Episode 16: Why retailers are racing to adopt AI-powered personalisation

On the ground at Shop Talk Fall, Amie is joined by Diane Keng, CEO of Breinify, to discuss implementing artificial intelligence in the retail space – and the challenges companies face with planning, resources and legal considerations. 

Transcript

Intro:

Coming up, on this episode of Retail Untangled…

AI power personalisation is really key for this upcoming year. And everyone has different flavours of how they're deciding if this is the right approach. Some are really focused on working with their existing technology and trying to activate it internally. Some are r eally focused on bringing in external help that can overlay across a lot of things. But at the core of it, comes down to how do we know AI is effective?

Amie:

Welcome to Retail Untangled. My name is Amie Larter and this is the podcast where we speak to retail industry experts and find our business hacks to help you succeed. You won't find these gems anywhere else and we have some superb stories from the coal face as well as helicopter insights from retail industry leaders. This week we're bringing you insights live from Shop Talk Fall in Chicago.

Today I'm joined by Diane Keng, co-founder and CEO at Brainify, a predictive personalization platform for retailers. Diane, you've had an impressive career trajectory from launching two startups before age 18. Wow. You then went to join Apple and fast forward to now founding Brainify. This journey has given you a deep understanding of marketing automation and now AI.

So to kick-start us today, I'm curious to explore the current state of AI adoption in retail. What are you seeing?

 Diane:

Yeah, for sure. I think right now we're hitting this new trend where people are like, we have to do AI. But they don't really know what they're doing with it. They just know that it's required now. And so many of the brands, when you think about retailers, they've been sitting on this treasure trove of first-party data. And now it's coming to the realisation that they actually don't know how to use it well. And for a lot of the consumer packaged goods brands, where they don't directly sell on their sites, they're realising that the importance of having first-party data is becoming really critical. And all these things, they empower how an AI is actually going to think and perform. 

But the truth is, many of these brands don't have data scientists and so you can imagine it's very hard to work with AI in a way that's practical and then be able to actually get and yield results from them. So one of the big trends that we're starting to notice is that over the last few years personalisation isn't a new initiative. It's been around for decades but what has changed is the adoption of AI and the concept around what are things that we can do that are low-hanging fruits. 

But what that's also created the last few months is a lot of kind of like piecemeal solutions where everyone gets one thing and inside there's some sort of AI component but all of it becomes super super disjointed. So whatever system you're using for your website doesn't correlate and bring the same experience inside your email and vice versa to other channels and so you end up with like very very different groups of silos of data and experiences. So that's one thing that we're noticing. 

The second trend that we're also starting to see is that I think marketing teams are really really interested to be able to do faster tests and learns. I think a lot of individuals have great ideas. They're just unsure of how it's going to perform, but they're really bottlenecked by the ability to actually run a test. And those tests can be as simple as just a layout on a site to something that's a little bit more complicated. Like, how do you define similar items? Is it similar because it's the same price? Is it similar because it has the same category or colour? So even just that concept of like, what is the right kind of strategy and approach for my business? 

We're starting to see that take hold and those challenges start to arise as well. 

Amie:

It's hard, right? I mean, we've got Wendy's and Dermalogica are speaking on implementing a data and AI strategy across whole organisations this afternoon on stage here at Shop Talk. I'm keen to understand your advice to businesses. It's almost like starting the journey. I imagine it, and as you said, it takes sort of a piecemeal approach to AI use, you know. Pick one thing, it's solving a problem, implement that, where to start?

Diane:

Yeah, so one of the big themes that we always like to recommend is this concept of crawl, walk, run. I think sometimes some brands get really overzealous and they're like, we're going to buy a solution that does everything, but nothing works until you're fully integrated. And so you blink and then nine months have gone by and you're like, I haven't launched my first experience. So you're absolutely right in the sense, like you should figure out what is a main problem point for yourself. One of the educational pieces that we work with our partners on is really thinking about the funnel, right? 

You're working so hard and paying for all these kinds of ways for traffic to get to your site. But then they're kind of reaching a dead end. They're not moving forward. They're not doing the and engaging the way that you would like them to, whether it's a checkout or some sort of like, enhancement of their experience. And so how do you think about that and saying, where is the actual problem? Is it getting them to a second page visit? And that's simple and straightforward? Or is the fact that they're not fully converting and so taking those approaches to then say, what is the right kind of location? 

We always say an experience is a mix of three things. It's where on your site and who it's for. It's the capabilities that's involved and last but not least, but what is the actual result you want out of it? We always say just because you get more clicks doesn't mean you get more checkouts and more revenue. So what is the actual importance and the primary metric of a test? And then given that, then you can find a good solution that moves quickly and is agile and is able to kind of give you the speed that you're looking for to get those early wins that you can then either scale or enhance or sometimes that'll test when. Then you should make that decision quick and move on.

Amie:

Yeah, it sounds, you've put this out, you've laid it out so logically for me. You're a software engineer, aren't you? 

Diane:

That is my background. 

Amie:

Yes, I love the logic of it. Okay, so are there any kind of, when we say pick something, choose something, a problem you need to solve, are there any kind of advancements or are there any kind of AI use cases right now that you're seeing people kind of pick and test and win? 

Diane:

Yeah, I would say there's three that really come to mind. The first one is maybe something that's a little bit more obvious, which is things around recommendations, right? So a lot of the solutions today either will be focused on content personalisation or they'll be really focused on like products. So the carousels that you see on product detail pages and home pages. But between those, there's so many ways that you can really think around which is actually driving incremental revenue. And so most people will start there because they typically something that's built out of the box from their commerce solution that they want to be able to see is it better to bring in some sort of algorithm that makes sense that doesn't require us like custom-build data science models. 

So that's usually an easy lift to get started. The second part is something that's usually a net new that people want to do but haven't had the capabilities to do and this typically falls between two things. One is this concept around dynamic categories. You can imagine when you go to a home page they typically will tell you shop women's or shop men's or shop some sort of major category, that's typically curated. 

But the truth is if you went to the navigation of a website, you'll find that they have a ton of subcategories. So why are those not now dynamically selected for you? So that's what we would call a net new one where people hear this and they go, yes, why are we not doing this and making our visual browsing experience better? And it just seems like an easy win, right? 

Then the third one that we typically see is this concept around are my initial first pages actually effective. So most of the Google and like paid media traffic typically will land on like a product detail page or a home page. And so how do you take that into account to now dynamically change the page and test out different things that really tailor to that first kind of visit based off the time of day, day a week and other factors that could be leveraged. Now that's one of the third use cases where people go we definitely want to do this but for us to go and wait for our engineering cycles it's gonna be three months before they can build us maybe like three variations of our homepage that is then AI powered and you know dynamically selected and reordered and so it becomes a big endeavour, where all of a sudden you're like I don't have the bandwidth to try this so let's put it on the back burner. 

Amie:

And I suppose then it goes back to your sort of test and learn in a sense because you don't have it sounds like you don't need to you know bite off more than you can chew you can start small and work out what's going to be the biggest win and then go from there. 

Diane:

Correct. And AI has really gone a long way. I mean, we started the company a few years back and when we first came out into the market, I remember we came across two very distinct groups of marketers. One group went, yes, this is going to be amazing, I'm into AI. I want to try things, makes automation easier and my team more effective. Then you hit the other side where they're very concerned about the fact that like, maybe they spend all week curating content and now they're kind of a little bit fearful that the AI is going to take over that. 

And so we have to do a lot of educational things around, well, we're not replacing that role, but actually now you get to be the strategist versus a tactical person sitting there for 40 hours selecting the right SKUs and the right images. But actually now you can say, I wonder when people come from, you know, search and organic search that lands on our site, if they're really interested on price because they're budget shoppers or if they happen to be individuals that really want to complete the outfit or maybe complete a bundle and that's what we want to go for. 

So now you as a marketing team can actually think around all these hypotheses and test them out, right? But not spend hours and hours executing. So overall, I would say that's really where the growth of the AI has gone is like, especially in the last few months, like, people are much more open to the adoption of it. And so we're seeing a lot more of these like lower hanging fruit like experiences, especially the three I mentioned as good starting points. 

And then afterwards,being able to take that same AI strategy, whether it's selecting content, selecting people, selecting the time, and applying that to your other channel so you have a holistic approach, almost like the same AI brain that goes through all of your existing solutions. And that's what people should be really looking forward to do within their first year of when they're thinking about the AI roadmap. 

Amie:

Yeah, it's the goal. Yeah, okay. So when we're looking at, and you're looking at brands that are using AI well, who are you thinking of? 

Diane:

Yeah, so for us there's probably a few. Some of them do build in-house, but I think for the majority of retailers and consumer brands out there, to find good data science talent is probably difficult. I always like to tell everyone that if you were a data scientist and you had a chance to work at a really bright and shiny AI company like Google, or you get to maybe go and work at like a dog food or a cat litter brand, right, which one would you choose? Not that one's worse than the other, but you do end up realising that talent goes more in one place than the other. 

And so how do we then supplement that? But there's a few groups that have been doing it really, really well. And for us, when we think about the AI personalisation, Wayfair is a good example. They've built a really strong ML team. And that's wonderful. They're able to actually test and learn a lot of their algorithms on their own. But if you're one of the 90% of brands that can't afford to have your own in-house data science team, that's usually where we see the struggle, right? How do you orchestrate and map the data schemas the right way? How do you have some sort of capability to put infrastructure into place for guardrails? 

Because at the end, businesses are messy. You can't have all AI or all human. So how do you take that into consideration and find some tool that gives you a spectrum that can say, I still want the AI strategy, but out of these 20 things I have to enforce, you know? And versus having to now call like 20 APIs and trying to fuse them together. 

So Wayfair is a great example, IKEA is another great example, and Target. But all three of those, I would say, have done a good job of building the infrastructure from the inside out. And they've been afforded the luxury to be able to do that. 

Amie:

Which not a lot of brands have. 

Diane:

Exactly. I would say most of the brands that we come across, they are eager to build something in-house until they try it. And they realise a year passes, and just because you might have good data scientists doesn't mean you have good applications of data science.

And so you can imagine you may have a strong AI engine that you've kind of put in place for promotions or something like that. But then to actually now put it onto the site with good turnaround time so you're not waiting like seconds or minutes before some results come back. All these things are holistic. It's not like being good at one thing will win you the end prize. It's actually a mix. You need to have the right business strategy of like, what you want to achieve. You need to have the right technical team that knows how important data. You need to have the right, data science and algorithms that power it. So all these things come into place to actually do something that's amazing for the end consumer. 

Amie:

Which makes a lot of sense. Now, what's the cost of falling behind in this space?

Diane:

I think this is a given answer, right? If you're behind, then you can just wait in two or three years, your business just won't be thriving. And given the fact that revenue right now is gold for every single brand that's out there, I think people are aware that they need to jump onto this bandwagon. But the strategy piece is still a big missing part. And so that's one of the big things that we always come across when we're meeting new brands, is they don't know where to get started. They feel like they need to get this whole big kind of concept in place, whether it's a CDP and marketing clouds and all these things before they can make a decision. 

But the truth is the world's still moving, even while you're implementing other tools. And so how do you kind of find a happy medium where you can still leverage the power of AI to do good, but at the same time build up infrastructure in the background? 

Amie:

It makes sense. And looking forward, I mean, we're here now. What do you see as sort of the future for AI, generative AI, very keen to get your perspective on this because you're just so close to it. 

Diane:

Totally. I'll start with where I think practically the space is going to go and where I think a lot of hope is for generative AI, but I think there's some restrictions that will kind of place and come into play first. So what we're seeing as a trend in the next year or so is this concept around AI orchestration. I think especially these last few months, a lot of people have realised that all these tools have started saying, hey, we have AI in there, right?

But what happens is those systems only have visibility into their own silo of work. And so it's really hard to keep the experience consistent across and holistic when you have multiple channels, whether it's in store, online, email, app, SMS. You can imagine that all these different tools have its own split testing, have its own AI system, have its own configurations, and actually they all lack data from one another and then the ability to consume it and understand the context. 

So this concept of like one AI that kind of overlays across everything is something that we're seeing come to the forefront. So, you know, just a personal plug, but like Breinify is very well suited for this because that's what we were built for. Now, where do we see Geno.3 AI going? So I can tell you today, like I think people are very excited about it and it works especially in areas that have to do with chatbots or customer success. We haven't really seen it in the world of content yet and the reason being that legal teams are still a little bit sceptical that if you do the wrong prompt, then maybe the wrong product image or the wrong combination of product images and like content would get generated. 

So we haven't seen a big push on those fronts. I would say it helps with copywriting, but not to the point where they would allow the AI to just come up with it by itself and push it out into the world without any sort of approval process. 

Amie:

Yeah. 
Diane:

And so that's gonna be one of the big challenges, right, its the same as how we had started with this concept of temporal AI selecting content, there has to be some ability for brands to input their own kind of like rule sets and conditions into place but still allow the AI to thrive. And so that's where I'm seeing like a lot of people are very hopeful that Gervais is gonna save a lot of like content building time and like asset creation and it will but in a different way where it's not gonna be as autonomous. But we'll see maybe over the next few months or a year, Legal will find ways to cover you know the disclaimers around it and maybe things will be a little bit different. 

But I would say right now, when we talk to our partners, the biggest fear is like the wrong image asset is created and then who is liable. So we haven't seen the adoption of that piece in content creation as heavily as we see it in the customer success piece around how do we create good chat bots that can help to solve a problem before it has to go to a person. 

Amie:

Diane, AI might just have been, I would say, maybe the second most used word at Shop Talk behind retail itself. What are people most commonly talking about or struggling with?

Diane:

Yeah, I would say the biggest theme that we're hearing across all, at least the major retails and consumer packaged goods brands is that AI power personalisation is really key for this upcoming year. And everyone has different flavours of how they're deciding if this is the right approach. Some are really focused on working with their existing technology and trying to activate it internally. Some are really focused on bringing in external help that can overlay across a lot of things. But at the core of it, comes down to how do we know AI is effective? 

And that's been the big struggle that people have been talking about on stage is that like all these different situations and solutions are suggesting that they're much more powerful, but then how do you actually tell that it's doing better than before? So having that ability to be very data-driven to understand whether or not it's actually doing better for your business from a revenue standpoint is going to be really critical. So you can't do personalisation and work on AI without an ability to easily test and learn how does it impact your lifetime value of your consumers. So that's been the major two themes that we've been picking up across all brands, whether they're a little bit more advanced in their journey or if they're just starting the crawl stage of their journey. 

Regardless, it sounds like the theme is everyone's just trying to figure out what is the right approach, and many of the solutions out there require you to kind of invest nine months of your time to figure this out. So the brands and the vendors that can really come in in like a month or less, I think will be clear winners for a lot of them. 

Amie:

And so I'm going to...guess that you can do that. 

Diane:

Yes, so most of our implementation are done in weeks, not months. And we're a big believer in a proof of value kind of way for people to actually try it because we're big believers that we can talk all day about how great AI is going to be wonderful for you and how it's going to create lift. But until you get to try it and you can actually experience it yourself, it's this intangible thing of like the software is going to be helpful.

And most of the enterprise solutions out there take a long time because they require you to implement into their data sets. Whereas we've discovered that being from my background of working at Apple and then Symantec, one of things we realised is that marketing teams have grand ideas that have to be done yesterday. Tech teams have sprint cycles that do require, you know, some rigid piece to it. And you can imagine those two things not essentially working well, it can clash a little bit. 

So that's one of the things that we've solved for. So although it's not necessarily predictive, we know that to be able to jump that hurdle and overcome the obstacle of using AI, you got to make the integration, the setup as easy as possible where it doesn't rely too much on the engineering teams for you to see value. 

Outro:

Thanks Diane for joining us on this episode of Retail Untangled. If you've enjoyed listening, feel free to like and subscribe to Retail Untangled on your favourite podcast platform. Thank you.