Creating our Digital World with Artificial Intelligence
A Q&A with Susan Etlinger on the Future of AI in the enterprise.
Susan Etlinger moderated an Adobe Think Tank titled the Future of AI in the Enterprise, held April 30 in New York City. As an industry analyst with Altimeter, a division of Prophet, she’s an expert in digital strategy and conducts research on artificial intelligence (AI), big data, analytics, and digital ethics. While in New York, Susan shared her thoughts with us on the impact of AI on the workplace, its implications on the customer journey, challenges, and opportunities with AI that we should look to address.
Susan, what was the most interesting part of the Think Tank discussion for you?
One of the things I thought was really interesting was the continued challenge of defining artificial intelligence. In the research I’ve done, I’ve asked many different people — computer scientists, programmers, statisticians and data scientists, business people, and others — and I keep getting different views. One way to define it is in terms of the actual technology that constitutes AI from a computer science perspective, and there is a lot of disagreement there. But when we think about AI in a business context, it really represents the ability for machines to sense, classify, analyze, draw inferences, act, and, most importantly, learn from past data and experience. That presents an opportunity for organizations to learn at scale in a way that wasn’t possible before.
How have you already witnessed AI transforming the workplace?
Today we are using data to see patterns that we couldn’t see before and learn from those patterns — that can take shape in many ways. One example is the customer journey and understanding what kinds of signals might indicate that a customer is ready to buy, or a customer is ready to churn, or a customer is having an issue that could be resolved.
Another area is conversational interfaces and the idea now that we’re moving from a browser-based world to a world in which we can talk or type to devices, and they’ll talk or type right back to us. That creates a really interesting set of opportunities for enterprises around fulfilling customer requests, and that’s where a lot of early work is being done around AI. In the case of chatbots, we need AI not just to understand explicit intent — say to transfer money, but also to understand implicit intent — not to put myself in a bad financial situation.
Then, from there, we start to understand things like state of mind. Am I frustrated? Am I worried? Did I lose my credit card? Am I traveling? And then we need to create a different kind of process that is empathetic and relevant to the customer’s situation and state of mind.
Those are great customer-facing examples — experiences the enterprise delivers with the help of AI. What are some more examples that are internal to the enterprise?
One of the things I heard somebody say once, and I wish I could attribute this properly, was that a lot of the most innovative technologies don’t really look like much at the very beginning. Think back to (or search) web sites from the mid-90s – they’re awful. So at the beginning, tech often doesn’t look like a lot, but now we have little devices that run our lives. With AI, we’re kind of on par with 1993 in internet years.
With AI in the enterprise (and with the consumer too), we’re at this really early stage where it doesn’t look like much, but it’s also a bit magical. One enterprise use case for AI, which sounds like the simplest, most boring thing, is managing calendars. However, the actual ability to create an appointment between two people is really a complex, multicycle task.
Another enterprise use case for AI is cybersecurity — helping employees recognize phishing emails. And the next step would be that the technology would be so well integrated into the delivery mechanism that the emails don’t even reach the employee anymore. In financial services, AI can help understand money laundering patterns, fraud patterns, and risk patterns. And in healthcare, one of the more challenging use cases has to do with electronic health records — for regulatory reasons and because the data’s very messy. Anything that has to do with paperwork and processes, and is very data rich, and structured, and has a relatively defined set of outputs is a great use case for AI.
What’s the most exciting and the most frightening part of AI?
AI absorbs and reflects and amplifies all the assumptions and bias that we have in society. It has the good, the bad, and the ugly. So when you’re using data that comes from people to train systems, guess what? All that stuff gets absorbed. There are some great papers out there on language bias and image bias, and luckily there are people in academia and industry who are working on how to remediate that. We like to think of algorithms as neutral – but they’re not neutral. They reflect who we are, which is also an opportunity, because it gives us a chance to make it better. That’s the exciting part.
What do you think AI is really good at, and what are some areas that maybe we expect it to be better, but it’s not?
A developmental psychologist named Howard Gardner in 1983 came up with the theory of multiple intelligences. This idea that human beings have nine specific types of intelligence. One of them is logical-mathematical intelligence, and that’s the ability to see numbers, understand them, and process them. Machines are just insanely good at that. But they’re terrible at some of the other kinds of intelligences humans have — for example, intrapersonal intelligence.
I think the Gardner model makes a really interesting frame to have a conversation about what machines do well, what do humans do well, and how we could create a complementary structure, where machines help people be better.
What do you want people to remember about AI technology?
We’re just at the very beginning, and every technology shift brings about behavioral shifts, so it’s hard to have perspective at this point. From a pragmatic point of view, however, we should start by focusing on the processes and the areas where there is a lot of data and a big problem to solve. From a more human point of view, we need to recognize that this is exciting technology, it’s powerful technology, and it needs to be treated with respect. Finally, AI gives us an opportunity to consider, and make real decisions about, the kind of digital world we want to live in.