Watson on the runway – Q&A with IBM’s Keith Mercier on Watson’s role in transforming retail

IBM Watson loves to ingest data (that’s the official verb). Most compellingly, Watson’s favourite meal is unstructured data. After helping out in the healthcare and life sciences sectors, the ‘cognitive computing system’ is turning its talents towards helping the retail sector and its customer experiences. Marketing chats to the man pitching Watson’s talents to retailers, IBM’s Keith Mercier.

This week IBM’s global VP for Watson Transformation in Retail, Keith Mercier, visited Melbourne Spring Fashion Week’s (MSFW) industry event to show a room full of retailers and fashion designers one of the many possible uses of the company’s artificial intelligence/cognitive computing system, Watson.

Watson took on the form of ‘My Stylist’ in a smartphone app, recommending an outfit for its user to wear. In coming up with the advice, Watson asked the user questions including, “What sort of mood are you in?” and pulled together unstructured data from weather reports, scheduled calendar events, and the latest looks from Pinterest. Mercier told the audience that apps such as this were changing consumer’s expectations of brands – and such expectations were often coming from other industries.

“My son orders from a pizza place in San Francisco,” Mercier told the audience. “He’s 14 and he and his friends always order from this one pizza place because he gets a text at every stage of the pizza preparation process. The first one when it’s being made, when it’s put in the oven, when it’s being boxed, and when it’s in the car. He’s 14 and he can track his pizza from order to delivery.

“How do you think he’s going to be walking into a retail store that can’t find an out of stock item?”

He also discussed the results of IBM’s recent global consumer study, which we discussed last month in this article.

Marketing sat down to chat with Mercier after the MSFW event.

M: That was a great little app you showed up there. Can you give me some examples of times where you’ve used something similar to that?

KM: Well we’re very new, so we just launched Watson in retail in February of this year. We’re a very new business so we’re really in the process of getting the word out and helping our retail partners understand what Watson is and that cognitive computing is and how that’s very different to the programmable world of computing we’re in today where we tell a computer to do something and it does it.

Watson is more of a learning type of computer – it’s not actually a computer, it’s a capability where it reads and understands information – and then you can ask questions in natural language back to Watson and Watson will go find the answer for you and say, “This is what I believe is the answer, and this is how confident I am that this is the answer, and if you would like to test my confidence, let me show you where I found the pieces of data that make me confident that this is the top answer”. So it’s very open that way, it’s not a black box, it’s not something that you just put in a question and it magically comes out with an answer.

I really think of Watson more as an advisor versus giving you an answer, because at the end of the day we still all make decisions on our own, but having more information quickly can make us allow us make a much more informed decision.


M: So it’s sort of a more intelligent version of a search engine in a way?

KM: Well, the thing about search is search doesn’t understand your intent. Search looks at keywords. Watson actually passes out a question, and says, “This person is asking a question about a place, potentially a time, with these certain constraints”. A search engine would just look at a couple of keywords, and it will bring up 50,000 results, and somewhere in there you have to find the actual answer.

Watson does that additional work for you, and says, “I believe after reading the 50,000 potential results, I’ve scored them, I’ve ranked them, I’ve actually hypothesised against them, and these are the top three”, and it does in about the same amount of time as search.

The other part you have to know is Watson only does that on data that it learns. So it doesn’t just know anything, you have to actually train Watson about we call it a ‘corpus of data’, a body of data – you say, “Watson, learn this”.


M: So if a retailer wanted to invest in, say, the My Stylist app that you had as an example, how do they go about it? How do you teach Watson what you want it to know?

KM: Well, the first thing that we do is we typically spend time with the retailer helping them understand what it would take to implement a cognitive engine like that. Where are all the places that you would want to find data? Purchase history, weather, in that example. Do you want to put in mood? What happens when they click ‘flirty’ versus ‘sexy’? What type of questions would you want it to ask next to help guide the dialogue? So there’s a process where we work with retailers for a few weeks just building and listing what we call the ‘use case’.

Then we actually go into a process called ‘configure and train’ where Watson starts to ‘ingest’, is the term, it sounds so weird, but thats the term, ‘ingest’ the data. Spreadsheets, PDF files, HTML, it’s all unstructured content, and that’s what Watson loves is unstructured data. So when you think about social today and how important social mobility is becoming and the amount of data that consumers are creating about their retailers, Watson is a way for retailers to quickly understand what people are saying in that data versus hiring big teams of people to constantly go through and review what people have said.


M: In that way its a bit of an investment in social listening?

That’s one way you could deploy it. I showed a way that you can have Watson help you figure out what to wear in the morning. There are ways that you can deploy Watson inside a company where if I’m a sales associate and I’m trying to help a customer on the sales floor who’s interested in this denim, I can quickly ask Watson, “What other colours do these jeans come in?”. Because I might be in a small store in a big city and I’ve got red, white and blue but it also comes in green, brown and these others, but I don’t even know that, so Watson can help me out.


M: So is this a way for retailers to almost control the message? People are walking into shops with their smartphones and maybe they’re searching a competitor. So is this a helpful way for them to keep it in the company?

KM: What I would love to see is retailers using Watson as the engagement part of their app. So retailers today are spending a lot of money and time developing apps, which really are about a store locator, some of them link to shopping. But consumers are really having a tough time engaging with them because there’s not a lot of utility there.

So if Watson was the place I could go open my app and start engaging with some questions, and go, “I’m looking for a little black dress under $20, what do you have in size small?”, versus going to the e-commerce site, going to black dresses, clicking size small, and then looking for all the prices. It can really change the way a consumer navigates a brand, on an omni-commerce site, on a device, or in the hands of a sales associate, helping a sales associate be more knowledgeable.

If you think about the Q and A that a sales associate has with a customer today, none of that is recorded. So once the consumer leaves that data is gone. But if that sales associate is now recording, asking questions based on what you’re interested in, that’s a whole bunch of new information I just learned about you as a customer, and if I have a customer profile of you and I can now add this to you and I can actually send you more relevant communications from my brand because now I know you’re interested in little black dresses, I didn’t know that before. Those type of opportunities, to capture that data and keep it and make it useful for the consumer for the future, I think is really where Watson is going to make a change.


M: I put together a story the other day from some data that said retailers are quite behind compared with other industries in adopting technology. The stats we had were 21% of retailers in Australia are developing apps and only 30% have websites. Have you found that retailers are struggling to work out how to adopt technology? 

KM: That seems low. Globally, yes, no doubt. Like I said when I first got up, the change of technology adoption in the last five years by consumers is faster than IT executives can get budgets approved. It really is amazing how consumers have just fallen in love with the smartphone, and other industries have used technology and other online brands – there are other industries that are training [consumers]. My pizza example – a small local pizza house that decided, “You know what, let’s just do this”. I don’t know what system they’re using, but its changing the expectations of my son and the way he thinks about shopping now.

It’s changing so quick. 10 years ago, it was “We’ve got to get into the e-commerce business”, and they have, and now you have this mobile thing, and now they’re all kind of coming together. I think the real challenge is integration, because the consumer says, “You’re one brand to me. When I pick up my phone, when I get on my tablet, when I’m sitting at home on my computer, it all needs to feel like the shopping experience is continuing”. That’s where they’re really struggling today, but they are focusing.

ALSO READ: Retailers slow to improve technology for mobile shopping – study »

Really where I see a lot of them going is inventory, because inventory is the crux of everything. Consumers want to know where it is and retailers want to know where it is because thats the biggest cost, so it’s a really great place to start. Then the data and understanding who my customer is as a single view – those two are the really foundational areas that we see retailers focusing on right now, as kind of the building blocks to get to the true omni-channel experience.


M: One thing we thought was really interesting was about Watson itself, from a marketing point of view, is that it’s this abstract computer system, which is difficult to understand, but it’s got a name – we thought that was a really great way of marketing.

KM: Is it a man? Is it a woman? Watson has a man’s voice, Watson has a female voice. Well it’s named after Thomas J Watson, our founder at IBM. Really, where Watson made its inroads was, our technology team does these challenges, and when they started developing the artificial intelligence really like 10 years ago, someone said, “What if you could get it to play a game show and win?” You’ve heard of the Jeopardy game show, and once Watson won, then they realised, “I think we might have something here. Now what do we do?” It was originally a challenge and then they said, “Well wait a minute, where can we deploy this in industry?”

The first industry was healthcare and life sciences, so you actually have cancer doctors that are using Watson to plug in proposed treatment plans for patients, and Watson is going out and reading a ton of data and bringing back reasons why it would work, why it wouldn’t work. It’s helping those doctors make better decisions and faster than they ever could today. It’s been successful there for the last kind of two to three years, and then in financial services and then in February it was, “Now let’s go out in industry”. So we are very young and new, we’re like a start-up to some degree, which is very fun, and at the same time we certainly see lots of opportunity for it.


M: So, out of all the industries you could have chosen, why retail this early on?

KM: Well it’s not the only one. We’re also in travel, transportation, we’re in legal, some others as well, but IBM has a deep history in retail, so why not? It made perfect sense. We have such great relationships with such leading retail brands that are now looking to us for transformational road maps because technology’s now playing such a big role in the consumer experience – it just made perfect sense to say retail.

It’s one of the things we’re now bringing to industry, and we’re the only one that has it right now. We’ve developed this, we’re very proud of it, and we want to partner with them to help kind of co-build this together, and I think that’s what’s wonderful about doing something so new, is that you don’t have to follow the old rules. So we’re working with retailers, saying, “You help us. You tell us what you would want to do with this”.


M: How does it work between big business versus small business? I’m imagining it’s going to be quite an expensive technology to bring on. What about for smaller brands?

KM: Well it’s new. So all new technology is very expensive at first. But it would be great if it could it get into small business, certainly. And one of the things we’ve done is when we launched Watson, we opened up the API, and we’re developing a whole ecosystem of third party partners, that are using Watson and coming up with use cases, coming up with their own solutions. Those companies are in a position to potentially work with smaller organisations.

So the technology will be available. Again, that’s very new, we have two or three signed on now, that are really doing interesting things. They’re doing things in a very different way, and their pricing can be different, and they may be more accessible to smaller companies. My hope is that over time, if you’re going to usher in a new era of computing, everybody needs access to it. Now its very nascent now and expensive, I wouldn’t say expensive, but for a very small company, I wouldn’t say it’s the first place that they should be spending money, right? But over time I hope would be that it’s available at every price point and every level.


M: It learns from what data you give to it, so would it ideally get to the point that you don’t need people that know about software development? Is it that smart?

KM: No it’s not that, it only knows what you train it to know. It can’t execute tasks yet. It can’t write code. Right now it’s helping be an advisor. Over time we’re developing, we’ve just launched a debater. So now we have a version of Watson that can help look at a problem where there’s no right or wrong answer – abortion. So Watson can go out and come up with reasons why thats okay and reasons why not, but reveal the data. So we’re expanding Watson’s capabilities beyond Q and A now, debating, but over time there’s a very long roadmap. We want to co-develop that with industry, so we make sure what we focus on is going to be valuable to our industry partners. It’s still early but lots of potential.


Michelle Herbison
BY Michelle Herbison ON 5 September 2014
Assistant editor, Marketing Magazine.