Cognitive Commerce

Phillip: [00:00:04] Welcome to Future Commerce, the podcast about cutting edge and next generation commerce. I'm Phillip.

Brian: [00:00:29] I'm Brian.

Phillip: [00:00:31] He's timid Brian from New York City this time.

Brian: [00:00:35] Yeah, no, I'm not going to lie. I was late to the show because of NYC. {laughter}

Phillip: [00:00:42] That's ok. That's ok. This is usually me. I'm usually the person in your seat.

Brian: [00:00:46] That's true.

Phillip: [00:00:46] So he's coming at us live today from The Row New York City. Don't try to camp out and get his autograph because he'll be gone by the time you hear this. But anyway, today we are talking about Cognitive Commerce?

Brian: [00:01:00] Cognitive Commerce. Watson. And really anything that our first guest really wants to talk to us about it. So let me go ahead and introduce you, Tom. Oh, well, maybe should we do our intro first?

Phillip: [00:01:15] We did our into. {laughter} This is, we're in the show.

Brian: [00:01:19] Oh we're in the show. Let's do it.

Phillip: [00:01:20] Yeah, this is it. So introduce yourself, Tom.

Tom: [00:01:24] Hi. My name's Tom Robertshaw. I'm an agency owner based in the UK, a company called Meanbee focused on Magento. But of late, I've been very interested in that ongoings at IBM and their move towards artificial intelligence with Watson. So I'm looking forward to having a bit of discussion about that today.

Phillip: [00:01:44] Great. And as always, we want your feedback on today's show. So you happen to be listening to this right now at FutureCommerce.fm. And if you are, just scroll down to the bottom and leave us some feedback in the Disqus comment box below. We also want to remind you you can subscribe and listen to Tom's sultry voice any time on Future Commerce the podcast at iTunes or on Google Play. Or you can listen right from your Amazon Echo with TuneIn radio with the phrase, "Alexa, play Future Commerce." And don't you start with me right now, Alexa. You go back to sleep.

Brian: [00:02:17] Not again.

Phillip: [00:02:17] You go back to sleep. There she goes. I got to get this thing out of my room.

Brian: [00:02:21] No you don't.

Phillip: [00:02:21] So anyway, Tom, how are things? Tell me a little bit about your background and what sort of brought you to the point of what I consider to be sort of a shift in your focus or the focus of your business at Meanbee.

Tom: [00:02:36] Yeah, sure. So my background has always been in computer science. So I'm about five years ago now, I graduated from computer science here in Bath and started an agency because building technology was something that I loved. And eCommerce seemed like a fantastic industry to be working in. After a few years, I feel I've had quite a few experiences. And still I'm somewhat of an expert in the kind of system integrations and building sites the right way and was looking for something a bit more of a challenge and new area to grow in, like many developers and with a bit of an interest in science fiction as well. I love the idea of artificial intelligence and robots and things, things like that that are going to take over the world. And so I started playing with that last year. Math has always been one of my difficult points. I'll admit. As soon as I see an equation, I start to kind of glaze over. So I was looking to see what, actually, was out there was an option. I did start with a couple of simple mathematical algorithms, but decided quite quickly that maybe is something that all that was going to help me, because what I kind of realized is that I would say that I'm an innovator, more of an applier than kind of groundbreaking when it comes to these things. Just seeing opportunities to be made. So excited about what we can do with things that learn. And that's the key to it. I think that we've got very good at programing things now and building systems. I think it's beyond time that we should have built them so that they could actually learn. So we're stopped this manual process all the time, ranking things and doing a bit more. And there's been plenty of progress that they said with kind of the big guys, whether that be Facebook or Google in the last few years. But it's trickling down now. So it's actually accessible to you and I and so for us, eCommerce store owners. And so when I started going about it, I was looking for what services are out there. And there's lots of small players, startups and Y Combinator, if you're familiar with the startup scene, they've had a massive influx of startups.

Phillip: [00:04:53] Sure. Yeah.

Tom: [00:04:53] I've been talking about machine learning and kind of moving in that direction. That's where the industry is interested in at the moment. But in terms of the big players, the big move for us is APIs. So we can now get machine learning, we can get machine learning algorithms without having to have a specialist knowledge to be able to apply them. That's a big, big thing for us. So you don't have to pay the money or spend the time. They expect expensive consulting fees for machine learning. Those things are actually available to you now. So there are APIs for the big guys like Amazon, Microsoft as well. But they do require a bit more kind of background knowledge and statistics, et cetera. But Google have some that are interesting. But I've decided to spend my time on IBM Watson learning what they've done. And I see the Watson program itself has been around for a few years, and the audience might be familiar with it's playing in the competition or the game show Jeopardy and going on to win against some of the best that have played. And now they've opened that up. So they've created a range of APIs that are kind of tailored for specific use cases. So it is a bit easier for you to kind of get underway and recognize how you can apply them to your kind of business use cases.

Brian: [00:06:21] Nice.

Phillip: [00:06:22] Wow, that's a phenomenal intro.

Brian: [00:06:25] Yeah totally.

Phillip: [00:06:25] I'm sure Brian has a million questions because I do a too, but go ahead.

Brian: [00:06:30] You kind of give us sort of like some sort of high level summary of Watson there. I think one of the things that I'm curious to know more about, and I have done a little bit of research on this, but it seems like IBM is kind of labeling a lot of things "Watson."

Tom: [00:06:51] Yes.

Brian: [00:06:51] And so if you could shed a little bit of light on what Watson actually is, you know, I don't know that we might not be able to run a comprehensive list of things that it can do, but maybe just a little bit more definition around what it is right now.

Tom: [00:07:08] Yeah, sure. I found the same problem and it has become somewhat of a branding exercise now, the direction that IBM is going in. But from my research the way they advertise it is they've not necessarily done lots of revolutionary things, but they've got a large number of algorithms, machine learning algorithms, that they've combined together into kind of one interface. So it is like essentially one big computer that is very much tailored for learning and it's applied some of those or that computer to different problem areas. So, yes, it can now work in many different industries and they working in the legal system and the health system and very much kind of advertising as such and going into new areas as well. But from the core of it is the machine learning. In essence, is automated classification is the kind of simplest description. So if you have kind of some training data to start with, so questions that you know the answer to and you give it that to start with, the idea being that later down the line you give a new question that has some similarities on the previous questions and it can give you what it thinks is the best answer. And that is kind of the cusp of just the same sort of question that is applied to different types of problems, whether that be audio, visuals, and/or simple classification of labeled data.

Brian: [00:08:43] Nice. Yeah, like that. Yeah, maybe, I think another thing that would be really useful for our listeners is to maybe give some good examples of things that Watson is doing kind of now and then maybe what you first see people accomplishing with it, maybe in the near future and then maybe also the long term.

Tom: [00:09:05] Sure. One of the the simplest kind of APIs that has available now for eCommerce is one called Retrieve and Rank. The retrieve part is your basic search. So a search that you probably already have on your eCom store. What it has is the second part, the rank, which is the learning part. So you're meant to train this search system with example queries that your customers have made and then either the answer you want, the answer, so the best products, and that should be shown in as a result. There's two ways you could provide that. One could be some the products for those queries that were likely to be sold if they were searched for, or it could be what they say, kind of expert review. So you are the expert of your store of your product catalog. You can kind of browse and search terms and suggest kind of the top products for each one. And as the training, and it can learn from that. And kind of taking back control, as I see it, as this has been a growth of improved search services and products that you can integrate with your stores in the last couple of years. I see this as kind of a step forward in terms of actually understanding and being in control of the products that are recommended, rather than completely offloading it to a third party without actually kind of getting any knowledge insight. And that's why you touched on Cognitive Commerce. And that's where I see that being is actually having a bit more knowledge and insight about the sales that you're making, your customers, et cetera. That's certainly one.

Brian: [00:10:44] But yeah, definitely.

Phillip: [00:10:44] Can just break in?

Brian: [00:10:48] Go ahead. Yeah.

Phillip: [00:10:48] One of the things I think we said right at the outset of the charter of our show was I'm really interested in what's available today. But I'm also kind of interested in the thought process about what the future holds. My sense is that we only... This is like such brand new ground that the only thing that we can think of at the moment that is like the hello world, step zero sort of an example for a merchant is that rank in and search, which is basically like what Tom just said.

Brian: [00:11:30] Yeah.

Phillip: [00:11:31] The things that I think are coming on the horizon, which maybe Tom can speak to, is more assistive to people, to humans that are doing the work, not necessarily suggestions like what we've seen at Amazon employee, you know, which has been great, product suggestion. But what the retailers I work with just don't have a catalog large enough to really support, you know, spending any sort of money or time to improve discovery. You know, most of the retailers I work with have a two hundred piece catalog. If anything, they're asking me to make their catalog look bigger than it is. So my sense is that most of this could be assistive in nature, like tools that would help a marketing analyst or someone who's doing merchandizing to resort and prioritize products in their store or to help them make decisions about promotions that are most effective over time. Do you have any thoughts in that, like the what's in the midterm for us?

Tom: [00:12:33] Yeah, sure. And I'd start off by saying I totally agree. Like the first things that we going to do with machine learning as we understand it is just going to find better solutions to things that we already had solutions for, and then once we get from a more familiar with it, we'll break new ground. And that is something that I think that doing those kind of two different tools that have been released, I think, are very pertinent to what you're talking about. The first would be Watson Analytics, which I haven't had exposure to, but have done a lot of reading up on. And as far as I can tell, it is that kind of that takes on steroids of really delving into the data as to whether or not that just be kind of you uploading a spreadsheet of data, but not having to kind of say which metrics are important, but it can kind of feed all over that in and start telling you things about it. Things are interesting. The IBM Commerce has Commerce Insights, which integrates Watson Analytics. I mean, you kind of have on the page you can browse your site and see kind of a category of products which products are performing well or poorly for this particular category. And you can start to delve into why, like right there in the interface and then merchandise and see if there's a display issue and things like that. So I think it's a very interesting area that IBM is after integrating into their own eCommerce system. Quite rightly so.

Phillip: [00:14:00] Is that so, though? Are we talking about, like, WebSphere?

Tom: [00:14:03] Yes. Yes. So Commerce Insights like plug in to WebSphere.

Brian: [00:14:08] Right.

Phillip: [00:14:08] Ok.

Brian: [00:14:09] Nice.

Phillip: [00:14:09] Interesting.

Brian: [00:14:09] Yeah, I think it kind of highlights a little bit more of I think how this really relates back to IBM's vision of Cognitive Commerce, because I think the idea is, you know, at least as much as I can I have been able to interpret it, is that they're actually looking for... They're trying to portray at least is that Watson is going to be like a member of your team and it's going to be able to accomplish things for you that you would have had to spend a bunch of time actually doing on your own. Like the story is basically, you know, instead of spending your time on one specific task, you let Watson take care of the things that it's good at and then you focus on the things that you're good at.

Tom: [00:15:05] Yeah. You're like, working on things...

Brian: [00:15:08] That's what I was reading about on the Commerce blog. On IBM's Commerce blog.

Phillip: [00:15:12] {laughter} He drank the Kool-Aid, I think.

Tom: [00:15:15] Yeah part of their Kool-Aid is that like following your instincts of a eCommerce today is not like a successful business strategy. You might get lucky in the short term or maybe even the long term, but, you know, for real results you need something that is providing you insights on the data. And we've gone through this phase of having lots of data available to us and lots of talk about it, very little in the way of tools that can actually give us true insight in our industry for it. So, yeah, instead of their movement to support rather than this kind of spray and pray eCommerce, where we kind of try and do a bit of everything. We know what we should be doing. We just can't get around to everything to the degree that we should be. It can try to lighten the load and do some of these things for us. You know, the visual merchandizing is a good example of that.

Brian: [00:16:06] Yeah. I think that that's really consistent with my experience with the merchants that I've worked with. They really do not have enough time and resources to really take advantage of the tools that are available to them. And so it's often, it's just so often that a tool will just sit around unused because they had to focus on something else or felt like, you know, their best shot at working on improving conversions was, you know, the merchandizing. Well, they spent all their time on that, but they didn't spend any time on improving their UX and focusing on CRO. And so I think the vision is, I think, an accurate one. I should say, it's a worthwhile vision because I actually feel like this is something that merchants really need.

Phillip: [00:17:02] Well, can I play devil's advocate or human's advocate? I'm on team human. How's that?

Brian: [00:17:08] Go ahead.

Phillip: [00:17:09] My sense is that there will always be external or mitigating factors that Watson will never, ever or any AI or any machine learning technology will ever be able to take into consideration. For instance...

Brian: [00:17:24] Never? Did you say never?

Phillip: [00:17:26] I said never.

Brian: [00:17:27] Ok. Ok.

Phillip: [00:17:28] Because there are things that I'm picking up on Tom's accent that he might, you know, he might understand what a Brexit might be. But there are things that happen in this world that are external factors that are very difficult or nigh impossible to predict. For instance, a really good example here would be supply forecasting. People have been using machine learning and AI for supply forecasting on the manufacturing side for a decade. But nobody could have predicted 10 years ago a tsunami taking out all of SanDisk and Toshiba and Cannon and taking out these major manufacturing facilities in Thailand and the Southeast Asia Pacific. So there are things that will always be mitigating factors that don't allow you to make accurate predictions.

Brian: [00:18:25] Sure. I think that there's definitely an element of that. I do think that we're going to get a heck of a lot better at predicting things that we don't feel like we can predict now. That said, I don't know...

Phillip: [00:18:37] And that's pretty nebulous. Give me an example.

Brian: [00:18:39] Ok. An example is, you know, the tsunamis. I mean, we might in the future, we might have a better way of detecting those.

Phillip: [00:18:48] Ok. Fair enough. I mean, maybe that's true. I'm thinking more about the things, the impacts of your business where if given a small enough sample set, you could look at, you know, a retailer starting out today training a machine learning algorithm for merchandizing could look back and and only have a year's worth of data or maybe a year and a half worth of data. And that machine learning algorithm could learn to interpret something that's a blip, like a downturn in demand from a Brexit as some seasonality or some seasonal factor. And it would take a long time to train that attitude out, whereas a human would understand.

Brian: [00:19:28] Time is certainly a factor. I think never is a very strong word.

Tom: [00:19:32] Yeah. I'd say, like, machine learning is never gonna be perfect. Like our aim is for it to either be as good as humans, so that we can kind of get them doing jobs that that machines can't do or ideally or even better, which they are plenty of jobs. But also like if we broach the subject topic of Brexit and like the financial chaos for that, the first trainers that have actually kind of been doing better in the financial markets have been the robots. They are like the machines that the automated transaction. So there's something to be said for it.

Brian: [00:20:11] For sure. For sure. Oh, I've got to bring up an example of something that plays to your point, Phillip, and I'm not disagreeing with you. I think that...

Phillip: [00:20:19] No. This is healthy debate because I don't necessarily hold the viewpoint that I'm arguing, but I think it's interesting.

Brian: [00:20:25] Yeah. No, I definitely think it's interesting. I think this definitely can be said. We are not as far down the AI path as a lot of people may believe that we are. A great example of this is Sun Spring. Sun Spring is a project that was directed by Oscar Sharpe and AI researcher Ross Goodwin to make a movie for the Sci Fi London Film Festival. It's starring Thomas Middleditch, the guy from Silicon Valley, and essentially applied AI to read a bunch of scripts and then write a short sci fi script for this film festival.

Phillip: [00:21:16] So the Netflix model, but for writing content.

Brian: [00:21:19] Yes. Exactly, and in short...

Tom: [00:21:21] Still better than Adam Sandler.

Brian: [00:21:23] Oh, definitely better than Adam Sandler. No doubt about that. But in short, go definitely wash. Go, go check it out. Everyone should go watch this, first of all, because it is I mean, I think Thomas Middleditch and Elizabeth Gray and Humphrey Care do a fantastic job putting a nonsensical script together as something that is actually hilarious.

Phillip: [00:21:50] Oh, wow. Ok.

Brian: [00:21:52] It's very much worth watching. I think it really demonstrates ok, look, you know, AI gets sort of grammar and like really high level what what's kind of going on, sort of. But it really cannot produce a script, not even close.

Phillip: [00:22:12] And it's probably a heavy ask on a technology like that to do something like that.

Brian: [00:22:20] Super heavy. Yeah. But I know it's kind of a ridiculous example. But my point is, you know, there's so much left for AI to do that's a very long way off. And so definitely temper your expectations, understand what it can and can't do. And I think getting in there and digging into Watson and seeing how Watson could help your business is very different than thinking about AI as just this autonomous practically being that can just help you.

Tom: [00:22:54] Yeah. As Phil has touched on already, like this machine learning stuff isn't new. It's been around for decades already. Even in terms of the algorithms that are being used now, there's kind of only been sort of one improvement in the last few years that has actually kind of made some of this possible with the rest of it is just kind of the cyclical nature of like fads in computer science and every other industry, as well as the increase in computing power kind of gives us. We can do a bit more than we could do in a reasonable amount of time now, which makes it interesting. But there is like kind of any new kid on the block. There's a bit of hype over it and it's kind of cutting through that to see what can actually be done like this time around. How far are we going to get with the tools available to us now? And I think that it's still so interesting to us.

Brian: [00:23:45] Absolutely.

Phillip: [00:23:46] I'm interested to hear, Tom, your take with you know, there's so much lock in that's occurring with these technologies at this point. You know, they always go through for mass market adoption. You have to have big players that simplify the use. And so you sort of democratize the access to these sorts of very complex mechanics. A good example is, is IBM producing an API for Watson so that you don't necessarily have to run Watson yourself at home or there have been machine learning and hive trainers for five years that are python packages for the developers that out there. Like you could get this stuff before. There's nothing that was preventing you from doing it. But when Amazon releases for its Web services a pay as you go machine learning, you know, like a consumer tool kit that is simplified just to the one or two things you might need. And it sort of locks you in to a vendor, I guess is what I'm saying. What do you think about, um, are we on the leading edge of that? Are we going to see tremendous amounts of vendor lock in? If you if you really want to take advantage of Watson, I really need to be running on SoftLayer, and I got to buy all all the way to IBM. Or do you see this as...

Brian: [00:25:16] That's a really good question.

Phillip: [00:25:17] Because for technology to grow and to be used by the mass market and to actually be groundbreaking, I really think that it needs to be something that can sort of be consumed a la carte. So what are your thoughts there?

Tom: [00:25:32] Yeah, I think it's a very good question. And I think that the kind of the the simplified API is going to be very good for one set of companies. There's going to be those companies that don't want to kind of, you know, even give their data to a third party. Certainly not all of it labeled and the concerns about giving that out to a potential competitor.

Phillip: [00:25:56] Right.

Tom: [00:25:56] So, yeah, there is going to be people on both ends of of that spectrum. And it's something to consider. I think, from talking to you in the past, I think we both value eCommerce store and having a hold on their own technology assets rather than being hosted or owned elsewhere and very reliant on like a third party company. Yeah, that being said. I think the simplified APIs are useful for kind of getting out to the mass market, as we said, in terms of developers and developers can get involved in that. But it's also the amount of effort they're going into to explaining this to the business people. So they couldn't recognize it because, yes, there might be many libraries available that developers are kind of playing around with, but that might not get the buy in from the business or not really understand it, but companies like IBM are helping on both fronts and actually educating in what's possible with it to get kind of business people kind of fluid up on it, which is very useful. The next step as well, which we haven't quite reached yet, which is actually for it to really do well, is for kind of actual end customers and users to understand the process, what's going on in the background. It's not going to kind of take off if it's completely hidden from the actual customer. There are definitely use cases for that. But they're going to want to understand in order to kind of trust the data that comes back. They're going to want to have some understanding of where it's coming from and why, I think. So, yeah, there's definite risks for the kind of basing it off of one technology or another. The only other thing I would say on that is if anyone has done some machine learning, the most costly part of it is not necessarily the actual kind of connecting to a system and using, say, IBM machine learning, but it's actually kind of gaining that sample data and manipulating it with the training data, rather. So that your queries and the labels against them in terms of what type of query is, what is the customer talking about, that is far more costly in terms of time than the time to actually integrate with a third party system. So it's quite likely that if there's more and more of these APIs from different providers are available, then same as if your newsletter company updates price lists or something like that, you'd be able to, you know, migrate across to somebody else.

Brian: [00:28:24] Oh, man. That brings up some of the questions now. {laughter} I think one of my next questions would be, do you see people sort of building out practices within their companies to spend that time and that effort to do that? Or do you kind of see them looking elsewhere out to third parties and other business practices to sort of develop that our as their specialty? Or I guess what I'm saying is, how do you see an ecosystem forming around this?

Tom: [00:28:58] I'm not sure. There's two ways of getting the training data. One is, from your point of view, users themselves. So they're going to buy their activity on the site. And they're providing information so that enables you to kind of build up an awful lot of information very cheaply. But if you want the kind of expertly trained data, then you or people within the team are going to have to kind of sit there and kind of manually classify, like, for example, the easy example of the product that should go to it should be sent to the customer with this search term, which is very arduous process that aren't going to necessarily produce great results depending on the person. But also how much they enjoy that task is going to have an effect on the results as well

Phillip: [00:29:48] Your job changes from the person who makes intuitive decisions to the person that just classifies the data and let something else make the decision, right?

Tom: [00:29:56] Yeah. And you're not going to get much buy-in from that. So this is going to be a very difficult integration process. And there's plenty of research studies, which I'm kind of still catching up with in terms of how to gamify the system, so that people are going to take part in it and they buy into the process. And normally that's to try and make the feel like your input is more valid rather than it just being a case of like, you know, send this product or this is this type of image or that type of image, actually kind of being able to give more feedback to the learning system and tell it what it's doing. Well, tell it where it's doing badly and it kind of recognizing that, being able to apply it. You know, that's kind of the next step, which, you know, I think it's not, the limit isn't technology at that point for that kind of interaction, it's more just our familiarity with how to build those kind of systems. So I think that would help. But yeah, the training data is your biggest stumbling block when it comes to actually implementing some of these systems.

Phillip: [00:31:04] And I have to wonder, too, what like there's always from a merchant perspective, you kind of bring it back to digital commerce for me is  how you can justify the ROI on some of these sorts of tools. Again, if they're not just built and ready for you and plug and play where it's, you know, it's something that's quantifiable as every percent of every transaction that results as a result of using this tool. Like, it's very difficult to to justify the R&D because, you know, what are we building here? You know, some ideas or some thoughts that I had are well, wouldn't it be nice if we could apply... I'm trying to think of the sorts of problems that merchants have right now. A great example for me is as a digital agency, I for whatever reason, I am unable to get my clients to tell me in advance when they're going to do a forty five percent off sale. I don't know why. I'm physically unable to get them to do that. And if I could build something that would just subscribe to their email list and the second that it sees that email come through it would auto scale everything up, that would make my life easier.

Tom: [00:32:18] Yeah.

Phillip: [00:32:18] It's not going to save anybody any money. I'll be honest with you. You know, there's no justification to building that sort of thing. It's I'm going to make that decision or the machine is going to make that decision. I'm just trying to think of, you know, there has to be... You know, some people are going to go out. They're going to cut a lot of new ground. They're going to forge the way. A lot of them are not going to be successful. And we'll learn from the mistakes.

Brian: [00:32:39] Actually, let's get into the super practical here. The real world stuff like 1 800 Flowers introduced Gwen, which is powered by Watson. It's a discovery tool.

Phillip: [00:32:52] Is it me or is 1 800 Flowers doing literally everything?

Brian: [00:32:55] They always do everything. It's actually pretty amazing. I'm always blown away. I can't say that I really ever use them, but I do see them do a lot of stuff.

Phillip: [00:33:07] I feel like whenever there's something new, it's like, well, you know, and they have a new Slack bot. Why do I need a Slack bot for... Anyway, whatever. I've become a curmudgeon. I don't know how. I didn't mean to.

Brian: [00:33:20] You might be getting old, Phillip.

Phillip: [00:33:24] That's probably true. Yeah. Yeah. So sorry I interrupted you. You were talking about practicalities.

Brian: [00:33:30] Oh yeah. No, I was just saying, you know, it's a discovery tool. It's in beta right now. Watson powers it, and this is going to lead into some more questions. But I think are they going to see a return on Gwen? I have no idea. Like, is this tool actually going to be something that people actually find useful to pick out the perfect flowers for whatever occasion? I mean, there are a lot of different flowers and arrangements and so on. I mean, when I've shopped for flowers before, it's been quite a challenge and maybe it would be useful to have something like this. But did you guys check out Gwen yet?

Tom: [00:34:10] No, but I did see was actually I think it was a couple of years ago now that IBM with Watson, again, did something with North Face. Very similar. Discovery tool.

Brian: [00:34:19] Yes, that's right. Right.

Tom: [00:34:20] So you have that kind of mocked up video of the CEO of North Face, him sitting there saying that he's going away hiking to this location.

Brian: [00:34:30] Right.

Tom: [00:34:31] Like in the next month and needs a coat. What should he be wearing, and Watson goes away and walks out from that catalog with this is going to be three degrees and this is kind of the best for you because, know, expecting rain and stuff like that. So even though I haven't seen much more, I was looking around news articles for kind of the results since I think it was Q4 2014 launch and I haven't found anything. So, yeah, that's the challenge for people that are kind of the pioneer, is that they're not necessarily going to be the ones that reap all the benefits because it's such a new experience for customers as well. And we're al going to be learning very much, learning the way.

Brian: [00:35:08] Absolutely. Yeah.

Phillip: [00:35:09] Something like this is a recurring theme now. Whenever we talk about Apple, we have to talk about how Microsoft did it first and they go through all the learning experiences.

Brian: [00:35:20] Yes.

Phillip: [00:35:20] Yeah, it's interesting.

Brian: [00:35:21] I think also... So let me take this a step further and get a little bit broader with it. I think it's really interesting to me that with 1 800 Flowers, they called their assistant Gwen. They gave it a name, but they said it was powered by Watson. And yet the same time, if you kind of watch IBM's Watson commercials and how they're sort of portraying themselves as being in their commercials, it's almost like they're trying to compete with Alexa. And I know you brought up Amazon earlier, Phillip, but I think my question for you, Tom, and this is definitely conjecture. So just spit ball here. But how do you see Watson actually interfacing with the public? Do you see it solely as being something that powers things behind the scenes and then we have all of these different named bots popping up like Gwen and others? Or do you see Watson sort of taking a more public face, as it kind of has or IBM has been sort of portrayed it and actually kind of competing with Alexa in some ways?

Tom: [00:36:31] Yeah, I think Watson is going to be the only one even on top of the kind of the Watson API or Watson API has that kind of dialog and conversation system. So if you wanted to create your own Slack bot, chatbot, you can already do that. So I think people could continue growing on top of that. And Watson will be going to the older brother doing the job of kind of telling the neighborhood about what he can do, or people like him can, what they can do. So I think we very much need that. So I think the marketing and the Amazon Alexa will support that as well just to kind of get people, and there's a lot of people, a lot of businesses putting a lot of money into this bit of conversational commerce, conversational interactions, and whether or not that's something that end customers are actually going to want. So it is still to be seen whether that's an interaction that is going to be the most successful but is certainly...

Brian: [00:37:32] Oh I lost Tom.

Phillip: [00:37:35] No, we still have Tom. We still have Tom over here. Yeah. Are you still there, Brian?

Brian: [00:37:41] I am here. I can hear you, Phillip. I just can't hear Tom.

Phillip: [00:37:44] To recap what Tom had said, he said something to the effect of, you know, there's a lot of build out and a lot of discussion around investment in conversational commerce at the moment. But whether or not that's something customers want or something that will be successful or provide a real return on investment is yet to be seen.

Brian: [00:38:05] Yes. And we talked about this quite a bit already. The truth is, this conversational economy is going to happen because people are investing in it, right?

Tom: [00:38:20] Yes.

Brian: [00:38:21] And so there's really no way around it. I think either you're going to completely ignore it and hope it doesn't really pan out or you're going to have to get out there and invest in it and plan it and kind of figure it out a little bit on your own as well. Otherwise, you're going to be missing out on a major channel.

Phillip: [00:38:44] It's so funny because I feel like three or four years ago we were saying, who the heck wants an app on their phone for X brand? Like, what's the point? And now you fast forward a little bit and there's a Taco Bell app. There's a Chick-fil-A app. There's a McDonald's app. There's like a million pay apps for these sorts of one off transactions that we always do. And I feel like those weren't necessarily pushed upon the companies from demand from the public. I feel like certain players were very successful. And so there were land grabs and rushes to that space.

Brian: [00:39:24] Yeah.

Phillip: [00:39:24] A good example is Starbucks. Starbucks was extremely successful with its loyalty app, and they had a very interesting and engaged and very affluent market. And now we see other food and restaurant chains and now even some department store chains that are moving to this model as well. Again, not necessarily because of market forces demanding that they go there, but but that's because the consumer, at least the affluent consumer is being trained to shop in that way. I think that businesses aren't going to miss out on sales opportunity or they're not going to miss out on opportunity because they don't have a chat bot. But their customers will eventually either be trained to want to chat to engage in commerce. And that will cause them to want to invest there. Or they'll be one of the first that sort of get into the fray and do it poorly. I just I don't know that the American consumer and maybe, Tom, you can give a different perspective. I don't know that the American consumer even understands yet that I would shop at Macy's on Facebook. That makes no sense to an American consumer at this point. Or Facebook messenger in particular.

Tom: [00:40:47] Yeah, right.

Brian: [00:40:49] {laughter} Yeah. Especially given a couple episodes back when we talked about how difficult it is to shop on a Messenger app at this time in America.

Phillip: [00:40:59] Yeah. I believe the pull quote was, "It's literally the worst way to shop ever." Something like that. You're taking the worst experience possible.

Tom: [00:41:09] But I can see the kind of the use cases that where it could provide a better experience like I don't want to cover old ground. So stop me if you have a way to discuss this. But I think the article that I was reading was by the guy that started Kick, the Messenger, so he was kind of weighing in on his opinion. And the app system, as we know, works for those that have customers coming back over and over again. For the smaller and not global brand is it's still the kind of way too expensive an option and unlikely to see the return on investment. So what do you do? I think his example was I mean, if you were going into a coffee shop or if you going into a baseball stadium or something, something like that, like trying to download this new 70 megabyte app to place your order for drinks and food isn't going to be the best consumer experience. Kind of going up to the line and waiting ages isn't going to be the best experience. And actually, you know, maybe if there was like a WhatsApp bot that you could talk to, it would give you a list of options for your order and then you go and collect it. You don't have to download and text. Chat is cheap and could... I'm completely basing this on no evidence whatsoever, but it could be a nicer way for customers to place orders.

Brian: [00:42:39] And certainly, I think, you know, when you say you have no data to base that on whatsoever, I think that, in America at least, and then maybe in Europe, this is still definitely early, really rudimentary. We talked about this also recently, whereas like the data that we do have is from China and in other places, other kind of developing places where purchasing through text and chat is actually already a very developed ecosystem, developed method of purchasing. And so the question is, how does that translate for us? And so I think you're right, there's not really that much data that we can go off of right now to really speak to how we're going to interact with it.

Tom: [00:43:31] Yeah, I think that payment integration is an interesting point. And some of it does depend on whether or not there are closer integrations. You know, I haven't necessarily kept up with if they have already been ones announced. But payment integration built into the messaging system. And that's what the kind of the biggest chat app  does have. So it does have its advantages there.

Phillip: [00:43:51] Which I believe is the killer app for Facebook, Facebook Messenger. You know, if you put your banking information, you trust that to Facebook. And if you have payment data stored in Facebook Messenger and in theory, if you could just show your Facebook Messenger QR code at a checkout like an in-person like a kiosk or a Pop-Up store, all of a sudden that, you know, it closes the loop on the multipurpose use of a chat app. It's not just... It can be used in real world scenarios exactly in the same way as your, you know, both to verify identity and for payments and for, you know, conversational exchange. So to your point, I do see some value in it, but I don't know that the value is there because the market is demanding that we have to go in that direction. The value's there, because I think there's a lot of people stating that there's potential future value there. I think what we need really to have a better conversation around this is to have somebody who really understands the Asian market in particular. You know, the chat apps that that are already doing this and people that are already seeing a tremendous shift in China.

Tom: [00:45:11] Absolutely.

Phillip: [00:45:12] WeChat and Weibo already have been doing this for five, six, seven years and somewhat out of necessity. Some of that market is underserved and the population density is tremendous. Something to the effect of I'm going to get this statistic wrong. But there's like over 30 cities with over a million people in China alone. And so the population density alone, I think allows for more innovation or allows for people to have to work around traditional means of commerce.

Brian: [00:45:50] I kind of wanted to cut back to Facebook for one second too. Just speaking of Facebook, last week, they announced some pretty big stuff.

Phillip: [00:46:01] Oh, yeah, yeah, yeah, yeah. Hit it up, because this is an interesting news that we didn't get on the last show.

Brian: [00:46:05] It is interesting news. We definitely hit on this. So last week, Facebook announced persistent menus in Messenger, which was severely needed. It was very apparent to me when I started using Masina Bot that the way that the UX worked was just not sufficient for actual interaction because you constantly had to scroll up to figure out what you wanted to do next and you couldn't just chat. I mean, you could chat some different shortcut commands to get back to different things. But who's really going to remember those things?

Phillip: [00:46:52] You have to learn them. Yeah, such a hassle.

Brian: [00:46:54] Forget that. Also announced gift sharing, which is fun and useful, and then this is pretty exciting and announced an entire series of guidelines and a blog just around...

Phillip: [00:47:08] I think the guidelines is the key here. Something that you and I were talking about a while ago, which is the people I think that are going to be successful are the people who are going to determine or put out best practices and documentation of the best way to interact with this new technology.

Brian: [00:47:27] Yes, absolutely. Very exciting stuff.

Phillip: [00:47:31] Yeah, interesting stuff. Very cool.

Brian: [00:47:34] Also, they announced that they have eleven thousand chat bots available on Messenger right now, which kind of blew my mind.

Phillip: [00:47:43] Ridiculous. I think they're just taking the Pied Piper approach to the growth model.

Brian: [00:47:47] A little farming.

Phillip: [00:47:52] It just seems ridiculous. Who in the world is out there creating eleven Facebook chat bots?

Brian: [00:47:58] I don't know. They are, from my understanding, they are relatively simple to create. I don't think that those eleven thousand chat bots are going to be used, frankly. But as Facebook sees more adoption of Messenger and usage of Messenger, the top ones will definitely get used. But all of those, I can't see people using all eleven thousand of those right now.

Phillip: [00:48:27] Yeah, I feel like there was some sort of statistic. Something like, you know how there's... I'm the worst that numbers, I should never do this. There is some kind of statistic of there are over like 20 million registered Facebook developers or something like that. People that are registered as Facebook developers, which is, you know, I don't know. So I suppose if you took like one half of one percent of those and they all just went in and clicked and created a demo app, then you could come into the 11000 number.

Brian: [00:49:03] Yeah. That's true.

Phillip: [00:49:05] Interesting. I think it is interesting. I'm more concerned about, you know, kind of getting back to the sort of AI, yeah the cognitive commerce topic of the show. I'm more I'm more interested really in the the consumer aspect to this. I think that that's the real transformative technology. They say it takes, what, seven years or something to that effect for a technology to be really transformative. We've seen, you know, it's been just about exactly seven years. I'm sorry, it's been longer than that now. It's been nine years since the iPhone, but seven years since the App Store and only six years since copy and paste on iPhone. So there you go. But I think we're still in the dawn of what, you know, mobile computing can do. We're still in the predawn of what consumer available artificial intelligence is going to do. But I really think that the the killer app here is the always on, always available experience rather than me having to, you know, interact with typing with my thumbs. I think me being able to speak aloud to a device like an Echo is really the consumer approach to the AI challenge here,

Brian: [00:50:29] I would definitely consider the release of the Echo is sort of that starting point for the clock, if you will, for this whole new initiative. I think, you know, before that, when Siri came out, I don't think that Apple ever really capitalized on that vision. And it was Alexa that really got it started the way that we're going to think about it even ten years from now. Alexa is going to be... We're going to look back at Alexa and be like, "Oh, man, that was hot new stuff when it came out, you know, cutting edge stuff." And fifteen years from now, we're going be like, "Oh, that was just so silly."

Phillip: [00:51:12] It just sort of burns my biscuits when you look at the people that are making real money right now that are essentially just applying these lowest common denominator applications. And they're out there making money and I mean, good on them. But there's so much more innovation in the space that we can do. You know, there are plenty of companies, there are tons of them, that you could even have VC funding for all I know that all they do is they put product suggestions on your website and in your emails. And, you know, that is the lowest common denominator. So I think there's a lot of room to grow. You're right. We're almost out of time. Tom, do you have any closing thoughts? I want you to be able to pump some stuff and tell us where we can find you. But I want you to have the last word.

Brian: [00:51:57] Also if you have a project you've been working on or something, just anything you want to say.

Tom: [00:52:04] Oh, wow. So many questions to end on. I think we've talked about kind of the natural language stuff quite a bit, which I think is definitely the area to watch. And that's, you know, the bias towards actual spoken voice. But the same applies exactly the same as text as well. But the interaction method via text, we're not so familiar with either. So actually just typing a sentence to add to something and it understanding what you're getting at. I think both those are definitely, definitely something to watch. So that helps with eCommerce, that helps with the discoverability, kind of the building up the rapport and whether or not we can. I don't know that that will be the Turing test of kind of faking it. You're actually talking to a real person. But I mean, they did manage that with a Columbia, the college university that it was. But there was a computer science course that was one where one of the tutors, it was kind of on the forums and was actually IBM Watson that had been trained on kind of previous forum answers, and they were the students were told it after the end of the class. And they didn't realize. So we do seem to be making progress there. So I think that can help us if we can apply that to that personal shopping assistant experience. That's definitely one area to look out. I mean, you did talk about this just then about the newsletter and kind of promoting products. That's how cohesive commerce that we've had for a little while now. And the next step is how do we make that even smarter? I think the Watsons API to try and help that is their personality insights that kind of sentient analysis. So you get a bit more of a natural kind of understanding of the customers and personalities. So you can maybe better segment them. You know, if you find that they can have very open customer than you might be kind of pushing kind of rewards based on social media sharing things. And you might be able to know better and more intelligent targeting rather than the targeting, which at the moment is more based off of how often they shop with you and the amount of money they spend on what shopping category. Kind to go into the next steps there in giving that, I don't want to use the phrase, but I will that truly personalized experience. So these are a couple of different areas for to be looking at with eCommerce. But the rest of it is seeing what the pioneers are doing and hedging your bets, doing some experiments. But as you said, there is money to be lost in the R&D of these. And depending on where your brand is as an eCommerce company, you need to decide how much money and what you might want to kind of be experimenting with whether or not you want to be the leader, because, you know, if that's important for you as a brand and if that's important for you or for your customer base or whether or not, you know, you wait a little bit longer, you keep an eye on it. But you wait for the the systems that kind of roll it up into a package for you and you recognize when the time is right to employ that particular package. So it does very much depend on you as a business and of your responsibility, so identifying what kind of business that you are, what's reasonable for you to achieve, and then understanding some of the opportunities that are available in order to choose the one that you think is best for you.

Brian: [00:55:34] Good thoughts.

Brian: [00:55:35] Wow. Well said.

Tom: [00:55:37] Thanks. And in terms of me, the company that I own, Magento Development Agency, is called Meanbee, and that's Meanbee.com. And myself, I'm @bobbyshaw on Twitter, and I'm sure to be doing more stuff with IBM Watson. I've played with a few of the APIs already. The visual recognition I think is really cool for actually being able to recognize objects within images that you train. So I've done that on a real estate site to recognize whether or not it was the pictures of a swimming pool, whether or not it was an inside picture or an outside a picture and just a drawing of, you know, once we get more informed information, the platform to be able to take it and take control over, you know, things that previously were just like we've got a media gallery. We don't know which images which. So that's something I've been working on. The concept Insights is another API that people should be looking out for. And that's kind of the next step in terms of being able to get the search that Google does so rather than just being your documents indexed by keywords. But Watson has some understanding actually being and talked about in terms of concepts. So that next again, all of this is around the next steps and doing things a little bit smarter to get slightly better results. I'm sure, to be playing around with another API next. I might try that retrieve and rank one for improved search for eCommerce. But there's also the tradeoff analytics, which is we're familiar with eCommerce filtering. So you've gone through your categories. You've got a bit of an idea what you want and you start narrowing down the products are available to you. You might choose, say, the price bracket. You might choose, you know, for a mobile phone, you might choose the battery life or the weight of it. But actually, like if we were browsing around, kind of physical store our requirements aren't quite so rigid and so tradeoff analytics is this kind of optimization between all of your objectives. So rather than just filtering down to this price bracket, it might come back and say, "Well, you kind of valued battery life and this one is far greater, but it is slightly out of your price range." And that's an experience that we don't really have online yet. And I'm not saying it's something that's going to take off for one, the kind of user experience and the user interface for that. I'm not entirely sure, because if you kind of change the filters as they are now to work like that, people might get a bit confused or kind of complain the site is broken. But that's, I think, another very interesting one to see how we can move some of that offline commerce and experience to people. People have to treat people as people in the online word.

Phillip: [00:58:24] Oh you just blew my mind.

Brian: [00:58:24] Love it.

Phillip: [00:58:24] I wish you hadn't done that right at the end of the show. {laughter} Oh, that's phenomenal. Well, thank you, Tom. Thank you for all your time. Thanks, Brian. I know you're having a little bit of connection trouble there, so I'll go ahead. Yeah, and thank you all. So anyway, thank you for listening. And this is a Future Commerce. We want you to give us feedback. And so if you are listening on the site there at FutureCommerce.fm, we we want you to scroll all the way to the bottom and leave some feedback on the show and the Disqus box. And we really do want you to make the show what you want it to be. So we need your feedback. Also, you know, we want you to subscribe, so you can always have the latest episode of Future Commerce. So you can get it on iTunes or on the Google Play podcast store. Or you can listen right from your Amazon Echo with the phrase, "Alexa, play Future Commerce." And I didn't trigger her that time. I'm getting better at it. But anyway, thank you so much for listening and we'll see you next week.

‍