Danny Tomsett, CEO, UneeQ joins us to talk through the rise of digital humans and branded avatars.
There’s a reason why the digital human market is forecast to reach $527 billion by 2030. And it’s not all about the metaverse. One of the companies leading the digital human charge is UneeQ.
UneeQ works with enterprise brands to provide super-realistic digital humans with integrated conversational AI to serve customers across the marketing journey.
Why focus on marketing use cases for enterprise? Well, there’s a funny reaction people have to digital humans in customer service today. We’ll explain all of this and more, in this interview with UneeQ CEO, Danny Tomsett, including:
- The perfect use cases for digital humans
- Why digital humans are becoming popular
- Technology breakthroughs that are enabling growth
- Customer/user behaviour when interacting with digital humans
- The importance of conversation design
- Why personality design is crucial
- The cost of a digital human project
- How to approach digital human strategy and delivery
- Ethical considerations for digital humans
- Much much more
In this episode
00:00 Introduction and VUX @ VOICE22. Find out more: https://www.voicesummit.ai/agenda
01:30 Welcome Danny Tomsett
04:20 Why focus on the enterprise use cases for digital humans
09:20 Where customers interact with a digital humans
12:55 What is and how to make Sophie, the NFT?
19:00 Animating digital humans with AI
21:53 Creating a personality
31:10 Decisions designing digital humans and ethics
35:44 Randomising choice
42:50 Potential for digital humans to raise expectations
48:00 Most complex part: NLP
51:29 Cost considerations for digital humans
56:30 The future of digital human adoption
Word for word
So Danny Welcome. Welcome to the podcast on a brand new platform.
Danny Tomsett 0:01
Thank you. It’s good on the guinea pig.
Yeah, it says our Yeah, seems that way all seems to be working now. So Kane Simms is live right now on LinkedIn apparently I’ve just got a notification coming through. So it’s live and kick in pleasure for you, mate. Thanks for joining us. It is a pleasure to have you.
Before we do get into the conversation a quick mention for those of you that are tuning in that in October, VUX world will be running its first live event conference at the voice summit in Arlington, Virginia right next to Washington, it’ll be on the conferences on the 10th of October through the third to the 13th. And our event will be on the 11th. It’s sponsored and brought to you by Kore AI, one of the leading enterprise conversational AI platforms. And we’ll be bringing all kinds of industry thought leaders and practitioners and clients enterprises basically who are using conversational AI in the enterprise. To improve customer experience, we’re going to be talking about the end to end story of contact centre automation. So there’ll be a lot more coming out about that in the show notes. I will include the link to the agenda so you can go have a look at the tickets, I would include it in here,
Danny. But as I was just explaining, this is the first time I’ve used this platform and how it gets links into the other end is not something I’ve not figured out yet but the event will be a great success and so I’m looking forward to that. But without further ado, though, let’s get into this conversation Danny over there in Texas looking very educated with the obligatory bookshelf behind you there. So why don’t you kick off by telling us a little bit about yourself and about UneeQ?
Danny Tomsett 1:48
Yeah, great. So look, you know, UneeQ is a digital human company. And for those that may not be familiar with the term digital humans, you’ll certainly be familiar with digital humans, because really, they are the CGI characters that you would have had plenty of exposure to through movies and film, gaming. But the digital humans that were really focused around our business are AI powered digital humans.
Danny Tomsett 2:16
These are Digital humans that are avatars that have autonomous animation, and can connect into conversational AI, or natural language understanding to really drive dynamic and interactive experiences. And we particularly focus on the enterprise market, because we saw that there are a number of really challenging problems in the enterprise market. That is all about taking something that’s quite complex. And using digital humans, you can make it more human and simplify it. And therefore creating better outcomes. Things like helping people through, say, mortgage application through taking medication correctly through to many, many other areas, in E commerce and other things. So yeah, this is what we do.
Danny Tomsett 3:05
It’s such an awesome privilege to work with this kind of technology, because we’re at the forefront of learning so much, not just about the technology, but actually how we as humans interact with machines in different ways. I really enjoy them.
Nice. I mean, for those who’ve tuned into the podcast, obviously, we tend to mention digital humans here and there. But it’s always in an in as part of something else, like you know, as one of those things that is coming up, you know, but maybe it’s not as prevalent as you might think it’s not as prevalent as a chatbot. For example, one of those things is definitely emerging. But when most people think of digital humans, I think it gets crowded or overshadowed by a lot of the talk of the kind of like gaming and Metaverse kind of marketing kind of use cases and things like that is something that’s becoming very hyped at the moment.
Whereas you’re focusing specifically on enterprises, which is where I actually think the main value of these things is, you know, to be really helpful to add a different layer of experience on top of your channels, wherever that might be. But I’m wondering whether you can kind of maybe shed shed a light on why if you want to build a digital human based company, why focus on the enterprise and not the metaverse for want of a better word or the shiny marketing kind of situations as we see quite a lot of
Danny Tomsett 4:34
Yeah, the unrealized market of enterprise has been very attractive, you know, as we worked out, you know, there were real problems that we could solve. Well, I think, you know, Metaverse in particular, this is going to become something that enterprises will also benefit from and utilise, and digital humans will be a key part of how they scale customer interaction in there.
Danny Tomsett 5:01
So I think gaming is very relevant to this technology has been and will continue to be for some time. And there will be some interesting convergence, I think with the metaverse and how we design experiences that create utility and interest for users, but also solve problems and become safe places where people might want to also interact with enterprise brands as well. Retail probably has the first leap on that. But you know, you could see that would extend into healthcare and other areas in future. I think digital humans, one of the things that I think it’s quite important to understand, particularly in the enterprise space is that, as you kind of said, like, there’s the Chatbot, the focus around conversational AI, that’s not foreign to enterprise.
Danny Tomsett 5:54
And in fact, you know, there are plenty of use cases where I would say, digital humans would not actually be beneficial to solving the problem, where you already have conversational AI, or a chatbot experience. I often talk about it in language like transactions and interactions. So when a transaction use cases exist, they tend to focus around very quick one off type requirements. So if I was to even go, what’s my bank balance, or things like that, where I’m trying to have, obviously a conversational interface to simplify the process, do a transfer, things like that using the sub financial services as the example here? You know, there’s no real need to have a digital human in that, you know, like, we’ve tested this plenty of times. And, you know, contrast and interaction tends to be where someone has a series of questions, or there’s the customer journey, where the goal isn’t always efficient, even though that’s always important.
Danny Tomsett 7:03
The goal often is confidence, trust, trust, that you can do something yourself through self service channels, confidence that you understand what you need to know, to make a good decision as well. And that could be anything from buying a new product.
Danny Tomsett 7:18
We work with clients like Deutsche Telekom, and help people, you know, decide on what is the right plan, how many Wi Fi access points do you need in your house, all those kinds of things. So you’re on a customer journey, to solve an ultimate problem of having the right internet connection, the right performance and your house. And so for some people, that’s really easy.
Danny Tomsett 7:39
And for others, that’s a big scary technical problem to solve. And having that more human conversational contact point, and digital humans being visual, can really elevate the experience. And so that’s where I really see enterprise thriving now is that, you know, a lot of customers are starting to understand that better. And I think the technology is at a point where even conversational AI and natural language is getting to a point now where you can design conversation to be more interactive, and not just transactional.
So where would be an example? I’ve got so many questions coming down. That’s why I’ve been spending a bit of time writing some notes, because as you were talking, I had so many questions come into my mind. So where would somebody see a digital human? So for example, we had, we had Robert Cunningham from LNER, London, North Eastern Railway were on the podcast a few weeks back, and he was talking about a proof of concept that they did in Newcastle train station, where they had a digital human wasn’t like a super realistic one, like UneeQ do like you do. But it was more like a cartoon style avatar. And it was there for like we’re finding in the train station and stuff like that, and genuine help around the station, train timetables, that kind of stuff, they found that it went really well. I’ll come on to maybe some of the challenges that they had.
I would love to get your thoughts on that shortly. But that’s one example of where you may see a digital human for argument’s sake is in a physical location or retail environment? Where are some of the examples of where either Deutsche Telekom or some of your customers have been deployed in these where would a customer interact with one of these things?
Danny Tomsett 9:30
For our customers there are probably two or three key channels. Physical location is really good because it’s hard for self service and automation to exist in a physical environment without it being highly visual, and really simple to interact with. So a digital human is a great fit for that because it now creates this visual representation of conversational AI, which is great. The other areas You see now even with COVID, as well, changing a lot of our behaviours and how we interact with brands and companies.
Danny Tomsett 10:06
So you’ll see a lot of our channels now are focused around websites online. So it’s, you know, ecommerce sites or online shopping. Or sometimes these are just customer websites that are helping people on board, complete forms, or these types of experiences that are online and that obviously transitions both between desktops, laptops, and mobile devices. And then the third channel, which is more new and emerging that this is where, you know, more immersive channels. So this is your VR AR type experience.
Danny Tomsett 10:42
And this is, again with the focus around Metaverse, and with hitting this is really trying to use more creative power to enhance the experience and to leverage the fact that it doesn’t have to be 2D. So can we have better representation of products, and then a digital human. So give me an example. I can’t name the brand. But there is an automotive company we’re working with. And you can actually see the car, learn about the car. And you’ve also got a digital human answering questions about the cat and a whole virtual showroom. So some pretty cool things like that.
Well, that’s good. And it’s as I am. Sophie, is that the was it meet Sophia? I am Sophie, the one you you’re working on recently?
Danny Tomsett 11:31
Yeah so we did, we partnered with an agency to launch the first kind of digital human focused NFT project. So I am sophie.io is a digital human, with full natural language interaction built into her so you can actually have a conversation and learn about her and learn about her NFT project.
Danny Tomsett 11:57
What’s really cool is we’re also plugging her into some really smart technology, using AI to generate art, based around obviously, a description that NFT holders can give her she’ll generate the art and then that will also be released to the future NFT as well. So it’s good, good for our company to learn a lot through this process as well, we want to obviously understand the process ourselves. Because a lot of customers are asking us like, what should they be doing around the NFT strategy? And what does this look like? So we put ourselves out first, to really understand that and it has taught us a lot and, ya know, we were able to help customers through the lessons that we learned through that process.
And one of the things that is not necessarily related to NFT’s but I’m showing, I’m showing the screen now, for those of you that are listening on the podcast, you won’t be able to see this. But Sophie, as a digital human, is super realistic. You know, I mean, you can tell when you look closely enough, you could probably tell that it’s computer generated, but just glancing at it. You know, if this was a small screen, and you just flick through it, if you were not, you know, really staring at it. It is really incredible how kinda high quality it is.
And so I’m wondering about sort of like, you know, I mentioned the LNER example, there was kind of, it was like more of a cartoon style avatar, and Amelia, popular conversational AI platform, they’ve had this kind of like cartoon style kind of Avatar, not cartoon, but it’s very clearly animated. Whereas, Sophie, Sophie, and I’m sure a bunch of others that you have, and you know, other companies Sapient X, I’m sure you’re aware of like a couple of companies like UneeQ have got to a point now where these digital avatars are just so realistic. It’s unbelievable. I’m wondering whether you can talk about, like maybe it whether it’s the technology advancement, like what is what is it that enables you to create something that is so human like,
Danny Tomsett 14:01
Yeah, the technology really has advanced quite a lot because of companies like Unity or Epic Games. So triple A game platforms focused around, obviously, the rendering platforms that typically game companies have used, but now can be extended to a range of other areas. And we’ve taken advantage of our partnership with Epic Games. So we use the Unreal Engine, you know, for a long time because we’ve been doing this a little longer than others. You know, we had to develop a lot of our own proprietary shaders and textures and other things that really enhance the look with the lighting and things like that. That’s become a lot the barriers to entry are on there have become a lot easier with lighter versions.
Danny Tomsett 14:50
So that’s the first part of it, though, that’s really just in the modelling. So the textures, lighting, shaders, and having really good quality Yeah, I just didn’t know how to build out, you know, hyperreal, or very real, digital humans and how that works. And that’s, that’s a known art. We’ve been doing this for many years. And gaming and film studios in particular, are very good at this.
Danny Tomsett 15:15
And, and so the second part to that, though, is much more complex, which is actually how it animates and how you, you effectively create autonomous behaviour, and personality. So the second part is where pretty much most companies fall short. And this is a problem that really requires heavy degree of science and art combined, to really design AI models and to build an architecture that the digital human can speak naturally, can respond contextually to what’s happening, you can see me even moving quite a bit when I speak, which is very normal, you’re nodding, when you’re listening, there’s all these kind of dynamics that when you break this down becomes millions and millions of variances, and how compensation can flow.
Danny Tomsett 16:09
And then how you emotionally respond appropriately. And how you express emotion. I’m really sorry, but we can’t approve that loan right now. Or Congratulations, you’ve been approved, eyebrows go up all these different things that take place. And you don’t want to build in, you know, a considerable amount of overhead and natural language design, so that you’re having to tell the character how to move, how to be, how to respond and become insanely hard and really complicated as an operating model. And the future. as you’re making changes, what you really need is the AI to understand that if the natural language, you know, processes that it understands the intent, and here’s the response, then the animation systems really need to be able to understand the context and speak that in an appropriate way.
Danny Tomsett 17:03
And that’s really the technology that we’ve focused on for a number of years knowing that this is going to be the game changer. And how people experience digital humans in a more natural and realistic way. Funnily enough, we’ve played this into non hyperreal characters as well. So we’ve had animal characters we’ve had, you know, put it into Disney characters. There’s all sorts of things that our animation system can drive, because it’s so contextual, and how you behave, how you move with that. But those two combinations are really how this level of quality and this experience is now in a place where I think it’s moved beyond Uncanny Valley into more natural interactive experiences, that most people, not everyone, but most people experience in a positive way.
Hmm, interesting. That is very interesting because I always thought that. So I’m going to make a note of that, because I’m going to come back to that. Not everyone experiences the positive way, it’s interesting because I always thought that this would be something that most people would just love.
But anyway, I’ll come back to that. The interesting part is that whenever I’ve interviewed people who have are creating something that is natural language based, but the result is not just a conversation, like Alexa, for example, is purely compensation because a chatbot, a voice assistant voice bot in the call centre, are purely conversation based, soon as you put a physical thing there, whether it’s an avatar, digital human, maybe it’s a robot, like a house robot, you’ve got that other stuff to deal with, which which you’ve just mentioned the animation side of things, you know.
We had Stefan Sherer on just before Christmas, and he was from Moxie. And Moxie, the little robot is fantastic. You know it, but when you design for it, they have to think about every single animation that that robot does. So if a kid says something to the robot, and the robot is surprised, it has to do that kind of thing like eyebrow raise and stuff. So when you talk there, am I correct in understanding that all of that animation stuff is all just part of what already exists. And if you were to build a conversation on Sophie, for example, as we’ve just seen, that Sophie can understand the natural that what she’s seen, basically based on understanding the natural language and therefore, animations take care of themselves.
Danny Tomsett 19:26
Correct. So Sophie, a little bit of insight is plugged into GPT 3. So if anyone goes to that I am Sophie.io, so you can talk to her. And we use a combination of an IBM Watson combined with GPT 3 and a few other things that we’re plugging into API services. So she knows all sorts of other information but GPT 3, like there’s no way you can really mark that up. You could do a sentiment analysis on the fly and try to do something clever, but it’s dynamic. The responses are dynamic every time but you’ll see that her motion to change in the way she speaks it and expresses it is really, really good. So she’ll have a cheekiness and so we have what you call a personality, kind of waiting that takes place through that animation system. So there’s a personality behind who she is and how she acts. And if you know anything about GPT 3, you can also design the personality and the way that you’re building the model and training the data around the model. So that combination of the two works really, really well.
Interesting. So the personality side of things, it’s a very hotly debated topic in conversational AI can’t spit something, everyone kind of agrees that you need to have some, some degree of personality. But others differ on what it should be in terms of what you need to actually capture, like some models will just do, you base it around the goals of the user based around the tone of the conversation, others will go more into like, this is Marc is 28 years old, from South California. And it’s like, his dad’s a doctor and his mom’s a vet.
And so for this, though, it is definitely different. Because you have a physical body, you have a physical person, you have to create something to look at. And people’s personality, whether it’s just a human brain that makes it so people’s personality is often linked, I think, to how they look in some way, shape, or form. If you look at Tommy, you know, Tommy Cooper, I don’t know if everyone knows Tommy Cooper the old, British stand up comedian always wears a red Fez, is about six foot five, is massive. And he just looks daft, you know, and his personality is exactly like that, when he’s on stage. His looks fit his personality perfectly.
And whether that’s just us making up that story, because we’re humans, and we like to find patterns and stuff like that, or not. But I do think that people’s personality is linked to how they look. So I wonder if you can talk us through some of the considerations that you have, when you’re creating a digital human from scratch for a client? How do you go about personality design? And how does that link into how it looks? If it does at all?
Danny Tomsett 22:07
I think one of the things that we misunderstood at the beginning was that we would design based on the user in mind, and the customers would go through a whole process of what does the look need to look like? And who’s the target audience? And how do they appeal to the target audience and all these different processes. So we focus so much around the, I guess, the external look of the digital human. And then as we observed how people interact with the digital humans, we quickly learned that they’re trying to actually know more about the digital human, the personality became even more key.
Danny Tomsett 22:49
So the tone of voice, how they answered questions, and responded all the way through to people wanting to just know, general chitchat, not everyone does. So you’ve always got to think about the fact that you’ve still got a very broad range of user interaction. But there was a very high percentage of people just trying to get to know more about the digital human, what do you like doing, you know, all these kind of interesting questions that we sometimes probably catch ourselves doing, if we’re talking to someone else, even a customer service rep. And like, how’s your day, we might just throw in that. And it’s very fascinating.
Danny Tomsett 23:26
Actually, even the demographic splits. And even older demographics, actually very relational. So they will ask more conversational questions and be more relational than that. I think that tends to be a demographic that prefers conversation, sometimes over some of these digital clicks and self service tools and a lot of feedback. As websites keep changing, I learned one way now: they’ve changed that again. Whereas if we can just talk, it just helps. I’m sure you’ve heard that before as well. And the digital human, I think it’s just that amplifies the more human characteristics of how we respond. So then is personality.
Danny Tomsett 24:07
But that’s really key. I think personality design is something that we’ve in, in marketing, as an example, you often are thinking about your end user personas. And so you create these personas, thinking about, you know, what are the typical personas that we’re going to encounter on this customer journey? And how are we going to design for that? I think it’s really a counter persona. So you’re thinking about it in the same way, what are the personas that really matter? And we think about the design and you know, are they going to be highly empathetic, you know, it could be a healthcare coach, a wellness coach, and so, you know, a little bit of humans is probably good. But, you know, I’ve really been hypersensitive to where empathy will play a key role. Really sorry, going through that right now. You know, that is tough, you know.
Danny Tomsett 24:57
So those are the types of things that we want. Higher levels of empathy come in through the language and the expression and so we build that into the, into the personality.
Interesting. And is that the first one of the first things you would do? And I wonder if you can walk us through what a typical project is? I mean, first of all, I’m assuming that you at UneeQ do a lot of the hands on designing and implementation of these digital humans. Whereas with enterprise conversation AI, it’s a mixed bag. Some companies just do it themselves, fire up dialogflow or Watson and crack on whereas you’re dealing with very particular technology. Very kinda like I’m assuming that all your NLU technologies and NLP technologies need to integrate somehow with what you’re doing. So I’m making the assumption that you do, UneeQ does the design and build and implementation is that correct?
Danny Tomsett 25:49
Yes, not so much always around that you know, us though, because a lot of enterprise clients, as you said, already have teams work on their own NLU. So what we try to do is, obviously, complement what is already there, often what exists, and most customers and in our view, we like to call it a skill set.
Danny Tomsett 26:07
So it often doesn’t have personality, and the NLU and often isn’t, isn’t designed in a way that feels more human to interact with. So we spend a lot of time with our clients actually thinking about the use case with a more human feeling interaction. So where do we bring personality? Where do we bring more empathy? You know, we don’t have to talk like a robot, we’re actually creating more human experience here.
Danny Tomsett 26:31
And so you know, and to be fair, when you’re building chatbots, or other channels, you actually don’t need to do that as much as when you’re creating a digital human. But when you do it this way, like we outperform, on all metrics, by a long, long way, like conversion is nearly 2x on just conversion alone, for working with customers, like say, an E commerce or product recommendation, things like that.
Danny Tomsett 26:57
So it’s very, very good when you get it right. So the process for us through the design, most of the time, you know, we’re focused around sales marketing areas. And the reason we do that is because we find that introducing a digital human, going back to your way, you’re going to ask this question, why doesn’t everyone like a digital human? Well, some parts of it are just polarising, it’s something new. And not everyone likes new things, right? It’s like this is different. So you are gonna get a percentage of people that just feel a little bit uncomfortable and interested enough those who continue to use it become very comfortable. And then it becomes actually a more preferred channel of experience. But you do have something that is polarising with digital humans right now, because it’s not around us every day.
Danny Tomsett 27:53
But the other thing, I think, with the design process with a digital human is, is really just thinking about, you know, the way that when we create the experiences, what is the mindset of a user when they’re coming in here, because if it is customer service, and my phone has been cut off, my bill is wrong for the last three months, you know, tada, here’s a digital human, to make you feel better, because I’m gonna post the thing on the face, if it gets one natural language thing wrong, right? Can you help me with this? I’m sorry, I don’t understand. Could you please try again. That’s it, you’re gone. You are dead to me, digital human.
Danny Tomsett 28:45
So we like sales and marketing, because it has a much better way of creating a positive atmosphere as the way you kind of introduce a new experience. It’s like the honeymoon of buying a brand new iPhone, right? Like, hey, gotta get a new phone and digital human thing. Let’s have a look. This is really great. So you do want to introduce a new channel and a new experience and far more positive way. The reality is a digital human can help with a lot of customer service and interactions, not transactions. But if people have already met the digital human, formed a positive connection to the experience, and then go into customer service, it has a much bigger impact. Because there’s a lot more trust , I guess, connection with the personality, and therefore I feel like they’re their representative to help them because I’ve helped them before they’ll help them again. And so that’s how we work with clients on that.
And then, another difficult part I mean, creating a personality that is reflective of the brand, the tone, the use case, the way the thing looks and stuff like that is all hard enough. But you also have to actually design the thing itself. And everyone has an opinion on everything, even if you go back to the days of building websites, you know, everyone has an opinion on what colour the button should be, or what image should be used here, or what language we use and how this should be phrased in a chatbot. Everyone’s got an opinion on what this piece of dialogue should say, and what it shouldn’t say, how it should do or shouldn’t do.
When you create a physical thing. Not only do you enter into the world of opinion, because everyone’s got opinions on everything, but also you’re in pretty steep territory of either prejudice or ethics. Because with conversational AI, if it’s a voice bot, really the only decision you need to make is should the voice be male or female. And as long as you do some good consideration around that I’ll give an option, you know, do what male or female to the user, or long as you consider properly the voice that you’re using, and why then fair enough.
Whereas you’re designing gender, you’re designing race, you’re designing age, it is adding a whole bunch of different things that have ethical impacts from your client perspective, you know, and so I’m wondering whether you can talk us through how those kinds of decisions get made? Is that something that UneeQ does yourselves? Is this something that you work with a client to define? Do you do a number of different ones so that if I go to the Deutsche Telekom website, and one minute, it might be one type of digital human, they’ll go back again, the next time? It’ll be something different? What are we going to talk about about how these sort of decisions are made? Because it is, it’s a minefield enough with voice AI, but you add all of these other layers on top, and you’ve got a whole lot more stuff to consider?
Danny Tomsett 31:40
Yeah, it’s a really good question. I, you know, where we are in the market, I think we’re fortunate in some senses that it’s very early. So it’s seen as more experimental, as organisations are looking to learn and adapt. So they know they have to make some decisions around this. I feel like all reality with us is that choice will be given back to the user. So the user will just choose who they want to have as their advocate representative. In any organisation, whether it’s in the metaverse as a digital human, whether it’s online channels, things like that, I don’t really subscribe to the whole, I’ll make some stereotypical assumption on what you want, and I’ll change immediately. So I think that’s dangerous territory.
Danny Tomsett 32:34
In all reality, the cool thing is that you can take a knowledge base of helpful information, and that can easily be transformed through any different look, style and personality. And so you can create multiple look styles and personalities. The reality is to date, we don’t do a lot of that with clients, because they’ve really focused on what can we do have one, and really validate and prove this out. And so we’re at that early stage.
Danny Tomsett 33:02
I think as we move more into the from early adoption into more mass adoption, I think we will see more considerations for variety and it is so easy to create for variety, that is the reality you know, this voice already synthetic voices, we could stand up 16 Different within like moments, and have that so we can we can do something similar digital humans as well, that when customers so do we really consult our customers decide they know, their they know their customers the best. And they also understand their brand, their brand guidelines, and we try and work with them around that to really try to create something.
Danny Tomsett 33:38
Demographic focus is a good one. Because we you know, we also identify what is the language style? Do we address you by your surname? Or do we have a more casual approach to things and just by your first name, so things like that, I think, you know, play into the audience and what we’re designed for.
Danny Tomsett 33:59
And then and then when it comes to race and fix, often, you know, surveys are used, you know that there’s quite a lot of surveying and a lot of information that already exists, I would suggest that a lot of that information just to because particularly when we deal with marketing departments, they’ve already been looking at demographics audiences, they’ve been testing, and they’re doing this because they’re casting people for advertisement. So they already had to have actors there. They’re already thinking about what are the right actors, the representation of our brand. So a lot of this information kind of comes into the process of what we’re looking to do. But most people you know, even under our testing, most people aren’t super hung up on Hey, you use the wrong nationality for the service. What’s going on?
Danny Tomsett 34:49
I think in most cases, people are more interested in solving their problems than trying to get you to do everything the way that they need that to look and feel effort. that friendly personality and the other day like, think about real life, we’ve come across someone, and that friendly and helpful, and just there to help us with what we’re doing. We don’t tend to say it’s a shame that you’re, you know, you’re white, that’s a problem. You know, like, that’s, that’s not what we do, we can do that fairly helpful. How good was that? Right? Just be good human.
Danny Tomsett 35:20
So I think in time, as it becomes more mass adopted, though, people will start seeing trends and trends will start getting questioned, why are all the digital humans female? Right would be a trend that might come through? And so you know, and those types of things, I think, a great compensation, they’ve really good things to be thinking about and talking about, and certainly solvable.
Yeah, the voice assistants went through that, you know, why are all the voice assistants, why did they all have a female voice kind of thing. And then, which is weird, because Apple, I think its male in America, female in the UK, you change it now. And then Google had like, coloured coloured voices like blue, yellow, green to try and separate self with agenda and stuff.
But it’s interesting, because as soon as you actually put a physical person there, it’s so much more to consider, ideally, ideally, the end result is that anytime you call a company, it just randomly selects a voice. And every time you log on to the website, and there’s a digital human there, it just randomly generates one, or at least one of the 10 or 20, that you have. and then all of that decision making is taken out of it. But it’s just, it’s just kind of getting to that point, I suppose they’re not not every client is going to have the budget for you to create 10 different variations of, of digital humans,
Danny Tomsett 36:40
Well, it’s actually getting much cheaper to do it as well. So it is very possible on that side. And we had one customer, we were a gender neutral, digital human. So it was really well put together and everything but in all reality. I mean, I’ve seen that even when people try to create a chatbot, that’s just called bot, but people still create a gender attachment to it. And the same thing happened with the gender neutral that everyone started saying she, and then it didn’t really matter how much effort we put in, like, I think you’re gonna have challenges where you can’t always control how people are receiving things. So it’s better to go with choice,than any other strategy, if that’s going to be a problem to solve.
Yeah. And then you mentioned that you’re not gonna please everybody. And you kind of touched on this a little bit around some people’s negative kind of reactions to it. Like, what typically are the negative reactions to digital humans that you’ve noticed?
Danny Tomsett 37:40
Well, there’s probably just two areas that I’ve observed. And the first one is very valid, which is the polarising effect. And that is just, you know, for me, and probably a lot of people that are similar where they might have, they’ve been brought up around technology, they like technology, they’ve been gaming, they love movies, they like sci fi, like this is just easy technology, right? That’s right. And that kind of that was cool. I really like it. And I just, it’s so amazing how good it feels to talk to like, new things and change, it doesn’t bother my persona, that much. And so you’ll probably get up to about 52%. And then there’s not all just one persona, but about 52% of people are in a very good category with this, this is preferred over other things.
Danny Tomsett 38:31
And then you start splitting into next two categories, which is like this is a little bit unsure. And I’m willing to try. But this is not exciting to me. Now, this is just does this help me? And am I really going to be interested in using this again and again, and get to know that. And we’re really successful with that group as well. Because if you’re an uncanny valley, you turn that group off, you’d lose them, because it’s just a little too weird and a little bit unhelpful. But once you’re designed your natural language, understanding it to be supportive and helpful, there’s a personality and it feels natural to interact with you when that group is over.
Danny Tomsett 39:13
Then the last group is just very freaked out by a digital Human Right. Like it’s kind of like, I don’t understand why I would talk to something that feels so human. That’s not and that’s totally okay, as well. It’s actually I think, I thought it would be more going into this business. I thought there’d be more than what there would be in there. So I was so surprised with how interested and how many people are wanting to adopt this and how many people were excited.
Danny Tomsett 39:45
So that last group, I think, is something that will always exist and it’s normal to exist because this is not a normal I’m sure many technologies and past have always had a group that didn’t even think like why would I want to have a phone with me all the time. That’s just ridiculous. That’s just creepy, then people will know where I am and call me, you know, like, these are the types of things that evolved through history and become a new norm and get more accepted. And so yeah, so that’s, that’s how that works.
Danny Tomsett 40:17
So the second part is a little bit more. Yeah, so it’s a little bit more demographic focused as well. So, you know, if I feel really comfortable with something, why do I need to talk to someone who tends to be in there. And how they interact with digital humans is very different.
Danny Tomsett 40:39
So a good example is, even so if you look at Gen Z, so familiar with using technology, but very command oriented, so they actually command and talk to a digital human and are more command focused. So if you’re talking to Siri, Alexa, things like that tend to be: I know what I’m doing, I’m really comfortable with this, this is cool, really comfortable, just do what I need to do those kinds of things.
Danny Tomsett 41:05
Whereas as I gave the example, before, older demographics, so 65 years and older, this is I never saw the technology becoming popular, and that demographic originally, but this is actually a technology, I think that really bridges the gap between the scary experience of digital and all the self service tools, that not really, they’re not very personal, they are changing frequently, all these types of things. And we can create something that just feels like having a chat with someone that you can actually get things done. And you can do it yourself online and not have to wait on the line or anything like that. And I think in that kind of language, you see a very different way of seeing technology, it’s to have a conversation.
Danny Tomsett 41:55
The hard thing with that is you’ve got to design the natural language understanding to have a conversation. And so it gets a little bit more tricky at that end. And if you don’t get that right, then yes, we see rejection of the technology, they’re the same in the lower. So the earlier Gen Z, Gen Z. If you don’t get a command base, just make it snappy and quick. And it’s talking too much and things like that, you’ll get disengagement there. So you really got to think about the demographics and how you’re designing. But I don’t think that’s too dissimilar to what you’re doing on invoice.
Interesting. Bless the old folks, they always get a bad rap don’t they, the old grey feathers, always get a bad rap for adopting technology, but they actually do, they give it a go, man, you know, Alexa has been fantastic for a lot of older people and stuff like that, you know?
So you touched on something there, which is quite interesting, which is around how Is there potential for the expectations around what digital humans are capable of, to be raised? When you put a digital human there, for example, if you call a call centre, people have been used to those terrible, clunky voice systems for a while, if you’re calling about credit, say credit, you know, if the chat bots that are just push buttons, you know, push 123. So people are being used to primitive voice and chat technologies.
But then when you have a natural language, one, which is you know, a little bit more allows you to speak a bit more verbose manner and stuff like that. The expectations rise, but not tremendously, really. So if you think about Alexa, when you talk to Alexa and Google Assistant, you’ve got an expectation, it’ll understand you. But I think now there’s an expectation that on understanding that it’s not going to help with everything, there’s certain things it does certain things it doesn’t.
Whereas you put a human in front of that, that looks so realistic, like a movie character sitting on screen. I would imagine the expectations get raised quite substantially, which puts a lot more pressure on the conversation design and NLP. I wonder if you can, if that’s what you have. Is that what you’ve noticed? Or is it can you just put a digital human or put a voice bot on its own?
Danny Tomsett 44:10
No, you can’t put it in front of a voice spot and it’s all rosy, although, so you have to take the good with the bad, right. So the good news is that a lot of companies actually struggle to get the adoption of some of their chatbots and their automation technology, because it’s so closely connected to the past of the horrific systems that we’ve all had to navigate. And so there’s little trust, and there’s not a lot of, I guess, ways that you can convince people unless they just try it one day, and then they realise that it’s better.
Danny Tomsett 44:46
And I think the key to digital humans has been successful as the improvements around the way that we understand what natural language can do well, but it can’t do well, the way that we design it. And obviously there’s some technology and advancements as well that have made this A lot better. So all these kinds of experiences could be better. But you still got this low engagement because people have been burned in the past.
Danny Tomsett 45:07
So put in a digital human on top, the good news is you’ll create all this new traffic, to use your new experience. Because people do believe that this is now more advanced, it will be a better experience. And away they go. If you don’t make it a better experience, than the expectations are that it was, and we will hurt the reputation of digital humans been the saviour of natural language, but but the reality is, you know, when we design with our customers, we understand that so we really think about how are we designing the use case being as specific as possible, and really guiding someone through as much as possible, as opposed to tell me your problem? Anything? Let’s go and you’re gonna disappoint them.
Danny Tomsett 45:55
So I think that’s the key, I think. Yeah, the other thing is that there is a little bit more, I think, I would say, grace or understanding, that sometimes can’t. So a digital human can do this and go, I’m so sorry. I still don’t understand. I’m trying. If I don’t get this one, I am going to connect you to someone. But could you try just one more way, right. So you can do something like that with facial expressions, everything. And it’s fascinating in some of our more intense studies that we’ve done to try to understand whether people have more empathy for digital humans and chatbots miles more empathy, miles more. So actually, you can get more forgiveness. If you design it. If you keep saying the same thing over and over again, you remind them that they’re annoying, but that hasn’t been programmed. Well. So I’m sorry, I don’t understand, please, again, I’m sorry, please don’t say again, is this. It’s so painful and bad when we see these experiences.
Danny Tomsett 46:57
So I started talking to the team about wanting to be able to turn off digital humans when people don’t design them. Well, because they heard you they’re hurting the ability to actually generate much better throughput and self service channels. If we design them well. And it’s a great opportunities, digital humans to do that. I think the metaverse will be another way that they will be introduced again. Which will be good.
Interesting. It sounds as though I don’t want to put words in your mouth. But where I took that last statement was that you’ve already built the ability to create digital humans, they look fantastic. You’ve mentioned that you’ve already got layers of software that do the animations and facial expressions and all that kind of stuff. So it sounds as though actually the hardest part or the bit the tics most of the work. Not the most important part, I wouldn’t say because I think they both go hand in hand. But the more complex part, from a day to day project perspective, sounds like a lot of hard works already been done on the the digital human side, would it be fair to say that the more complex part for you now that you’ve already got the digital human engine setup, the more complex part of these things is actually the natural language
Danny Tomsett 48:13
100%. Number one dependency for digital humans to be successful, is still natural language and the way that people design and build that, like our technology is so productized, that you can deploy a digital human now, on your website, within 30 seconds, you can actually have an app running, we’ve got modals, and all these pre built SDK, so you can actually have it sitting on top of your ecommerce site, ready to talk about your products. But you still need to design the natural language to be a really good experience around that.
Danny Tomsett 48:47
And that’s where we spend most of our time with clients really tuning the overall experience, there are some visual elements that are quite fun with digital humans that you don’t necessarily always do. So SDKs allow you to manipulate the website as you talk. So as you’re talking to a digital human, they’ll bring up a product and they can look at the product and say, this is actually what I think will be a really good fit for your skin type. You should really consider, you know, having a look at this for the next three months. Things like that are quite fun as you bring that into both the natural language and the orchestration layer that is pulling, you know, various API’s and data points to to fill out the conversation so you are really thinking more like a TV show versus a radio show, right? Like that’s, that’s what you got to think you’ve got a lot more visual elements you can bring in. It’s not just about the people. It’s about what’s happening around the people and what’s happening on the site. And that’s how you really start amplifying your metrics. That’s how you get better conversion rates. We have way better engagement and throughput completion rates or those types of things.
Interesting We spoke as mentioned before, we spoke to LNER’s Robert Cunningham. And they tried it in Newcastle train station, but don’t know what they use and what platform they use. But as I mentioned, it was more of an avatar kind of thing rather than a digital human. But anyway, the fact that there was a person there talking, their results were very positive. And the only thing that was preventing them from rolling out widely was the cost. Because I mean, if you think about it, you could have a digital human on every platform, and on every train station, you could have one on in the concourse, you know where the information stands, every shop potentially could have one on the train itself could have one. So the potential for it’s a huge help on hand, wherever you are.
But what Robert was saying is that it was cost prohibitive to do that. Not all because of the software. Because in order to set these up in a train station, you need a good internet connection, every single square inch of the building, you need screens that are good enough to have an engaging experience on you need microphone technology, there’s a whole bunch of stuff that you need, aside from the actual digital human itself.
But you kind of alluded earlier that the actual pricing of digital humans is not incredibly expensive. So I’m wondering whether someone, let’s say, there’s someone listening to this now who’s debating and doing a chatbot, or they’re thinking about voice enabled, and now call centre? Now they’ve heard about digital humans, they’ve got a bit of an idea about that. How do you think it’s I mean, I don’t have to go through the numbers, but how does it kind of stack up? From a business case perspective? If you wanted to do a project involving a digital human? Like, is it 10x of a typical kind of voice or Chatbot? Project? Is it double? Like? What ballpark ballpark range is what you’re kind of looking at? Yeah,
Danny Tomsett 51:48
I think, yeah, maybe just the interesting thing is, if I look at the competitive landscape around us, right now, you have a lot of digital human companies, that or avatar companies, I mean, it’s the same sort of game, really, I mean, they’re trying to create an interactive experience with a human like character, the, the nature of a lot of them, they’re trying to deploy complex tech into a PC server that has to have the GPU or the big screen, and they have to solve all these microbes, and all these things. And so the cost of a kiosk just to do that becomes very, like, very hard to scale. Not only just the cost, but managing that over and over again.
Danny Tomsett 52:36
So we solved this problem about five years ago, because we started this in 2015, when we started our r&d journey. And we recognised that that was never going to work. So a lot of what we do is we render everything on the cloud. So just like you’re connecting to a Google meet, or a zoom call whenever that everything can be super lightweight, and hardware side. So you don’t need to spend a lot of money on complex kiosks, you can stand it up, just having a video called straight from the you do need a good internet connection. And then what we did is we actually recognised microphones and noisy environments.
Danny Tomsett 53:17
So what you can actually do is you can just put a screen with a video and a QR code on it. And people can come up with their own phone, scan the QR code, because it’s in the cloud. And now they have a digital human on their phone, because who doesn’t have a phone and talk to that digital human while they have wayfinding on their phone and moving around all of a sudden just changing the game to our retail customers. We have public transport as well. And that there is super cheap, like with weed like literally talking like 20 cents a conversation of that. You know, so, so now and that’s, that’s including the digital human like, that’s the kind of thing.
Danny Tomsett 53:55
So yeah, because we work with enterprise customers, there’s some design, and there’s a lot of implementation thinking and things like that. So you know, they need to be thinking probably the better part of 100k to more depending how big that project gets. But at the end of the day, when you look at scale and consumption, this really does scale now, particularly if you use companies like UneeQ, who have solved it in this way.
And especially if you’re applying it for those enterprise use cases for big companies that are dealing with millions of conversations, and you if you’re demonstrating the ability to convert twice as good as a chatbot, as you mentioned earlier on, then, you know, whatever it is north of 100 k costs pay for themselves anyway because there’s so much value in being able to deliver so much better service.
Danny Tomsett 54:44
It’s very, very easy to build an ROI model. As long as you can do two things. You can either redirect expensive traffic to lower cost channels, or you can obviously help with conversion and completion that turns into revenue and digital humans have, you know, have a really good track record and doing this now. So they I think, you know, we’re probably going to move out of that early adoption, as people understand this more, I think mass adoption in the next probably five years is definitely very realistic with this technology.
Very good. When you say mass adoption, there’s different ways that you could kind of classify that like, for example, it was said that Google Assistant Amazon Alexa, those devices will be in every home by x date. There’s been stats out there, which is like every company, I’ll have a chat bot by x and a juniper actually put something out a few months back, which was kind of the exact numbers, but it was a staggering amount of spend on cognitive services forecast by 2023.
Lots of companies have chat bots, you know, there’s this notion of a million or billion assistants, which is every company having their own digital assistant, whether it’s chat or voice, whatever it might be. And so this whole market, which I would actually class digital humans as part of because the NLP technologies as a core part of it, it’s all growing, it’s all, you know, forecast to continue to grow quite tremendously.
For the digital human side in particular. Has there been any research done on that kind of market forecast? Is it forecast that every company will have a digital human by x date? Or be, you know, how would you compare it to the adoption of other things like chat, like mobile, like social, like voice assistants, and all that kind of stuff? Like, what if you could shed a bit of light on what you see as, as the future of adoption
Danny Tomsett 56:38
More recently, last year to two years, you’ve seen, like Gartner and others now have specific emerging tech reports that are focused on digital humans? I’m trying to remember now I couldn’t quote exactly. You know, it’s 100, and something billion in the next six years or something. But, yeah, a lot more focus around digital humans now. And the key players, and you’re seeing big players like Microsoft and others, that to get into this, Baidu just announced the work they’re doing in space as well in China. So that’s done to happen.
Danny Tomsett 57:13
I wouldn’t, I wouldn’t say that I’ve seen anything around every home or every brand or every enterprise. But I would say in the next five years, you’ve already seen the evolution of and home devices with screens. We’ve seen that across Google and Alexa already, those screens, we already know there are conversations taking place around more interactive experiences than a visual that makes total sense that you know, Geico, we’re going to have the Geico Gecko or Gordon Ramsay is going to help you design a meal and your home while you’re cooking. And then it connects into your local supermarket, provides the supplies, just like that, wherever it’ll probably be whole foods actually, if it’s Alexa. But yeah, that’s, that’s the natural evolution, I think of natural language and what has been designed, the evolution makes sense, because as we’ve just identified, the key dependency is you need to have a conversational engine, that will be able to meet the needs of the use case. And once you’ve solved for that to a sufficient level, you amplify it with visual experience around it. And we’re at that point now where that’s really starting to happen.
And your potential market is already laid out for you. Because all you need to do is just go to the enterprises that have got sophisticated conversational AI setups, with good chatbots and decent voice assistants in the call centre, like everyone’s doing now. And you’ve got a market there, because inevitably, once they’ve got skilled enough, and they’ve taken their channels to where they can get them to, the next step would be adding that visual component. And I think that they would also find it easier as well, about a bit the same language modelling the exact same conversation, as we’ve discussed, but at least, the skills to be able to do it, the foundational technology to be able to do it. So your market is basically laid out in front of you
Danny Tomsett 59:07
Exactly. The one, I guess challenge I would say is that, you know, there’s a lot of investment around the customer service contact centre side. Still, I think it’s always best to bring digital humans and a positive use case. So in sales and marketing, they are less invested. And that conversational side, so we’re still educating and helping and connecting at this point in time, but I think a lot of the no code solutions are actually for natural language understanding and designing the conversational flow. A lot of these no code products, I think are helping a lot in that particular area. And we’re obviously plugged into anything so as long as it’s able to be designed as a good compensation as low latency then then it’s a good fit.
Fantastic, wicked. That is really Good. Thank you so much, Danny for joining us. Where can people I put the I am Sophie, link in the show notes and whatnot. But what about the unique website? How can people learn more? How can people kind of like, you know, reach out? Like, what’s the best way for them? Yeah,
Danny Tomsett 1:00:13
The website, you can actually go to www.digitalhumans.com or www.uneeq.com, they will take you to our website, lots of information there, lots of ways to connect them with the team. And myself. And yeah, we always love to have conversations with those in the industry or people interested in trying out the technology we can help with that as well.
Fantastic. Sounds really good. Well, this has definitely been an education. For me. I really enjoyed the conversation. It’s yeah, so much potential to be honest for this stuff. I can see it. See it laid out because he has all kinds of opportunities for it. So it’s been really good for you to spend some time with us. Yeah,
Danny Tomsett 1:01:01
Thanks, Kane. Appreciate it as well. It’s been fun.