We’re back with Kunal Chopra for Part 2 of our discussion about artificial intelligence. In Part 1, Kunal defined artificial intelligence and gave us a high level understanding of what it. Now we’re going to talk about the application of artificial intelligence to the world of retail. In this episode, Kunal is going to tell us what’s possible now, who’s doing a great job using artificial intelligence, and what to look out for in the future.
Melinda: Hi, I’m Melinda and you’re listening to Think Retail. Today we’re back with Kunal Chopra for Part 2 of our discussion about artificial intelligence. In Part 1, we listened to Kunal define artificial intelligence. So, hopefully, now we have a clearer idea of the line between fact and fiction. And today we’re going to be talking about the application of artificial intelligence to the world of retail. Kunal is going to tell us what’s possible now, who’s doing a great job using artificial intelligence, and what to look out for.
Welcome back. Thank you for joining us again. Last time we left off with you describing what the biggest benefit of AI for retailers is right now. Can you just recap on that?
Kunal: Yeah, absolutely. It was really around the benefit of client segmentation in a much deeper or more intimate understanding of individual client preferences and buying patterns and then being able to deliver individual products, more customization at a client level.
Melinda: And how would that work? I understand theoretically how it works, if you have this information then you can have customization, but how do you get the information in the first place?
Kunal: I mean data is gold at this point, right? So, if you’re not making use of your data, you’re not collecting it from multiple sources, you’re doing yourself a disservice, whether it’s point of sale, whether it’s your receipts from actual interactions, whether it’s looking at patterns and viewing behavior for clients who are logged into your website. There’s a number of ways you can get data. I think you have to be careful with that. There’s a lot of data privacy issues that are at the forefront of a lot of this right now and will continue to be.
Kunal: We’re seeing laws passed such as GDPR in Europethat will find themselves or find their ways to North America as well. We’re a little behind in some of that at the moment. But getting data is part of the problem, and then being able to make sense of it is another part of the problem. I think the biggest challenge though is great. Once you’ve collected data in a transparent and ethical way and you’ve done the analysis, how do you actually deliver a customized product or solution to your clients? Customization is not often easy, especially in a mass retail environment.
So, it’s going to be how do you figure that out?
Melinda: Yes, absolutely. And what would be some other applications for artificial intelligence that retailers would be smart to think about right now?
Kunal: If we look at how it’s typically being used right now there’s a lot of backend, supply chain, inventory management uses of AI that are at the forefront of how retail is using it. It’s hard to sit here and say, “Here’s something you should do.” What I would encourage retailers to start doing is thinking about how AI fits into your broader strategy. It’s a tool like any other automation or non-automation related tool. There are benefits. There are pros and cons. So, really start to think about who you are in the market, what is the service experience that you want to deliver whether it’s a bricks and mortar experience, whether it’s an online experience, and then work to figure out how AI fits into that and enables that.
Melinda: Okay. So, you mentioned supply chain. If you’re working with a brand and talking about things like “just in time inventory,” a lot of these technologies come with a really big upfront investment. How do you approach that? Where would you, if you were a brand owner, where would you start?
Kunal: That’s a great question and one that Munish and I get all the time. Our guidance is: don’t start with a multimillion-dollar investment. You’re doing yourself a disservice. AI is different in terms of what we’ve traditionally seen in terms of technology and automation. Start small. Our guidance is start with something that you can put a box around, test and explore with a small sample internally within your organization first and your employees and then with a group of clients that’s outward-facing as well and really understand what it is and how it fits into your organization, how it fits into your processes, how it affects you culturally as well. There are cultural implications of automation on your staff, there are perceptions that your clients will have about what you’re doing with their data. You need to really assess this before you go down the multimillion-dollar path. Start small, keep it boxed to something which has value in terms of the insight you can derive or the service experience that you can measure, but isn’t mission-critical to your organization and your financial success. And then quickly learn from it and understand, great, does it make sense for us and our clients and our employees? And if so, then start to think about the bigger and longer-term investment.
Melinda: A couple of times you’ve talked about data privacy and yes, GDPR is something that we can probably look at. In North America, we can look at that and say, “Yes. We’re probably going to be dealing with that here as well.” How would we plan for that? If we know that it’s coming, we don’t know when but we can anticipate something down the pipeline. If I’m a retailer and I’m thinking about collecting data, how can I future proof and make sure that if I am investing in this even this small boxed testing of an artificial intelligence program, how can I get ahead of that to be sure that I’m not going to have to switch gears suddenly down the road?
Kunal: I think there’s a couple of ways. One is look at some of the more stringent regulation that does exist in the world today and try to adhere to that. But the other one which I think is just absolute common sense is, be fully transparent with your clients around how their data is being collected and how it’s being used in terms of the types of analysis, who it’s being shared with, whether there’s third parties involved or whatnot, and get their consent. And that’s just full transparency, which will always keep you out of trouble.
Melinda: And we’ve seen there’s been a few people who’ve gotten into trouble not being fully transparent.
Melinda: So, what are some of the big barriers that brands face when they’re starting to think about AI and they’re maybe trying out this little testing environment and wanting to implement something? What are some of the barriers that they’re finding?
Kunal: I think there’s a few. One is really the operational perspective. And you touched on this with the multimillion-dollar investment but it’s really around the investment that’s involved. Having touched on that, start small, keep it basic, and then understand whether it makes sense.
Melinda: Is that because there’s a value perception that we’re not seeing what is the return on this?
Melinda: Is that where that’s from?
Kunal: Yeah. It’s partly that and partly from a lack of understanding of what it is, how do we derive value? Does it make sense for us as an organization to go down this path?
Melinda: You also mentioned that it’s going to change your internal company culture. Can you describe that a little more?
Kunal: Sure. There’s going to be resistance. There are jobs which are going to be affected in terms of you may lose some analyst jobs to programming jobs. There’s a cultural aspect of this which has to be considered. AI and the systems involved and the analysis that it produces is not 100% accurate. So, if you were relying on systems that were more accurate than a typical AI system, and not to say that they’re entirely inaccurate, but how are you accounting for those inaccuracies in your processes, in your employees and their knowledge? How are you making sure that those are detrimental to your long-term business model? So, there’s some things that need to be considered as people test and learn from this as well in highly regulated industries like banking. For example, having inaccuracies in things like credit models can be detrimental.
Kunal: And then how do you prove to your regulator that you’ve accounted for them with process controls and whatnot to make sure that you can mitigate any of the issues that come out of those inaccuracies. So, there’s a lot of learning that needs to happen with this.
Melinda: And are there any others, so aside from not understanding AI and not understanding maybe the value that it might bring to an organization, are there any other logistical barriers?
Kunal: Yeah. So, getting data, we touched on this, getting data. It’s not only the collection but making sure that the data is usable.
Organizations have mounds and mounds of data but to have systems which deliver the right data in the right format, in a clean, structured way that an AI system can readily make use of it, is not an easy thing to do.
Melinda: Why is it so difficult?
Kunal: I don’t think organizations have been built historically and data systems built historically for this type of work. So, whereas organizations have gone and collected data for 5, 10, 15, 20, 50, however many years they’ve been in business, it wasn’t for the intent of “In 2019, we’re going to build an AI system and we need it structured this way.” So, it’s almost like we’re now reverse engineering and I have to take all this historical data and cleanse it and structure it and find a way to deliver it to a system in a manner where that system can quickly analyze it and do something with it to produce intelligence.
Melinda: I know you and I have talked about this other times is that, for example, the banking industry where there’s lots of different consolidation, mergers and acquisitions, when you’re bringing different organizations together, can you talk a little bit about the challenges with data in that situation?
Kunal: Yeah, absolutely. Banks are notorious for these multiple systems throughout processes and not just within–across the organization, but within a process as well. You often have a sales system, a credit system, a back-office system or multiple systems at each point in those process. Do those systems collect data in a similar way? Do they or can they share data? It can be as simple as the wrong acronym or an inconsistent use of acronyms across systems, so, as opposed to spelling out road, do you put RD? Right now, you have an inconsistency in your data and how it’s collected. So, it’s even down to that level where you need to be very consistent and very clean. And there’s a lot of effort involved. Other challenges and barriers with respect to the AI front– aside from the data and aside from the investment–is the human interaction issue.
Kunal: And this is one that organizations I think need to be very cognizant of. Do your clients really want to be entirely engaged by a machine? Are they comfortable with that interaction beyond getting consent for how their data is used and the analysis that it’s doing, and the recommendations that come out of it? I ran focus groups years ago for a large Canadian bank and we were interviewing their clients on their mortgage process. We ran focus groups across Millennials, first-time home buyers, new immigrants, people who were renewing their mortgage, and even the most technologically advanced Millennial would tell us openly, “I still want human interaction in my process. I want to know that there is somebody intelligent who knows what they’re doing, who knows how to navigate this, who can guide me. I want to talk with them at least once. After that, I may never talk to them again. I want a fully digital experience.” But what is that balance between technology and human that your clients want and that your business model needs for you to be successful? Finding that balance is not going to be easy.
Melinda: Are there any examples you can think of where people maybe have gone too far? I’m thinking, myself, about the self-checkout kiosks which are very controversial, it seems like.
Kunal: I think that’s an example of broader automation that me personally, I won’t use them. A, I don’t think it’s as seamless as they should be but, B, I actually look at the jobs lost and I’m like, “Really? You’re replacing cashiers with this and now I’m not getting that human interaction?” So, recently, it was McDonald’s. If you go into a McDonald’s in the last couple of years, in Canada at least, I’ve seen the self-order kiosks that they have introduced at McDonald’s. So, I intentionally will avoid them.
And just go to the cashier or the order taker simply because I just think it’s ridiculous.
Melinda: I also find I had an experience where I used a self-checkout kiosk and then as I left the store, one of my items had a security tag on it and it set off the alarm. And I then had to go back and get in the lineup to go through the teller and have them remove it and she says, “Oh, yeah. The self-checkout kiosks don’t remove the security tags.” And I was like, “Well, why didn’t I know about that before because I would have just gone in the line right from the start and now I’ve doubled my time?” So, I agree. It’s not as seamless as it’s supposed to be.
And the other thing that you mentioned is you said, “What is the human interaction balance that the clients are going to want?” Are you finding that sometimes the excitement about this kind of thing, does the client focus or the customer focus get sometimes left behind or people forget about it?
Kunal: I think there’s cases where it does. Again, this stuff is so unique to each brand and how they engage their clients. Can I see the risk of that being left behind or getting lost? For sure.
Melinda: We talked about a bunch of barriers. How can they overcome some of these barriers? I don’t know if you want to pick one or two that you think are most for all that.
Kunal: From the data side, be transparent. And we talked about this earlier. Just be as transparent as you can with your clients around how their data is being collected and used and avoid any of the pitfalls we’ve seen or heard of organizations in recent past run into with that. Data issues can be entirely detrimental to your business, so be careful and be transparent. The next I would say is really, with respect to finding that balance between automation and the human interaction, engage your clients and engage your employees in the design process as you re-envision your business, as you re-envision your client experience. There is no better way to understand how far you should go or need to go than by engaging your consumers and your employees in that process. I know it sounds like common sense, but we live in a world where time is money and there’s a lot of pressure on organizations to move forward at a very rapid pace. And things like this often get lost but you don’t want to simply go to market without having done this and then find that you’re losing business as opposed to gaining it because your clients are entirely alienated or disengaged from the experience that they’re now involved in.
Melinda: Can you tell me if there are any brands out there that you can think of as examples that are using artificial intelligence well?
Kunal: One of the interesting ones is Sephora, the cosmetics retailer. They’re doing something and they have been for the last few years now where they have an application where a consumer can go onto their website or into their store, I believe, and take a picture or upload a picture of themselves. And what this thing does is it then analyzes their individual skin tone and recommends shades of their cosmetics which would be appropriate for that skin tone. And I believe it actually gives the consumer an image of them and how they would look having used this. So, this is a tremendous example of customization and a very unique individualized service experience for something very sensitive to people in terms of how they look and how they present themselves.
So, I think that’s a great example of it. Obviously, we all know how automated companies like Amazonare and how that experience works and how they’re using data to suggest what else you should be buying and looking at and we look at how they run their warehouses and there’s YouTube videosall over this. Well, robots running around these warehouses. But the Amazon is Amazon and that’s an entirely online automated model. But there are companies who are using it at a consumer model and Sephora definitely stands out.
Melinda: Sephora, I was wondering if you might mention that brand because they have a lot going for them. They have a reward system that’s very, very much used. Lots of people have rewards programs but theirs is very popular and their subscription box is also really popular. They’re collecting a lot of information about you and able to personalize stuff. I think that that’s on the top of everybody’s list.
Are there any trends or directions that you think won’t last, or that you don’t see the application that people imagine, as they imagine it, you don’t see it working the way they think it will?
Kunal: I still think it’s too early to tell. For me personally, and others would likely have a view on what will last, I need to see more use cases emerge. I need to see what works and doesn’t work with consumers. As I mentioned I believe in Part 1, we’re still in the experimental phase with a lot of this. So, I want to experiment, and I want to see what comes out of those experiments before I pass judgment on what will work.
Melinda: So, this is kind of a fun question. If there was one myth or dream or fantasy about artificial intelligence that you wish would actually come true, what would that be?
Kunal: You know, I laugh as I say this but living in Toronto I think we’re all just appalled at the traffic and the commute times in this city. So, I would love to see governments really take transit data and analyze it and use artificial intelligence to make life easier from a commute perspective for the citizens of this city.
Melinda: If the whole city could all be on Waze at the same time?
Kunal: That would be a fantastic use of AI.
Melinda: And I think everybody probably agrees with you on that.
Kunal: And to be honest with you, after being raised entirely in Toronto, it’s probably a fantasy at this point given how the city works.
Melinda: Well, thank you so much. And I’m going to leave a link to your site. And if our listeners have any further questions about artificial intelligence, leave us a note and we’ll get you in touch with Kunal.
Kunal: Thank you very much. This was a pleasure.
Melinda: So, just to summarize the key points Kunal is advising retail brands to think about when approaching artificial intelligence. One, he emphasized quite a few times about transparency in collecting data. That means going above and beyond what’s regulated in terms of privacy, because depending on where you are, there may be very little regulation but even if you’re within a jurisdiction that is regulated, consumers are very sensitive about this issue and are becoming more so. Thinking about how you would want your data treated and how it’s communicated to you is becoming more important. We’ve heard this from many of our guests so I would say this is a critical issue brands need to approach very thoughtfully. The other key point Kunal reiterated was about taking a slow strategic approach by creating test and learn environments, engaging your customers and employees in the process, and really figuring out what value AI can bring to your organization.
If you do want to learn more about how AI could benefit your organization, we’ll provide a link for you to get in touch with Kunal.
- Kunal Chopra is a Financial Services Transformation Strategist for SLD. With a focus on business effectiveness, Kunal has optimized complete service delivery models from the sales force and channels, through product and service fulfillment.
Think Retail is a podcast where top designers, strategists, thought leaders and business people discuss what’s coming next. For more information, email email@example.com.