with their customers, with their employees and their workplaces. It is permeating throughout our lives and in ways where we don't even think twice about it.
Are you ready to learn from these industry experts in how you can sustainably, ethically and impactfully bring AI and Machine Learning into your work?
Then start your journey today with our online three-course series!