Episode 5: Tom Sherwood from Koios

[00:00:00] Hello and welcome to the Conversational AI News Podcast and today we're talking emotionally intelligent AI. I'm joined by Tom from Koios. Tom, hello.

[00:00:09] Hi.

[00:00:09] How are you doing man?

[00:00:10] I'm good thank you. I'm good. How are you?

[00:00:12] Very well, very well. It's good to have you here Tom. I am very excited to be talking through the technology and the capability that you've got because we got a preview at the Conversational AI, the mini summit recently. But would you, before we begin on the company, who are you and what's your team?

[00:00:29] What's your background Tom?

[00:00:31] Of course. So, yeah, first of all, thank you for having me. I really appreciate the event last week, a couple of weeks ago, it was great fun. So my background, I am quite different, I think is probably the best way to put this. I left school at 16, which gave my dad a heart attack. My sister's got an MBA, my dad's got a PhD and at 16 I was like, I hate school. I am out of here. Education is the worst thing in the world. I was a 16 year old at the time.

[00:00:57] So it's a different perspective. But I actually ended up joining the British Army. So I spent five and a half years in the British Army, including operational tours of Iraq, Afghanistan, all before my 21st birthday.

[00:01:11] And then after doing that, I kind of had this mini checklist to do when I joined the Army, which was play football for the Army, go to Iraq, go to Afghanistan, ticked all three and was like, well, time to go.

[00:01:24] Right, right, right, done.

[00:01:26] Yeah, done, served its purpose. All of my friends were at uni or whatever. And obviously they were graduating. I was leaving the Army. They were graduating with loads of debt. I was leaving the Army with no debt. I had savings. They didn't, which was quite a nice transition.

[00:01:41] And then kind of entered the big bad world. It was lucky I had those savings because my first job was agency recruitment. So I started working for a big US recruitment company, which starting salary of £18,000.

[00:01:54] It was certainly a reality check of living and working in zone one London and then having £18,000. The salary in 2011 was, yeah. Anyway, keeps you hungry is what they said, like literally keeps you hungry.

[00:02:11] So, yeah, spent seven years in agency recruitment and then I moved in-house and worked in talent acquisition at Spotify, pre-IPO and post-IPO, Apple and Snapchat. And that was the kind of grounding and understanding of the world of human relationships that kind of led us to today.

[00:02:29] Right. Great overview there, Tom. So, yeah, take us into today now. Talk to us about Koyos. How was it formed? Where did the idea come from? And what's the service?

[00:02:39] Yeah, of course. So we've actually gone through quite a transition, but it all started around psychometrics, psychology. Something that kind of referencing back to my time at school, I hated school quite clearly.

[00:02:53] But something that was quite apparent was there were a couple of teachers that really got through to me and I excelled and I did really well in those subjects.

[00:03:00] The other subjects I really struggled in, not through lack of academic capability, more the fact of I just wasn't bought in.

[00:03:11] I backed my head with a teacher. I mean, I was a 16-year-old boy, right? You're always going to fall out with teachers, just who I am.

[00:03:20] But I started to get an understanding of, well, hang about. If you treat people differently and you communicate it, how does that work?

[00:03:27] I suppose subconsciously I was thinking about this for 15, 20 years.

[00:03:32] And the idea and the big thing was when I was at one of these big tech companies, they brought in psychometric testing as part of their hiring process.

[00:03:40] And it's been used for generations. And I mean, you could argue that it's used in the highest performing teams in the world.

[00:03:46] U.S. Delta Force do psychometric testing as part of their hiring process to enter special forces, which you could probably argue is the highest stressed, highest performance organization in the world.

[00:03:57] U.S. Delta Force do so. They brought it in, but then quite quickly dropped it because the method in which they're acquiring this information or gathering this information was very old, very expensive.

[00:04:07] It was self-assessment. I'm sure most of your listeners and most of the people in the professional world would have gone through some kind of personality psychometric testing because it is incredibly valuable.

[00:04:17] If you as a human can understand exactly how the other human you're engaging with thinks and is motivated and likes to be communicated with, it accelerates that relationship.

[00:04:27] And you can, I mean, you will speak differently to your mom, your dad, your friends from school as you would to someone you just met because you understand over time subconsciously and consciously how to deliver your messaging differently based on who they are.

[00:04:40] And that was something that always fascinated me. And that's obviously why organizations wanted it because they wanted to have better onboarding, wanted to be able to coach them more effectively.

[00:04:48] And they only get a small amount of time really to kind of coach people in a company.

[00:04:52] So they want to make sure it's efficient and they can hit the ground running and using psychometrics and personality and psychology is the fastest way to do that.

[00:05:00] So that's kind of how it started. But we wanted to change the way it was acquired.

[00:05:04] So myself and my co-founder, Martin, who PhD in computational science, computational social science, which is the application of AI to sociology.

[00:05:15] For those who don't know, wanted to change this.

[00:05:18] And over a three to four year period, we bootstrapped. We're working on this in the background.

[00:05:23] We've been able to develop a series of models that are able to accurately predict and understand your personality and psychometric makeup purely from the way that you speak.

[00:05:32] So the acoustic and linguistic signature of your voice is a massive indicator of your personality.

[00:05:38] And that's where we got to today.

[00:05:41] Goodness me. So can you say a little bit more about how that works then?

[00:05:46] So you listen to me for what, a period of time and then you can infer things about gone. Say some more.

[00:05:52] Yeah. Yeah. So there's there's about two and a half thousand data points in your voice.

[00:05:56] So linguistic and acoustic, the pauses, the speed, the rhythm, the emotion and the words that you use, the structure of the sentences are just some examples.

[00:06:06] So, I mean, a great example of this is the word fine.

[00:06:08] The way the word fine is delivered has so much context behind it.

[00:06:13] So, so, so that thing.

[00:06:16] Yeah, yeah, yeah.

[00:06:17] Exactly. So those, those things that your teachers and your parents told you and you're probably your manager in your early career, it's not what you say, it's how you say it is scientifically true.

[00:06:26] And essentially, it's a massive, massive indicator of who you are.

[00:06:29] So that's, that's what our kind of foundational model is based on.

[00:06:33] And we, we understand personality and we did a lot of research around what obviously different sounds and different speech relates to with regards to personality.

[00:06:41] And we base ours on the big five personality model, which is really, really, really robust.

[00:06:47] And any behavioral or organizational psychologists, all of their research, all of their practice will be based on the big five.

[00:06:53] And because it's the most robust version.

[00:06:55] So if you haven't heard of it, it's probably because your company are still using Myers-Briggs, which without going into too much depth,

[00:07:02] you ask a psychology, sorry, you ask a psychologist about Myers-Briggs, they'll probably come out in a cold sweat.

[00:07:07] Because it's just not scientifically applicable.

[00:07:10] Before we get to the use cases, because I can imagine quite a few, how does this tech work then?

[00:07:15] You're, are you taking an audio sample from me real time?

[00:07:20] It depends on the context.

[00:07:23] So we are able to analyze real time communications.

[00:07:26] So if you, you and I on this at the moment, we can have our API running and it's able to capture, obviously, snippets of your voice.

[00:07:37] So once we have 90 seconds worth, we can start running the algorithm.

[00:07:41] So from 90 seconds to 10 minutes is where we see the highest performance.

[00:07:45] From in the zero to 90 seconds, it's quite a rapid curve to kind of get to a level of accepted accuracy.

[00:07:51] And then from 90 seconds to 10 minutes, it's quite a gradual increase.

[00:07:55] But then from 10 minutes, it kind of flatlines.

[00:07:56] We can't get much more accurate than that.

[00:07:59] It can be used retrospectively.

[00:08:00] So any kind of video communications or any digital form of podcasts, YouTube videos.

[00:08:08] We were just doing some experimentation recently where we're building a prototype of a product that we'll talk about in a minute.

[00:08:15] And we wanted different personas.

[00:08:17] And we thought, well, what better than just find some famous people this week?

[00:08:20] And we analyzed Oprah Winfrey.

[00:08:22] We analyzed Elon Musk.

[00:08:23] We analyzed Donald Trump, obviously.

[00:08:27] I mean, why wouldn't you?

[00:08:29] You want to know about that guy, right?

[00:08:32] Rihanna.

[00:08:33] But just kind of see so we could have some examples of different personalities respond to different things.

[00:08:39] So, yeah, any kind of recorded conversation we can take from.

[00:08:44] Then, okay, so use cases.

[00:08:46] And I can think of quite a few here, Tom.

[00:08:48] What are you – where are you beginning?

[00:08:50] Yeah.

[00:08:51] So with regards to our API, so just our ability to create the prediction,

[00:08:56] a lot of usage is coming from the recruitment space or interviewing space,

[00:09:01] which actually comes into venture capitalists as well.

[00:09:03] I mean, we're working with a VC firm at the moment who wants to be able to understand the potential founders.

[00:09:08] So every time somebody applies for funding or they submit their pitch deck,

[00:09:13] they're asking them to send a two- or three-minute video about why they have this company.

[00:09:17] And that's a great, really rich audio source for us, obviously, because we can analyze that and provide the insight.

[00:09:23] But also for the VC, they are – especially at early stage, they're investing in the founders.

[00:09:28] They're not necessarily investing in the idea.

[00:09:30] So the ability to provide insight there is really valuable to them in recruitment.

[00:09:35] So obviously more and more recruitment platforms now are completely video-based.

[00:09:41] But also a large majority of organizations, especially at the enterprise level,

[00:09:45] as part of the application process, ask you to upload a five-minute video about why you're interested.

[00:09:50] I want to be the VP of sales at Coca-Cola because I've worked at PepsiCo for this amount of time.

[00:09:54] Blah, blah, blah, blah, blah.

[00:09:55] And because at the end of the day, you are hiring humans.

[00:09:58] You're not hiring CVs.

[00:09:59] Yes.

[00:09:59] So we're running the analysis on that and providing insights for them in hiring.

[00:10:05] Then also within the sales enablement space or essentially any relationship industry,

[00:10:10] whether it be sales, customer service, we are providing insights to the other human on the other end

[00:10:16] to say, okay, well, I know who I'm speaking to.

[00:10:19] How can I accelerate that understanding?

[00:10:21] How does that work then?

[00:10:22] So I get that I'm calling your API, but I'm giving the audio to your API.

[00:10:27] What do I get back from the API?

[00:10:30] Am I getting reports?

[00:10:31] Am I getting PowerPoints or something?

[00:10:33] Yeah.

[00:10:33] So it depends on the use case.

[00:10:35] So the two main APIs that are out in the market at the moment for us are our Talent API, which is essentially what's on the tin.

[00:10:44] It gives you a summary, like a one-page summary, a bit of an overview.

[00:10:50] This is you and this is what his personality is like.

[00:10:52] This is how he comes across.

[00:10:53] And then we go into a bit more detail about how you like to communicate, how best to communicate with you, how not to communicate with you.

[00:10:59] Because that's just as important as how to.

[00:11:02] What motivates you?

[00:11:04] And then very clear instructions on how to motivate you, what your business attributes are.

[00:11:08] So what people don't realize is actually a lot of really key business attributes that are based on personality.

[00:11:13] And that's hardwired from your teenage years.

[00:11:16] So, for example, I am, as a person, I am at the kind of mid to high level of neuroticism.

[00:11:21] So I'm very emotional person.

[00:11:23] I get a lot of energy from my emotion.

[00:11:24] I'm also extroverted, which means I'm always coming out of ideas.

[00:11:28] I'm thinking.

[00:11:29] I'm kind of thinking about what next, what next.

[00:11:32] But I struggle with being sat down doing a really mundane task.

[00:11:35] I really struggle to do that.

[00:11:37] Whereas someone with a different personality to mine, so a different big five structure, would excel in that task.

[00:11:42] So it just highlights, get Tom on projects where it's like, think about the future.

[00:11:46] That's where you'll get the best out of me.

[00:11:47] But then also, the growth opportunity with me is I need to practice my kind of detail-oriented stuff,

[00:11:54] kind of build discipline in my administrative tasks.

[00:11:57] So that's what we do in the talent world.

[00:12:00] In the sales world, it's very much about relationship building.

[00:12:03] So, say, if you and I are having a conversation, I'm the seller, you're the buyer.

[00:12:10] It will provide me with insights in, again, how to communicate with you.

[00:12:13] What kind of cadence should I be engaging with you?

[00:12:18] What kind of information should I be providing you with?

[00:12:21] Not just the type of information, but how it's delivered.

[00:12:24] So, do you want really detailed 10-page reports to every single data point,

[00:12:29] or do you just want a high-level information?

[00:12:31] Different personalities require different deliveries.

[00:12:35] That's the kind of stuff we provide in the sales space.

[00:12:38] And then also, there's some other things to do with different kind of buying science.

[00:12:42] So, for example, as I was saying, I'm kind of at the mid-to-high level of neuroticism.

[00:12:47] The best way to sell to me, and this is probably going to give a bit of a secret to everybody who's listening to this podcast,

[00:12:53] trying to sell me a service, is take the deal away from me.

[00:12:56] Like, create an element of fear or loss and kind of use the emotional side of my personality to make that deal.

[00:13:04] If you were to use that technique with me, I'm more likely to respond because I'm emotional, I'm impulsive.

[00:13:09] I think with my – I trust my gut a lot on these things.

[00:13:13] If you were to use that same tactic with someone who had the kind of opposite personality to me,

[00:13:17] so someone who's very, very emotionally regulated, probably an introvert,

[00:13:20] and try to use fear or kind of time pressure,

[00:13:24] they'll just be like, okay, why are you telling me that?

[00:13:26] I will make my decision in my own time.

[00:13:28] So it's all about psychology and understanding the human on the other end.

[00:13:33] So if I was trying to pitch you then, Tom, is the best way of doing that,

[00:13:40] to go and do some research, find a podcast off you, find a video off you,

[00:13:44] run that through my COIL system, get some insights then?

[00:13:50] I'm asking because I've seen mention in the past of call center style systems

[00:13:56] that will measure sentiment or the different tone of voice or is the customer breathing heavily.

[00:14:04] But that's kind of a real-time feedback.

[00:14:06] This isn't real-time, I think.

[00:14:08] Is that right in saying?

[00:14:09] Yeah.

[00:14:09] So ours is more about – ours is a lot deeper than sentiment.

[00:14:13] And sentiment is – yeah.

[00:14:16] But also sentiment will lead into how you are that day, your emotion that day.

[00:14:20] If you've just come out of back-to-back meetings and you've just lost –

[00:14:24] or something's gone terribly wrong and Vodafone call you because you need to renew your contract,

[00:14:28] you're going to be in a very, very different mindset that day.

[00:14:31] And if you've just – I don't know, it's your birthday, you've just posted a couple of deals

[00:14:34] and the world's amazing, blah, blah, blah.

[00:14:36] You'll be like, yeah, cool.

[00:14:37] Yeah, let's go for international calls.

[00:14:40] I don't even travel, but let's do it.

[00:14:41] Like, you're going to – that's kind of sentiment rather than – it's that kind of, yeah.

[00:14:46] That's a – or behavior versus personality.

[00:14:49] Behavior can change and evolve, and personality is relatively fixed.

[00:14:55] So, yeah, that's the kind of key difference.

[00:14:58] So that would indicate that the use cases for the Koyos technology,

[00:15:03] the two APIs you have at the moment, is planned.

[00:15:08] Planned, as in talent, as you say, people joining or, you know,

[00:15:13] I'm going to look at sales here, so I'm going to do my analysis before we do –

[00:15:18] you know, before we interact, that kind of problem.

[00:15:20] Yeah, absolutely.

[00:15:22] If it was a one-hit sales call where you're trying to sell me double glazing

[00:15:26] and it's not about building a long-term relationship,

[00:15:30] obviously if you've spoken to me before, you'll have that information

[00:15:32] because you might have bought that data or whatever.

[00:15:37] But predominantly it's about relationship building, which is B2B sales.

[00:15:41] It's about, yeah, customer service or account management, et cetera, et cetera.

[00:15:46] So it's just accelerating that relationship between the two individuals.

[00:15:51] Okay, okay.

[00:15:52] Business model-wise, how does that work?

[00:15:55] Give us a perspective on how – if I'm adopting one or both of your APIs,

[00:16:00] how does the compensation or the business model work?

[00:16:04] So if we – for our APIs, we charge per API call.

[00:16:07] So for each individual that we analyze, that's essentially one pin each time.

[00:16:13] Right.

[00:16:13] And then I would imagine there's a lot of people listening who are thinking

[00:16:19] of various different applications that you haven't necessarily explored publicly

[00:16:25] at the moment, Dom.

[00:16:26] How open are you to hearing from different use cases?

[00:16:32] Yeah.

[00:16:33] It's kind of a weird situation because it's an incredibly powerful tool

[00:16:40] and kind of in the wrong hands it could potentially be used.

[00:16:43] But then AI, you could argue, in the wrong hands of anybody can be used in a detrimental way.

[00:16:48] But anyone who's looking to build better relationships understands individuals

[00:16:53] who they're speaking to in more depth to give them that advantage to –

[00:16:58] and I'm not saying advantage in an ethical way.

[00:17:01] But, for example, we've listed in each other a while now.

[00:17:05] But if I had like a how-to guide of how to work with you within the first two minutes

[00:17:10] of knowing each other, it would accelerate our relationship even further.

[00:17:13] Yes.

[00:17:14] It just accelerates everything, which is just a massive edge in any kind of –

[00:17:19] I mean, if you knew as a – whether it be customer service, whether it be financial sales,

[00:17:24] obviously there's a lot of research we've been doing internally around how different personalities

[00:17:28] reflect – are quite heavily linked to different investment strategies,

[00:17:32] different kind of – someone who's really, really extroverted is more likely going to want

[00:17:38] a high-risk, high-frequency trading portfolio, whereas someone who's more conscientious,

[00:17:47] a bit more introverted probably wants a more longer-term investment strategy,

[00:17:50] and there's some research being done into that.

[00:17:52] So if you think about how a model can aid in that, or whether it be in kind of big enterprise negotiations

[00:17:59] that go on, I don't know, obviously BAE systems doing a negotiation with Rolls-Royce

[00:18:03] to do with something, that might take five years to negotiate, and there's 25 people involved.

[00:18:08] And if over that time you're able to get the insight of, okay, we know that – I don't know –

[00:18:15] John Smith, he actually needs a load of detail, and you know exactly what kind of information

[00:18:21] to use to influence that decision. So, yeah.

[00:18:25] Is it fair to say that – I'm looking at it from an ethical standpoint here – is it fair to say that actually

[00:18:32] your system will give a lot more accelerated, immediate perspective that some individuals could probably infer, right?

[00:18:43] Some people who are very experienced, if they've got three minutes watching someone, they can go,

[00:18:49] oh, I support you. You can begin to infer some of this.

[00:18:52] Yeah.

[00:18:53] I'm trying to think, but it's not necessarily that bad, right?

[00:18:57] Because some people can go, oh, my goodness.

[00:18:59] We aren't – yeah, we aren't kind of like opening the kind of – the box of dark secrets that people have.

[00:19:08] It's not like, oh, we know – I mean, things like credit rating, right?

[00:19:12] Like, credit rating, I understand why banks do it.

[00:19:14] But you could argue that if it's a B2B environment and you know that the seller has a really bad credit rating,

[00:19:20] you can be like, look, we've got to do this deal because you need that bonus because I know that you're in debt.

[00:19:25] That's an unethical use, right?

[00:19:26] Right, yes.

[00:19:26] Because that's private information.

[00:19:28] But actually everybody in the world can build a relationship with you.

[00:19:31] Yes.

[00:19:32] And, yeah, a human can do this already.

[00:19:35] So you think about – or I'm sure you've met numerous – the best sellers in the world.

[00:19:39] I'm not talking Wolf of Wall Street.

[00:19:42] I'm talking people who are enterprise sales leaders, people who work at Oracle, blah, blah, blah, blah, incredible chief commercial officers.

[00:19:49] You ask them what their difference is as a seller, it won't be, oh, well, I made more calls.

[00:19:54] It will be I understood the human I was dealing with and I was able to build a better relationship faster.

[00:19:59] I don't know whether they had that understanding written down, formalized or gut essentially.

[00:20:05] But gut is a whole lot of different – I've met a lot of individuals that fit that profile, Tom.

[00:20:10] And they're fantastic at interacting with you.

[00:20:12] Fantastic.

[00:20:13] Yeah.

[00:20:13] But that's because they've honed that skill.

[00:20:16] You learn over time.

[00:20:17] Yes.

[00:20:17] It's very rare you'll get an 18-, 19-year-old, 20-year-old in their first job who has the ability to just connect and influence C-suite overnight.

[00:20:28] That takes a lot of time.

[00:20:29] And, yeah, the use case in sales is certainly what we're seeing is it's not aimed at the top 1%.

[00:20:35] Because they can do it already.

[00:20:37] They're so good at – I mean, you can argue that – I think a good example of this would be in sport.

[00:20:43] Or – I'm a Liverpool fan.

[00:20:45] Jürgen Klopp, his ability to communicate with individuals.

[00:20:48] He knew exactly what to say, when to say it, and how to say it.

[00:20:52] And that's how he got the best.

[00:20:53] That's why people trusted him.

[00:20:54] And that is – sales is trust.

[00:20:56] You're trying to build trust to that individual.

[00:20:58] If we're able to give the 21-, 22-year-old new seller who's up against it and it's like, right, you've got to do 12 enterprise deals this year,

[00:21:07] the ability to go, oh, I know exactly how to live my messaging for that individual.

[00:21:11] Because they've got their MedPick, they've got their Spice, they've got their frameworks.

[00:21:15] But a framework's a framework.

[00:21:17] A framework's not a human.

[00:21:18] You're supposed to transcribe and adjust that framework to the human.

[00:21:22] What we do is – that's a bit of a cheesy phrase, but we raise the floor.

[00:21:27] We don't raise the ceiling.

[00:21:30] Does that make sense?

[00:21:30] Yes, yes.

[00:21:30] We're bringing everybody up to the same level rather than just saying the top 1% can do better.

[00:21:34] No, it's about the other 99% bringing them up and giving them the ability to be a top performer.

[00:21:38] Gotcha.

[00:21:39] What are you thinking then in terms of roadmap?

[00:21:42] Because I'd imagine to an extent a lot of this is dependent on customers coming forward to you and saying,

[00:21:48] can it do this or could I use it here?

[00:21:49] What are you thinking?

[00:21:50] Yes.

[00:21:51] So the big roadmap for us is actually human-to-machine interaction.

[00:21:57] Right.

[00:21:58] Very interesting.

[00:21:59] Human-to-human interaction will never be replaced in a large part.

[00:22:05] But actually, a lot of human-to-machine interaction is changing and is evolving.

[00:22:09] And people have seen people like 11X and these kind of technologies which allows humans to interact with machines,

[00:22:18] which is impressive.

[00:22:19] But the machine is still emotionally dumb.

[00:22:24] And essentially, communication, every time you communicate, whether it be email, voice, whatever,

[00:22:29] you're trying to do three things when you communicate.

[00:22:32] That is to speak credibly.

[00:22:33] That is to communicate credibly, communicate logically, communicate with emotion.

[00:22:38] Machines are incredibly good at the first two because, obviously, the power of machines.

[00:22:41] They have logic.

[00:22:43] They have credibility.

[00:22:46] Credibility is actually probably a bit of a blur.

[00:22:48] But actually, the emotional side, machines can't really do that.

[00:22:51] Whereas with the huge acceleration we are now seeing between voice and machine interaction,

[00:22:57] but using our model, when you speak to the machine, we can then inform the machine exactly the human they're speaking to.

[00:23:05] And that creates the personalization on a kind of back end for the machine to go,

[00:23:10] okay, I know I need to speak to you in this way.

[00:23:12] I know I need to speak to Tom in this way.

[00:23:14] Or blah, blah, blah.

[00:23:15] So we're essentially adding a layer to human to machine interaction that creates a deeper relationship,

[00:23:22] a deeper understanding, and is able to accelerate that kind of interaction.

[00:23:28] So if you gave my agent or my chat GPT or Claude, if you gave it or I want that because it's going to be better,

[00:23:37] it will then be able to formulate answers and present information in the most effective way for me to consume it.

[00:23:44] Precisely.

[00:23:45] That's very powerful.

[00:23:47] Very powerful.

[00:23:48] Because I think it was only recently at the event, I think you were saying something like,

[00:23:54] you were touching on this topic, Tom, but you were highlighting that the chat GPT or any conversational AI,

[00:24:01] it gives the same answer typically.

[00:24:03] Yeah.

[00:24:04] It doesn't matter.

[00:24:05] Yeah.

[00:24:05] In the same way to everybody.

[00:24:08] Yeah.

[00:24:08] When actually it could be way more effective if it was emotionally intelligent.

[00:24:14] Precisely.

[00:24:14] There are some really cool technologies out there.

[00:24:17] I mean, Hume AI, a quite intelligent platform based in the US that we were talking about earlier about the sentiment stuff.

[00:24:24] So they can pick up, are you getting excited or you're really low?

[00:24:27] They can pick that up and allow, they essentially have their own LLM,

[00:24:31] that's able to pick that up and then mirror and match.

[00:24:35] But actually mirror and matching is kind of like stage one of emotional intelligence.

[00:24:39] Stage two is that I know how to adjust.

[00:24:42] And copying someone doesn't always mean that's the right thing to do.

[00:24:46] It's incredible.

[00:24:47] It's amazing.

[00:24:48] I mean, for those who haven't seen it, go and check out Hume AI.

[00:24:50] It's very, very cool.

[00:24:52] They've just released their second version.

[00:24:55] But obviously what we do is, the power of both of those together,

[00:24:58] our models and their models allow that whole other level.

[00:25:02] So yeah, it's the way it's presented, the way it builds a relationship.

[00:25:06] And obviously using memory because of the memory we now have,

[00:25:10] essentially the machine will have a better memory than you and I, quite clearly.

[00:25:14] Every single conversation that we ever have is stored.

[00:25:18] And then once you engage again, it's, oh yeah, you spoke about how you were going abroad last week.

[00:25:23] How did it go?

[00:25:23] You're like, whoa, how do you remember that?

[00:25:26] But that's the kind of thing that I've always done as a human.

[00:25:29] I've always found that, I mean, one of the guys I used to work with,

[00:25:32] he was like, I know a little about a lot.

[00:25:35] It's kind of my kind of human USP.

[00:25:38] If someone says, oh, I love field hockey.

[00:25:42] I'm like, oh yeah, the GB girls team in 2012, they won gold, right?

[00:25:46] I always had little snippets of everything.

[00:25:49] And obviously machines can do that at a thousand X level.

[00:25:53] But is that even relevant?

[00:25:55] And obviously using personalization, you can use that.

[00:25:58] I think it's really important to note that since the 1940s,

[00:26:02] obviously when computers were first, or artificial intelligence,

[00:26:05] you argue with Alan Turing being the father of AI back in the 1940s to today,

[00:26:12] the only way that humans can interact with machines was hard text.

[00:26:16] You input a button, you input a format or a sum, whatever it is,

[00:26:22] and it returns that exact same format.

[00:26:25] But because of the acceleration, especially the last six months,

[00:26:29] more and more human to machine interaction is going to be voice-based.

[00:26:34] And we were speaking to a company like the Backender last week

[00:26:37] who specialize in real-time avatars, human-like avatars.

[00:26:41] It's incredible.

[00:26:42] They're quite early on.

[00:26:44] Really great business.

[00:26:46] Second-time founders, they just sold their first business to Meta.

[00:26:49] So they're pretty good.

[00:26:51] And they've developed a way in which you wouldn't know

[00:26:54] if you were talking to a human or to an avatar.

[00:26:56] It's that lifelike.

[00:26:58] But obviously, so they're very good at the body and the face and the voice.

[00:27:03] The brain bit, they allow other companies to do it,

[00:27:05] whereas obviously we have that brain.

[00:27:07] So for us, it's about, yeah, it's whenever you...

[00:27:12] So where we're seeing a lot of interest in this part of our business

[00:27:18] is coaching, so kind of human coaching.

[00:27:21] And for those who find L&D and ed tech interesting,

[00:27:25] and like I said, a lot of people don't,

[00:27:26] it's actually a trillion-dollar industry.

[00:27:28] It's absolutely huge.

[00:27:30] Last year alone, globally, $534 billion was spent

[00:27:35] on learning development programs.

[00:27:37] And that's professional.

[00:27:39] That's not schools and colleges.

[00:27:40] That is doing a course internally.

[00:27:43] And that only goes to probably the top 1% or 2% of organizations

[00:27:46] because you don't...

[00:27:47] The only way to deliver this traditionally is by human coaching.

[00:27:50] And humans aren't scalable.

[00:27:52] I mean, Elon Musk might challenge that,

[00:27:54] but humans aren't scalable.

[00:27:57] So what we're doing is we're essentially creating virtual versions

[00:28:01] and emotionally intelligent kind of virtual versions of coaches

[00:28:04] to allow not just organizations to get more information,

[00:28:07] but allow a lot of these individual coaches, L&D managers,

[00:28:10] to scale and monetize themselves.

[00:28:13] Can you think about...

[00:28:15] The world of coaching has an incredibly long tail.

[00:28:17] You have a handful of big consultancies, obviously.

[00:28:20] You've got the big...

[00:28:21] Actually, KPMG do coaching, et cetera, et cetera,

[00:28:23] but you've got the likes of Mind Gym, Arcadia, Black Isle Group,

[00:28:28] who do a lot of programs.

[00:28:29] But actually, a huge amount is individuals.

[00:28:33] So I am a...

[00:28:33] So, yeah, there's numerous people you could kind of find who are leadership coaches.

[00:28:37] They work with C3, blah, blah, blah, blah.

[00:28:39] But they only have enough hours in the day to do that coaching for those people.

[00:28:42] A whole organization would love to get access to that coach.

[00:28:46] That coach isn't available.

[00:28:48] And also, can they afford £500 an hour for the administrator or the storeroom guy

[00:28:54] to get leadership from Jordan Peterson?

[00:28:57] It's never going to happen.

[00:28:58] But what they can do is go, well, why don't we create a virtual version?

[00:29:01] Yes, it's not replacing the human.

[00:29:03] What it does, it allows them to scale that human.

[00:29:05] And we can obviously train it.

[00:29:06] We can fine-tune it on their knowledge and their information base

[00:29:09] and their theory and their technique.

[00:29:10] We then, in partnership with one of these avatar organizations,

[00:29:13] can create a human-like avatar.

[00:29:16] Whether you do a phone call with them...

[00:29:17] I mean, think about walking the dog in the morning

[00:29:19] and basically being like, I've got a big meeting this afternoon.

[00:29:22] I'm pitching to my manager.

[00:29:24] It's my first chance to kind of speak to middle management.

[00:29:27] How am I going to cope with this?

[00:29:28] You would love to be able to speak to Jordan Peterson around that, right?

[00:29:31] Not just him, but some kind of mindset coach, leadership coach.

[00:29:36] He's not going to answer the phone to you.

[00:29:37] But if it's a virtual version of him that has all the knowledge,

[00:29:39] has delivery, is able to know you,

[00:29:42] because obviously using our models, it personalizes,

[00:29:44] and talks you through that.

[00:29:45] Okay, well, what are you trying to get out of this meeting?

[00:29:48] What are your kind of goals?

[00:29:49] How has it made you feel?

[00:29:51] Do you think there's anything in your personal life

[00:29:52] that you can use to embrace that meeting?

[00:29:55] All that kind of stuff.

[00:29:57] And it's scalable across the whole organization instantly,

[00:29:59] in every language.

[00:30:00] Because we can train it in English,

[00:30:02] because of the advancements in LLMs,

[00:30:04] it could be in Cantonese,

[00:30:05] it could be in Portuguese,

[00:30:07] it could be in French.

[00:30:09] 60, 70 languages instantly overnight.

[00:30:13] And every employee has that ability

[00:30:15] to engage with that coach within the business.

[00:30:18] It allows the coaches to monetize.

[00:30:20] It allows the individuals to have access to this information.

[00:30:23] And it's a win-win for all those involved.

[00:30:25] Can I take it in a slightly different direction there,

[00:30:27] Tom?

[00:30:27] Because the ramifications are just astonishing

[00:30:29] for everybody.

[00:30:31] What's your message or recommendation

[00:30:34] to those running conversational AI systems

[00:30:36] at the moment?

[00:30:37] We've got a lot of listeners

[00:30:38] who are responsible for conversational AI

[00:30:40] for telecom company, bank company.

[00:30:44] And increasingly,

[00:30:44] they'll either be looking at some degree of audio

[00:30:47] or they're already using audio.

[00:30:49] So what's your recommendation to them?

[00:30:52] Remember the human.

[00:30:53] I think it's the biggest thing.

[00:30:55] You aren't building a conversational AI

[00:30:57] to chat to another conversational AI.

[00:30:59] There is a human involved in this process.

[00:31:02] You need to humanize it.

[00:31:04] And I think a lot of organizations,

[00:31:05] especially,

[00:31:06] and obviously there's some heavily regulated areas

[00:31:08] which are using conversational AI,

[00:31:10] whether it be financial complaints,

[00:31:12] obviously everything.

[00:31:13] They have incredibly strict workflows

[00:31:15] because they have to.

[00:31:17] But what about the human on the other side?

[00:31:20] At the end of the day,

[00:31:21] you're building all this technology

[00:31:22] to engage with a human.

[00:31:24] But all you're thinking about at the moment

[00:31:25] is potentially about what the machine is doing,

[00:31:28] not about how you get the human to get on site.

[00:31:31] And that's the big leap forward.

[00:31:33] They had written information recently.

[00:31:35] They believe that 85% of individual

[00:31:39] to business communication by 2030

[00:31:41] will be human to machine.

[00:31:42] And whether that be through text,

[00:31:44] that is conversational AI,

[00:31:47] which is a great time in you.

[00:31:48] And I can understand why you're in this space.

[00:31:51] But that's where it's going.

[00:31:54] But it's the human side.

[00:31:56] Remember the human.

[00:31:57] And if you can get your conversational AI

[00:32:00] to build a relationship

[00:32:01] and build trust with the human,

[00:32:04] you're going to have more success.

[00:32:06] And just make it better

[00:32:07] at communicating with individuals, right?

[00:32:10] Yeah, 100%.

[00:32:11] I mean, I received one of these virtual sales calls

[00:32:15] a couple of weeks ago about Irish whiskey.

[00:32:18] And within 10...

[00:32:20] Well, no, within that.

[00:32:20] Within a second, I knew it was fake.

[00:32:23] And then I asked it a question.

[00:32:24] It just spazzed out.

[00:32:25] It was like, oh, you can't cope.

[00:32:27] I think it was like, hi, Tom,

[00:32:29] have you been interested in Irish peat whiskey?

[00:32:32] And I was like, look, can I just stop you there?

[00:32:33] Would you rather have a fight

[00:32:36] with one horse-sized duck

[00:32:38] or 100 duck-sized horses?

[00:32:41] Because I knew it was a machine.

[00:32:42] And it did not know what to do.

[00:32:44] It just went,

[00:32:45] it just went,

[00:32:46] that isn't something we've thought about.

[00:32:48] But we can get a case of Irish whiskey

[00:32:49] It just literally just...

[00:32:53] A human would go,

[00:32:55] that's really funny.

[00:32:56] Actually, I think I would probably...

[00:32:58] But they could probably spin that

[00:32:59] into something and in doing so...

[00:33:02] Yeah, it's human you

[00:33:02] and understand you.

[00:33:05] And as a salesperson

[00:33:07] or as a business,

[00:33:08] obviously as a founder of a business,

[00:33:10] I'm always trying to build relationships.

[00:33:11] If someone did that to me,

[00:33:12] I'll be like, that's memorable.

[00:33:14] And I am going to...

[00:33:15] And you can kind of go with it

[00:33:17] and go, how about I throw one back at you?

[00:33:19] And you can...

[00:33:19] And you build that relationship

[00:33:20] with an individual.

[00:33:22] So yeah, remember the human

[00:33:24] because that's why we're here.

[00:33:26] What's the best way

[00:33:27] of those individuals

[00:33:28] doing something with you, right?

[00:33:29] Moving things on.

[00:33:31] You've got the API available.

[00:33:33] Should they be coming to your website,

[00:33:34] getkoyos.ai,

[00:33:36] signing up,

[00:33:38] interacting with you?

[00:33:38] What's the call to action, Tom?

[00:33:41] Reach out to me on LinkedIn

[00:33:42] is a really good way.

[00:33:43] I mean, I've just kind of mentioned,

[00:33:45] obviously,

[00:33:45] you kind of heard the podcast

[00:33:47] and we'll kind of see

[00:33:49] the newsletters, et cetera.

[00:33:50] Because as you can imagine,

[00:33:51] I get a lot...

[00:33:52] Everyone gets a lot of spam

[00:33:53] in their inboxes nowadays,

[00:33:54] so it's kind of hard to filter.

[00:33:57] Yeah, go to our website.

[00:33:58] We are actually building

[00:33:59] a new website.

[00:33:59] They'll be coming out soon

[00:34:00] with a kind of holding page.

[00:34:01] But there is a wait list there.

[00:34:04] But yeah, reach out to me.

[00:34:06] We're always incredibly keen

[00:34:08] to hear about different use cases.

[00:34:10] We are kind of fundamentally

[00:34:13] a white-label technology.

[00:34:15] So we can either build

[00:34:16] your conversational AI for you,

[00:34:18] we can enhance it,

[00:34:19] and we can take that for ownership.

[00:34:20] I mean, there's a couple

[00:34:21] of organizations

[00:34:21] where they have a platform

[00:34:22] and they have some kind

[00:34:24] of conversational AI,

[00:34:25] but they're actually going

[00:34:26] to outsource that to us.

[00:34:27] So we all own that part of it.

[00:34:29] I mean, we've got

[00:34:29] an incredibly talented team

[00:34:30] internally that makes

[00:34:32] these agents phenomenally well.

[00:34:36] So we can do that for you.

[00:34:38] Or if you say,

[00:34:38] look, I've already got

[00:34:39] my conversation AI.

[00:34:40] We're experimenting with voice.

[00:34:43] I want to enhance it

[00:34:44] and come to us

[00:34:45] with a use case,

[00:34:46] then let's chat.

[00:34:47] Brilliant, brilliant.

[00:34:48] The final question

[00:34:49] I've got for you, Tom,

[00:34:50] is what's your favorite vacation?

[00:34:55] I've only been to Bordeaux once

[00:34:57] and it was amazing.

[00:34:59] I spent five days in Bordeaux

[00:35:00] before I had kids

[00:35:01] and my wife and I

[00:35:03] just had the most incredible time.

[00:35:05] It's cheaper than you realize.

[00:35:07] Beer's expensive

[00:35:08] because they don't want you

[00:35:09] to buy beer.

[00:35:11] The wine is obviously,

[00:35:12] it is a wine town.

[00:35:14] It's got amazing history.

[00:35:15] They love the British.

[00:35:16] All of the pubs

[00:35:17] are named after British icons.

[00:35:19] So Houses of Parliament,

[00:35:21] the Westminster,

[00:35:22] the Trafalgar,

[00:35:23] it's all British named pubs.

[00:35:25] They have the Bordeaux Wine School

[00:35:29] in the middle of their square

[00:35:30] which every afternoon,

[00:35:32] whatever wines

[00:35:32] they were teaching with that day,

[00:35:34] they just sell the rest of them

[00:35:35] because they're like,

[00:35:36] oh, we've got half a bottle

[00:35:37] of this left.

[00:35:37] It's three euros a glass

[00:35:39] and it could be

[00:35:39] like a 2,000 euro

[00:35:42] kind of like Premier Grand Cru

[00:35:43] and they're like,

[00:35:44] well, it's open now.

[00:35:46] You might as well.

[00:35:46] Yes.

[00:35:47] It's just amazing.

[00:35:48] The flight's about an hour.

[00:35:51] Wow.

[00:35:51] Really easy to get to.

[00:35:52] Yeah.

[00:35:53] So Bordeaux, I think.

[00:35:54] There we go.

[00:35:55] That was my left field question

[00:35:57] for you then, Tom.

[00:35:58] Good idea.

[00:35:59] Good idea.

[00:36:00] Well, can I say

[00:36:01] thank you so much

[00:36:02] for taking the time, Tom.

[00:36:03] My pleasure.

[00:36:04] Thank you so much

[00:36:04] for having me.