This is a recap of a webinar "Artificial Intelligence: Facilitating the Omni-channel Experience" held in partnership with Yellowfin BI.

‍

Across various industries, organizations face heightened challenges in the form of rising customer demand. 70% of customers expect organizations to provide tailored frictionless interactions powered by AI, that are transparent and fair (Capgemini).

In this session, learn how enterprise organizations use their endpoints to redesign their in-store customer journey, expand their dark warehouses and facilitate the omni-channel experience with vision AI — harnessing data responsibly to drive actionable insights, as well as utilizing synthetic data (not based on actual/identifiable persons) to train AI models.

Featuring:

• Stephen Borg - CEO & Co-founder, meldCX

Transcript:

0:00  
Hi, everybody. I'm Steven Borg. I'm the CEO of meldCX. Thank you for Yellowfin for inviting me to this session, we're going to take you through the meldCX journey and what we do around data AI, and how they intersect to the physical space. As you can see that the customer experience and expectations around experience is changing. And COVID driven that is people were shopping from home and using their mobiles more and more and become adept to information at the edge, and that type of expectation. So we're seeing there's expectation around personalization, there's, you know, people have been in lockdown, they want to get out there, they want it to be a great experience is that's what they see the difference being.

0:52  
And we're still seeing where customers do open our economies to open up that a lot of decisions are made in store, and how do we feel sellotaped better that better? That's what we're here to discuss. So one of the things we look at, we see this such a rich experience on mobile, and at home. How do we take all of that, and bring that into that physical world where there's multiple challenges, right? You don't have someone dedicated to an device, you have distractions, you have multiple people in the one place, it's not a controlled experience.

1:28  
So how do we get that to be something that's even richer than those two experiences? And how do we maintain that attention? One of the ways we do that is provide the right insights, provide retailers, the tools, they need to understand their customers to understand what drives behaviour, to understand the patterns of behaviour that are successful, and the things that aren't really working. And right now, when we talk to some very retail customers, they don't understand that they think they do. But when we drill down into it, they really do not understand their retail environments, as well as they understand their online web and mobile. So what do we do as meldCX, we really we teach autonomous devices to see analyse and make decisions seamlessly.

2:20  
That means that we put technology in or we integrate to existing technology. And we sit there seamlessly, and really provide that data. And sometimes we provide a method to intervene or create a cause and effect in real time, like you would do on a website, or a mobile device, but in a physical location. So when we drill down into it, our role is to unlock the potential of data at the at that physical location. And we work closely with Yellowfin to represent that in an easy, digestible way. Because one of one of the biggest feedback we have is the amount of data we can generate, how is this consumable? How's it put in a format where multiple departments can use it effectively. And that's where our collaboration for Yellowfin comes into it. What we've built is effectively an actionable platform. So it allows you to look at your physical location, and we operate in stores, we operate in hotels, we operate in government entities, how do you take that physical location and get the best out of it? And we have real really three phases, the person level analytics, what are people actually doing? What is the behaviour? And is that behaviour conducive to our alcohol?

3:51  
The store level analytics is the layout correct. The products that we want to promote, or our hero products are then placed in an area where maybe customers or walk past or engage with other products that aren't destination products. And then the phase three is we take all of that and make sure that data is paired with your omni-channel data to try and create a single vision of what customers do, how they behave, and how we can maximise every square foot space to become actionable.

4:27  
To make this happen, we have an ecosystem of partners. We work very closely with Google. We work closely with Microsoft. Yellowfin is is a fantastic partner they really have helped us visualise our content and allow customers to customise their content. Intel we work very closely with all of these edge devices need compute, and they work closely with us to make sure that the compute is the most cost optimise the possible. And then Cisco work with us very closely to enable existing networks or existing camera networks to operate our ecosystems. And then make sure that what we do is secure. And it can be seamlessly integrated into the retailers or the target environments, ecosystem. And then we work with quite a few enterprise customers. And you'll see, with some enterprise customers, we've done some firsts. But ultimately, we work with a lot of partners and system integrators.

5:40  
So this boots on the ground, no matter what country, and we can extend our capability forward. And some of those partners know retailers more intimately than we do, and our operations and how they work, our take. So we do get a lot of questions around what we do, and how we work with cameras. We made it a point when we started this journey, to do to have some guiding principles. And one is we don't use facial recognition, we we simply don't touch that space. And even when we're processing our feeds out, everything's blurred, we've gone to the extent of making sure even our data that we've trained content off or trained behaviour is also not real data. So we synthetically created it. So our system does not have the ability to understand things like race, or things that might be sensitive.

6:39  
So we're really started from that place of having no human data in in how we approach. And it allowed us to actually provide more data, because it's anonymous, it's private. And I'll show you some of the ways we provide data. This is some of the things we do. So every every contact, we see, we create a unique number. It's not identifiable. So we don't know who that is. But we we do create a persona, when the session if they're happy, what their retention time is on a particular shelf, or it could be on a particular piece of content on screens, what are their average visit times spends, and we do that in spending a few different ways. And we can even have a look at what type of clothing they might be wearing. So if we're doing, we're about to do some work with a very high street retailer. And in that case, they want to know if a competitive product such as Hermes is being carried, because then they will, they will be able to establish what type of customer and what type of approach.

7:52  
So they're the type of things we do. And we take that into a meaningful data set, where they retailers can really understand what their audience is doing. And not just basing the content on conversion and other data on conversion. Now I'm going to show you some case studies of how we've gone about things and how we trying to take that experience further. This is one of my favourite ones, because we went to Paris Design Award for it, the first time we applied. So it's it's a really good use case. And it's has a responsible approach to analysing this in branch data. So one of the first things we did was look at content effectiveness. And we find this often where we enter a customer environment. And the content is just not suitable for the area.

8:48  
So for example, the content could be too long for a dwell area. So any of the call to action or the key messaging that the creator of the content wanted you to see, you don't see. So there it has a massive issue in that your network is not being utilised. So what we do is we provide meaningful content about if someone's seen it, if they've watched it all the way through and what the engagement score is, and then we can live tailor content, or do some A B testing to get that best content appeal ratio, all while analysing data on that those type of audiences that respond to that content more effectively. And we even find that some content has a different impact based on time of day, based on audience and even in some cases temperature.

9:49  
For example, a Holiday Loan ad might work better when it's hot, and then we do branch level activation. So we look at zones in a branch that allow asked to convert. So what type of content do we need? Where do we lay things out? What type of staffing levels do we need? So this is all the things that we we look at when we're converting when we're looking at branch level activation. So we're not just working out who's gone to the counter, and what the the inquiries are, we're actually looking at the catchment area and how that's been converted. Why, and what audiences are working best with that content layout, and the products that are offered in that location, and zone engagement.

10:37  
So some branches are bigger. What we try and do is look at what is the purpose of that zone? Is it an ATM area? Is it a Discovery Zone where you've got multiple electronic brochures of different products where consumers can go and discover? Or is it an area that is designed for engagement? And consultative? Are these areas working? Is this branch format working? And are we are the customers behaving for appropriate to the intended purpose? So what we may find is you've created this zone of engagement, but customers actually don't want to stay there. They might like another area where they might sit down and engage.

11:22  
So these this is the type of content we understand we we look for. And we provide that live dashboard of what's occurring, what's working best. And we can compare multiple branches, multiple locations, or multiple zones to see how that's working. And does that actually create a lead? Or are they getting through their transaction quickly if it's a transactional type zone, and this is this is some of the dashboards that our work with Yellowfin allows us to achieve for our customers, we look at the recognition of, you know, what is the response to certain content? What are the activities that triggers? What type of content works in that branch? And why based on the the sort of Age of approximate age of people coming in? Is it a whole lot of young families? Is it, you know, sort of young single, sort of up and coming? How do we tailor that content? How do we turn that into the right?

12:32  
Or have the right product specialists that suit that audience in those locations? And how do we match all that to make sure that we get effective conversion, and really maximise that space. Some of this content also is used out facing. So we actually found in this case that any of the COVID content or any of the COVID advice got a very, very high score. And that's because it was the topic at the time. So they did use this technology for the purposes of, you know, informing, informing consumers about community messages and what we should be doing around COVID. Australia Post, so Australia Post is an interesting one, it's it has a lot of elements to it. And we're using data in various ways. So one of the objectives of Australia Post is to drive or make it easier for small business customers to effectively use postal services in either a self service fashion, or do the transactional related things quicker. So then it reduces lines, it reduces wait times, and effectively, that allows people to do what they want to do in post and have a convenient experience.

13:52  
They also want to inform about other products they have. So when we look at posts, we looked at in multiple ways. We looked at their they have a gadget zone, what type of gadget sell the most and what creates the most engagement. It's a small area within posts, but it's a very important area. And some posts only have one tiny gadget zone. So we need to get it right. So we analyse what gadgets engage people and what is more important to the audiences wherever, wherever possible area as well. So what we're doing is making sure that when customers scan a parcel, they effectively drop it off. And we can record those. And here's the proof of lodgement.

14:37  
So this was a real key one, that when a product is scanned for postal services, we ensure that they've put it into the chute. And this is really important because it allows less steps and allows customers not have to not have to wait at a counter. But we also are achieving the compliance that posters service needs, that that product that was scanned actually went through the chute. So this allows speed, convenience, self service, and compliance all at the same time. And this is another one we've been working on with post. This is a very interesting one is handwriting recognition and box detection.

15:22  
The purpose of this is to make sure that when you've dropped off a parcel in self service, that the address is correct. And if it's going overseas, that are asking the correct questions. Really, what this is about is making sure that your parcel gets to the other end. And you've corrected any misspellings or bad handwriting or anything that's going to confuse the system live. And in that manner. Now, this one was a massive undertaking. And what we achieved was quite interesting one of our approach to synthetic data right from the start, it means that we can ramp up, I guess, an artificial environment to create that data. So for example, in this case Post gave us 200 parcels. And we quickly realised, realised that we needed millions of parcels, which was physically impossible to do.

16:20  
So we use our synthetic engine to effectively get to create all these variances in parcel technology, parcel sizes, handwriting types. And we really got to 11 billion synthetic parcels in two weeks, which meant we delivered the project in 12 weeks, rather than two years. So this, this brings us back to the physical presence. So what we do, we essentially bring all that data back, allow you to have actionable insights, or effect how content is played on the screen. Or the it could be a message that you're getting, or could it be a prompt for assistance at the store. So what we do is really bring that all back and make sure that every square foot of space is effective as it can be, and that we understand how customers engage, interact and play in your environment. Again, we're really excited at this opportunity to present and we have a lot of content on our website. We have white papers from Intel. So please, by all means, jump on our website and take a look. We also have a YouTube channel that shows some of our videos and how we've created these environments with our customers. Thank you