Digital Humans: AI Ethics, Applications, ROI, and Future of Customer Interactions w/ Danny Tomsett

[00:00:02] One of the frameworks I like talking about is the idea of applying ethics to AI implementations. Well, it gets interesting when we start talking about replacing humans. If you watch my live show, we've done some experiments with AI versions of me. But what if we instead did a digital human? What that might look like with Danny Tomsett, the CEO of Unique, on this bonus episode of The Business of Tech.

[00:00:27] Well, Danny, welcome to the show.

[00:00:30] Thanks, Dave. Good to be here.

[00:00:32] So I got to understand because you're building unique digital humans. I need to know more about what a digital human is and where you're seeing that AI avatar being applied.

[00:00:46] Yeah, when I started this company, I did a lot of explaining about what AI was, let alone what an AI digital human was. So we live in an interesting time now where a lot of people are more familiar with AI.

[00:00:59] But digital humans, what is a unique digital human? A unique digital human is an interactive avatar or an interactive character that exists in digital channels, whether it be a kiosk or whether it be a website, mobile app, that people can talk to in real time and get assistance and all sorts of use cases. I'm sure we'll talk about a few of them today.

[00:01:23] Well, so give me a few of them today.

[00:01:25] Well, so give me a second. I would think that this applies really well in customer service or marketing. Like, give me a couple of the standard use cases where this technology makes a lot of sense.

[00:01:36] Well, I think one of the interesting areas this technology is really proving to be successful is just the nature of accessibility and natural interaction. What's been fascinating over the years, as you'll know, is humans had to be more machine-like to interact with machines.

[00:01:56] You had to learn the machine to then be able to use the machine. With technology like digital humans and AI, now you can be human and the machine knows how to be more human.

[00:02:06] And so that's important because to that very question, where does this play out best? And often we have a huge amount of challenge when we're trying to get customer service. Sometimes that feels like the worst experience in the world, sitting on a call center or trying to wait in line in a retail store, for example. Digital humans are always on. They listen to you. They personalize the responses and they're actually able to understand emotion as well.

[00:02:35] And that's a really interesting dynamic with emotional intelligence, not just knowing what you say, but addressing how you feel and responding to make you feel more at ease.

[00:02:45] So let's start with, because I want to explore a little bit about your thinking and the intentional decisions you've made around both positioning and ethics.

[00:02:53] And I'm going to start with kind of positioning because it's a very deliberate choice to choose digital human because we can obviously say they're not human.

[00:03:02] They are very much not human.

[00:03:04] And so you've actually had an interesting juxtaposition of acknowledging it's digital, yet also saying it's human.

[00:03:12] Tell me a little bit about the decision and the word choice to describe it this way.

[00:03:18] Yeah, we thought right from the beginning that this technology was only going to be effective if people could trust it and feel that this would be a natural interface to use.

[00:03:29] So we actually, as a company, purposely designed digital humans to be obviously digital, obviously characters, not deep fakes or these types of AI technology sometimes that are actually used in many ways to fool people.

[00:03:46] I think these things are problematic.

[00:03:48] And we've seen through our research and our customers that this is not actually something people want to use.

[00:03:53] But when it comes to a character, why is this important?

[00:03:57] And why is this human connection actually an important position?

[00:04:03] And the reality is that most of us don't operate just on our left brain.

[00:04:08] We use our right brain too.

[00:04:10] We have this feeling, creative, emotional part of how we interact with even companies on purchasing decisions, customer support, training and learning.

[00:04:19] And so we know the human aspect of an interface is really key to engaging that part of the brain.

[00:04:26] And when we do it well, it's super effective.

[00:04:29] So that's also part of the ethics of what we do too, leading to the other part of that question is ensuring that a digital human is one, authentic to the brand.

[00:04:39] So it's not trying to be something that it's not.

[00:04:43] So it's representing the brand well, but also something that an end user or customer can trust.

[00:04:49] They know what it represents.

[00:04:50] They know it's AI and they know how to safely use it and effectively use it.

[00:04:55] You've clearly given a lot of thought to the kind of ethics and the best way to present this technology.

[00:05:00] So can you kind of give me a little bit of the sense of the framework you've put together to guide the business?

[00:05:06] So when we talk with any business, we often break it into four categories.

[00:05:11] So the first category that we talk about is what does the digital human look like?

[00:05:17] How does that personality project as well?

[00:05:22] So it's not just aesthetics.

[00:05:23] It's not just clothing.

[00:05:25] It's not just how they look, male, female, age, whatever that might be.

[00:05:29] It could even be a character, right?

[00:05:30] Like a Geico Gecko.

[00:05:33] And so that's the first part is who is your audience?

[00:05:38] And what are you trying to convey to your audience?

[00:05:40] What's authentic to your brand?

[00:05:42] Your brand isn't just a logo.

[00:05:43] It should be a feeling that's invoked through even embodiment, embodiment being the characters, that personality.

[00:05:50] So that's the first part we go through.

[00:05:52] The second part is actually what is the actual job?

[00:05:55] Think of it like a digital worker.

[00:05:57] What is it tasked to do?

[00:05:59] And how does it do that really well?

[00:06:01] And what do we need to integrate that into to do that task well?

[00:06:05] And often you've got information systems, whether that's a CRM or it could be financial systems, could be call center systems, whatever that might be.

[00:06:13] We can integrate into that to ensure that what we're trying to create is a super brain, something that's consistent with knowledge and information as well as consistent with personality and the experience you have.

[00:06:25] The third part of that is now how do we want to make that accessible?

[00:06:30] Where are people going to interact with this?

[00:06:32] Is it physical?

[00:06:32] Is it in a store?

[00:06:34] Is it online on a website?

[00:06:36] Is this a VR experience?

[00:06:37] Are we going to create something super immersive and what that looks like?

[00:06:41] And then the last part of that is actually the operating model.

[00:06:44] What does the team look like that looks through the reports and sees what it's doing well?

[00:06:50] What's it not doing well?

[00:06:51] How do we retrain it to do better?

[00:06:53] Everything we want to have is supervised.

[00:06:55] And that also ensures that you're catching things that might not be helpful and you're correcting them and making that even better.

[00:07:02] And that tuning process never ends.

[00:07:04] Like any good employee that you might work with, you constantly should be reinvesting and training and helping them to get better.

[00:07:11] Now, this is interesting in that it's one of those technology implementations that is not human in the loop, right?

[00:07:17] So one of these, a lot of AI advocates will talk about the fact that augmenting humans is an ideal solution for AI.

[00:07:24] But here, this is clearly a replacement for some roles, potentially expanding, but it also could be viewed as replacing.

[00:07:31] How do you think about the kinds of roles that are right for this kind of use and the kind of roles that are not right for this technology?

[00:07:39] Well, I think also in that, I think like anything, there's a spectrum.

[00:07:43] Like think self-driving cars.

[00:07:45] It's no different, right?

[00:07:46] Like if the self-driving car follows another car a little bit closer than maybe what it should do, right?

[00:07:54] But there's no major consequence.

[00:07:55] You're going to be pretty forgiving of that error if it doesn't result in anything catastrophic.

[00:08:01] As opposed to the car sacrifices you, drives off a cliff, and you're like, what the heck just happened, right?

[00:08:07] And so that would not be a good error.

[00:08:10] And so it's no different in many other realms of AI.

[00:08:13] There's a spectrum in conversation and digital humans.

[00:08:16] And the use case often determines the criticality of the answers and the level of accuracy.

[00:08:24] So if you said it was in health diagnostics, you don't want to get anything wrong, right?

[00:08:30] So the level of acceptance for accuracy and perfection would need to be very high in that.

[00:08:37] But what if we were using it in training?

[00:08:40] So we have a sales training product.

[00:08:42] And what we do is people can interact with a digital human.

[00:08:46] That digital human is simulating a customer experience.

[00:08:49] So a new salesperson who needs to learn their products and learn how to pitch and all these things, they can practice with that digital human.

[00:08:57] And it's proven to be super effective.

[00:08:59] But it's also, you know, if the digital human says one thing that's slightly off or whatever that might look like, the consequences are pretty low in that.

[00:09:07] Ultimately, the end goal is still achieved and everyone enjoys using it.

[00:09:11] And that's equally relevant to maybe it's a product expert and things like that.

[00:09:16] What I will say is just like self-driving cars, this technology is just getting better and better and better and better.

[00:09:23] Like it's very rare now for us to not be able to even say we can guarantee what it will say that it will stay within the guardrails.

[00:09:33] Think of it like if you're ever in a bowling alley and you put the guardrails up on the side, we can kind of make sure the bowling ball is always going to hit through the center.

[00:09:42] It's not going to fall in the gutter.

[00:09:43] And so we can do that really effectively.

[00:09:45] Now, this is interesting in that it's one of those technology implementations that is not human in the loop, right?

[00:09:51] So a lot of AI advocates will talk about the fact that augmenting humans is an ideal solution for AI.

[00:09:57] But here, this is clearly a replacement for some roles, potentially expanding them too.

[00:10:03] Also could be viewed as replacing.

[00:10:04] How do you think about the kinds of roles that are right for this kind of use and the kinds of roles that are not right for this kind of technology?

[00:10:12] Yeah, I think the best way to describe where digital humans are best often talk about interactions versus transactions.

[00:10:22] And there's actually a lot of, I think, parallels to human to human connection.

[00:10:27] So Dave, you and I become good friends and we're going to text message.

[00:10:32] We're going to call.

[00:10:33] We might video call.

[00:10:34] We might meet in person, right?

[00:10:35] This is just a normal part of friendship.

[00:10:37] But we use these different modalities based on what we need.

[00:10:42] Is it just a quick answer?

[00:10:44] And is it asynchronous versus needing to be real time?

[00:10:47] All these things we do instinctively.

[00:10:49] That's just part of being alive, being human and having relationships.

[00:10:54] Well, it's no different for digital humans.

[00:10:55] If you want to create an experience with a digital human, it's really in comparison to a more in-depth conversation back and forward and what we call an interaction.

[00:11:08] Right?

[00:11:09] If I use the use case as an example, we work with a telecommunications company.

[00:11:15] And telecom products can be quite complex for some people.

[00:11:19] And we went and understood some of these products and we're like, okay, what if we put a digital human at the front of the site and people could talk about in their own language just what they need?

[00:11:30] I'm moving to a new house.

[00:11:31] I've got four kids.

[00:11:32] They do lots of gaming.

[00:11:33] I like Netflix, whatever that looks like, you can just talk about the things that matter to you.

[00:11:38] The digital human is listening, asking questions.

[00:11:41] And at the end of that, the digital human is able to say, hey, that's really helpful.

[00:11:45] This is what I actually think you need.

[00:11:47] You need this plan.

[00:11:48] You don't need to understand what a gigabyte is.

[00:11:50] It's just you need this.

[00:11:51] Right?

[00:11:52] And here's some of the things that we want to add to this.

[00:11:54] You probably need four wireless access points for a house that size and et cetera, et cetera.

[00:11:58] Well, we ended up increasing the conversion rates by 500% just by putting that digital human in that scenario.

[00:12:06] And you say, well, couldn't you just do that with a chat bot?

[00:12:09] Well, they did.

[00:12:10] And the problem was that the engagement rates were less than half of a digital human.

[00:12:15] Then you've got to ask yourself why.

[00:12:17] And the reason why is the digital human represents an interface that's more interesting, more effective, and actually really emotionally responsive.

[00:12:27] Like confidence and trust actually comes from more than just what you say.

[00:12:31] Confidence and trust comes from how you say.

[00:12:34] And that's why this tech works so well.

[00:12:36] So one of the things that I'm really interested to understand then is talk to me a little bit about the calculations for ROI and investment that you're doing here.

[00:12:44] Because it feels like my first instinct is, well, you could apply human level ones to this.

[00:12:50] But maybe you can.

[00:12:52] Maybe it actually has to be a little different.

[00:12:53] Tell me a little bit about how you're measuring the ROI there.

[00:12:56] Well, I think one of the biggest challenges for most businesses actually is how do we multiply what we do without huge overheads of staff?

[00:13:06] It's not really always about staff reduction.

[00:13:09] Sometimes that does apply, but it's not as common as people would think.

[00:13:13] Like most of the time, it's how do we expand and reach a larger audience, more customers?

[00:13:19] How do we do this at a better level of quality and experience?

[00:13:23] So two key factors in what they're looking for.

[00:13:26] Typically, if it's revenue orientated, it's going to be how are we generating more revenue?

[00:13:30] And the other metric on the other side of that is what's our cost to serve to do that, right?

[00:13:36] So can we maintain or grow our margins while reaching a larger audience?

[00:13:41] And that's obviously something with the digital human.

[00:13:44] You can have 24 by 7, 365 days a week.

[00:13:48] Now you're actually going to new regions, new countries, new ways of supporting customers.

[00:13:53] You're also taking load off existing staff that might be not high value use of their time.

[00:14:00] And that's another big area that often our customers and companies look at is around, well, what are people focused on?

[00:14:09] And how do we ensure that their time, attributed time to high value, right?

[00:14:14] And so they're looking at how much of their time is actually spent on high value tasks that are generating, again, more revenue, higher levels of customer experience, retention, lifetime value on contracts or extending, things like that.

[00:14:26] So it's quite business fundamentals that go across what this does.

[00:14:31] And training, a big part of this actually that we've seen is just more effective workforce.

[00:14:37] So a massive metric that we just have like that sales training product that we have, you know, most companies spend up to like a million, two million, depending on the size of the company.

[00:14:50] But in a smaller business, they can spend millions on pipeline burn, just new salespeople coming in and actually just practicing on customers.

[00:15:00] It's the worst way to practice.

[00:15:01] They get it wrong.

[00:15:03] The customer goes away and they do that over and over again.

[00:15:05] Then you bring a new salesperson and then a new salesperson.

[00:15:08] They just do it over.

[00:15:08] And so in a year, you're burning so much profit because you don't have an effective way to train.

[00:15:14] So it's a really great metric for just using this technology in an internal training way and makes people more effective, gives management and owners more visibility on how effective are their people, where do they need to train them more, things like that.

[00:15:29] Now, as we sort of get to the end of this discussion, I want to get a little sense of the speed of this.

[00:15:33] You talked about the fact that the technology is getting better and better.

[00:15:36] And of course, most technologies do.

[00:15:38] Now, one of the areas that's interesting to me, particularly about those related to generative AI, and I know there's got to be a generative AI component to what we're doing, is that there's some perception now that the model maturity is starting to slow.

[00:15:49] And that we're starting to see some of the charts.

[00:15:52] Talk to me a little bit about what you're seeing in terms of what you're implementing and the speed of that versus the model's potential slowdown.

[00:15:59] Yeah, it's actually really interesting because the models themselves are very, very good at understanding context and language and really bringing together meaningful human connection, right?

[00:16:13] So it feels relevant.

[00:16:14] And look, you're listening to me, the information, anyone that's done ChatGPT has had that wow moment, right?

[00:16:20] Well, effectively, the models themselves cannot really effectively work with real time or relevant information because it wasn't used in the model.

[00:16:30] The model might have out-of-date data, things like that.

[00:16:33] So you've got this now agentic architecture.

[00:16:36] This is now really complementary architecture to leverage the model to benefit the conversation, but taking over more the accuracy and information layers.

[00:16:49] And so that used to take a long time.

[00:16:52] Now what we're talking about is customers implementing within a couple of weeks to do that.

[00:16:57] Obviously, it does matter about where your data is, how good that data is, right?

[00:17:02] It's only as good as the data you're using.

[00:17:04] But most customers now have data systems have for a long time.

[00:17:08] And so that makes this a lot quicker in what they're doing.

[00:17:12] Average deployment times are normally one to two months across this technology now.

[00:17:18] So and what are you finding are the biggest hurdles or issues to get over to get really successful implementation?

[00:17:27] Well, I think probably in the past, one of the biggest was how do we ensure hallucination and some of those larger model issues don't play out in the real world.

[00:17:37] Now, as I said, with those new architectures, we have a lot more control.

[00:17:40] So that's a big area that we're seeing huge improvement with.

[00:17:44] So it's super exciting.

[00:17:45] It's one of the most exciting times, really, because we can leverage this technology in the real world in a safe way.

[00:17:51] But I think still one of the one of the bigger areas is, again, just ensuring that when people are using this technology, how do you ensure that you're creating trust and authenticity with your audience?

[00:18:04] Right.

[00:18:05] You don't just throw technology and go, hey, here's tech.

[00:18:08] You should love it.

[00:18:09] Right.

[00:18:09] You really need to think through about how do I present this to our customers so it's to their benefit that they understand how this is going to help them.

[00:18:17] They understand that their data is safe and how is this going to treat their data and their information?

[00:18:23] And are there people observing and looking after it and making it better and those types of things?

[00:18:27] And I think that just requires really good communication and transparency.

[00:18:31] And you've just outlined what I keep talking about is the real value of service providers in this equation.

[00:18:37] Well, Danny Thompson is the CEO of Unique, a pioneering company in AI-powered digital humans.

[00:18:43] He's been instrumental in advancing conversational AI to develop and deliver emotionally intelligent avatars that enhance brand-to-customer relationships and drive innovation in marketing and customer service.

[00:18:54] Danny, this has been really interesting to learn more.

[00:18:56] Thanks for joining me today.

[00:18:57] Thanks, Dave.

[00:18:58] It was really great.

[00:18:58] Are you ready to get your brand in front of the tech leaders shaping the future of managed services?

[00:19:06] Here at The Business of Tech, we offer flexible sponsorship opportunities to meet your needs, whether it's live show sponsorship, podcast advertising, event promotion, or custom webinars.

[00:19:18] From affordable exposure options to exclusive sponsorships, our offerings are designed to fit businesses and vendors of all sizes looking to make an impact.

[00:19:27] Prices start at just $500 per month, making our packages a fraction of typical event sponsorship costs.

[00:19:36] Be a part of the conversation that matters to IT service providers worldwide.

[00:19:42] Join us at MSP Radio and amplify your message where it counts.

[00:19:47] Visit MSP Radio dot com slash engage today to explore all the ways we can help you grow.

[00:19:56] The Business of Tech is written and produced by me, Dave Sobel, under ethics guidelines, posted at businessof.tech.

[00:20:03] If you like the content, please make sure to hit that like button and follow or subscribe.

[00:20:08] It's free and easy and the best way to support the show and help us grow.

[00:20:13] You can also check out our Patreon where you can join the Business of Tech community at patreon.com slash MSP Radio or buy our Why Do We Care merch at businessof.tech.

[00:20:26] Finally, if you're interested in advertising on this show, visit MSP Radio dot com slash engage.

[00:20:33] Once again, thanks for listening to me and I will talk to you again on our next episode of The Business of Tech.

[00:20:42] Part of the MSP Radio Network.