S1. Ep#6 - Neurodiversity, The Wizard of Oz and solving big business problems with AI. With Marton Gaspar, product whisperer

Unknown Speaker : 0:00

Hello and welcome to conversations worth millions. The podcast all about the business of conversational AI. We talk with leading experts in voice chat messaging and emerging technology, about the strategies and practical applications for business. In this episode, I talked to Martin Gasper, who has a huge amount of experience in voice chat and a background in psychology. We talk about continuous discovery, neuro diversity, strategies for building better bots and why the Wizard of Oz is your new best friend. He also answers the question, how do you create a conversation worth millions? Stay tuned. Enjoy. Welcome everyone to the latest conversations worth millions podcast. Today I am talking to a really interesting guest Marco Gasper, who has a wealth of experience across product, chat voice and a background in psychology. So really looking forward to this chat and getting Martin's take on where things are heading next. Welcome, Martin.

Unknown Speaker : 1:02

Hi, everyone. Thanks for having me.

Unknown Speaker : 1:04

Martin. The first question I love asking our guests on the show is how did you get into this whole space this emerging technology voice chat conversational AI

Unknown Speaker : 1:15

piece in 2016. When the first boom started, I was working as an AI researcher. And I started working on the largest chatbot project at the time for RBS with IBM and mindshare. We have to define what personality should the bank banks chatbot be? How do people want to talk to this new medium? And how should that conversation be like?

Unknown Speaker : 1:38

Was that a challenge to work within those constraints?

Unknown Speaker : 1:40

It was a massive challenge, because at the time no one talked about this was really, really at the beginning, people have no idea of how they want to communicate with AI, they didn't know how they want to communicate to a bank. What tone Do you strive for? What what features Do you give, we had to run a series of experiments and tests to figure out what even the right direction is, and then coming up with something called decent from that.

Unknown Speaker : 2:04

And since that early stage, you know, it feels like a long time ago, that's only like four or five years ago, right? So what was the path from there to now

Unknown Speaker : 2:11

so I've done quite a few, both projects in the space based on one of the fork leadership pieces we did a major company built a whole department on it. And I started to get to know the experts and got to know the field a bit. So I realised I have a wealth of connections and experience quite early on, which is quite powerful. So I created the chatbot agency. And that's slowly merged into a data science agency. And then I went on to work as a product consultant in the space as well.

Unknown Speaker : 2:45

And understand you've got a background in psychology, which must make the science and the emotion kind of interesting for you in terms of products.

Unknown Speaker : 2:53

Yeah, it's fascinating. I think psychology underpins everything from how you manage stakeholders to how you interview users to do how do you create that empathy within your company and within your user base as well? And how can you translate that into a digital experience? It's been hugely helpful.

Unknown Speaker : 3:11

So you've got some great experience over the last few years probably seen a lot of it from the corporate side, with a business perspective, from product and also from psychology. So taking the 30,000 foot view, where are we at the moment when this whole kind of conversational chat space Do you think

Unknown Speaker : 3:25

we are really at the beginning of the journey? If you were to plot both on Gartner's hype cycle, I would say that that AI bots are still at just the throat of disillusion. Actually, there's quite a lot of business problems that are very simple to solve. And using these they're really simple to solve. And we can already do some fantastic job with having scripted conversations and iterating really quickly to find out really user needs. Having really complex AI, I think we really far away from it, we don't have the layers of of sarcasm, we don't have the layers of culture in there. And, frankly, if it would work really, really well. All the major assistance on our phones who just understands us. And that's just not the case you can't have a proper conversation with I haven't seen any boss apart from Mitsuko that you can have a proper conversation with it's Yeah, we're a bit far away from from being seamless and fantastic. But we already see a lot of value from solving the more simple straightforward, narrow use cases. Sure. So do you think that

Unknown Speaker : 4:44

it's been overhyped? A little bit, so maybe at the beginning, you know, we've been involved in the same space for about the same amount of time. And there was certainly that kind of view at the beginning that, you know, AI is going to take over the world and its general AI is just around the corner. And that hasn't proved to be the case.

Unknown Speaker : 4:59

I would like to make a clear distinction between AI powered bots, that the conversation is driven through AI and the bots that are scripted and have occasional AI functions in there, the ones that are fully AI powered NLP based, and there's always a drop off after each conversation, I think where we're at the toddler level as most, whereas in the bots that are more task oriented, the ones that that have really clearly defined flow and have additional features that can use AI if you want those, I think we're at a good team, maybe that there are some really, really solid business cases, and I've done some stuff person as well, that that really changed my view on how useful some of the more simple applications can be. If you really put product discovery and users first, I think that's one of the key issues that I see not happening.

Unknown Speaker : 5:58

Let me talk a bit more about those use cases. Because you know, when we talk to some of our clients and customers, sometimes that education piece and expectation that, you know, ai conversation can do everything. And then when you actually get to reality is okay, well, the good thing to do would be to take this step first what what are those steps are the use cases, as far as using

Unknown Speaker : 6:17

what I always try to advise people to do, as they need to really understand the users first. Right, rather than just jumping into your technology solution, you really need to understand the whole landscape and what the problems that you want to solve. So people often go too broad, and they just want to wrap their whole experience into a conversation interface, which will not work. And they often want to solve challenges that are not even tied into their business outcomes. So why real life to have is figured out what people actually need? What, what are the use cases that will work in in your in your instance of what you translate well into your bot, and then come up with a really narrow scope, and start iterating? Day by day? Don't Don't start with coding first, don't start going too broad, don't choose the technology vendor first. You just have to develop, exploring the problem, and then going to two technologies.

Unknown Speaker : 7:24

So in terms of when you're talking to potential product owners, or clients or customers of yours, how would you approach that?

Unknown Speaker : 7:31

So there are three different types of customers I have one of them are just thinking about bots, that is normally a strategic conversation, how can this tie into your to your outcomes? What do you really want and what can work in a conversational interface. There's a next phase when they're already building a team. And they already have some ideas about what they want to do. That will be normally have to go back to to the beginning and just test what actually works. I'll go into a sec how that works. And I also have customers, what are the odds on the journey built a team and the boat is just not performing? And it's troubleshooting. And quite often, it's normally a total rewrite of the whole system. And we have to restart. But one of the ways that I really like to do is just visit of was it just put a person behind the boat and have conversation with actual people and see how your information architecture how your conversation design goes. Check your assumptions do people even want to do what you want to do? You have an agenda, people may not think that's what they want to do within the bot. They may not understand it, there's a lot of things you need to iron out from, from language through feature through information architecture. And once you've got the basic stuff out, like how do people want to talk to what and what do they want to get out of it. That's when you can jump to the next next steps when you when you when you build a prototype from it, and you keep iterating on it. And testing with qualitative and quantitative methods joined together.

Unknown Speaker : 9:11

You heard it here first on conversations worth millions all AI bots says bolson are Wizard of Oz.

Unknown Speaker : 9:20

It's it's the Mechanical Turk all over again. Right?

Unknown Speaker : 9:23

Yeah, this ultralight said. But yes, so if you Wizard of Oz that you can get a really accurate view of what people really want. And that's that that's fantastic. Instead of you jumping in with assumptions or starting to build robust solutions, which no one will use, I've seen it so many times over and over again.

Unknown Speaker : 9:44

You know, if you haven't really thought through your use cases and what the bot or the voice skill or the automation experience is for what the US gets out of it, you're kind of doomed to failure right from the beginning.

Unknown Speaker : 9:54

Oh, absolutely. And I'm sure you see that quite often as well. Do you see See voice and chat as the same thing? Or do you see them as very as part of the bigger conversational space? Are they just completely different? Well, good question. They I think they're completely different. They they're used differently. I understand there's a lot of similarities if you're communicating with an interface, but that's what we do with our keyboards as well. And I think they're wildly different. You have to design for them really differently. There are really different problems they can solve. There are overlaps. But yeah, I think they're very, very different.

Unknown Speaker : 10:34

So where is the biggest ROI at the moment within the conversational space,

Unknown Speaker : 10:38

internal stuff can be hugely powerful. optimising workflows, getting internal things done, repeated stuff that usually you can just automate away really, really easily with a reminder on the phone, for example, like simple solutions like that, especially at large organisations can be massive. And also in customer service, scaling, customer engagement. So some stuff like asking for user feedback at scale can be fantastic. I create continuous discovery cycles by getting people to do user interviews, scheduling automatically by the bot, fantastic stuff. Also, people want to get answers and what to get simple things done. companies who you can change the menu, booking 24 seven, or get really simple answers out of about 24 seven, we'll just provide a better experience, your users will love you more, they will churn you less. These are such simple things we can implement within weeks. And they are there forever. And they just keep giving and giving to the company and to the users as well.

Unknown Speaker : 11:52

But how did you create those experiences? For a practical level? Are you using NLP training data? are you how are you doing that continuous learning cycle to get to the real gold?

Unknown Speaker : 12:04

You can't I think one of the issues I see is people try to separate it and they want to build a bot. Whereas you have an organisational strategy that has goals, you want to increase engagement, you want to know more about users, you want to reduce churn, whatever that may be. And there's there's a number of different channels that that you're exploiting. First, you have to understand what channels you have, and how do people communicate with you on those channels? And what are the pain points there? And then you have to see, is there an opportunity for voice or or chat to help you somewhere and fit into the larger strategy rather than let's just create a bot in silo? And let's hope it works with everything together.

Unknown Speaker : 12:51

And where do you think voice will be in the next couple of years? Do you think

Unknown Speaker : 12:54

I'm still yet to see fantastic use cases to emerge, there are some really great stuff like for example, in the gaming space that does some some really good companies like lovebirds doing fantastic games, there's the new payment going through Amazon Alexa, that you can pay for your few those starting to be really good features. I think this still yet to see where all of this platform can go. I really hope that I would really want to tell you that in the next two years, there's going to be a big major boom, pot. I personally can't see it just yet. payments, subscriptions Those are probably where where we're gonna see a lot more activity because it's tied to money and it can solve your pain point being the voice could be really quite quite useful as as an as an identification as well. But whether you carry it whether you apply it stuff question,

Unknown Speaker : 13:56

are you going down the route of voice as utilities, rather than a nice to have for

Unknown Speaker : 14:01

years and years now we're all trying to achieve digital cohesion right? When you have one personalised AI that does everything for you that that controls your chatbots your voice, your health vaikka puts everything together and just makes your life seamless and easy. AR VR tell you what to buy in the shop creates your diet fits joins everything together. It would be amazing. I'd love that. I just don't think that within two years we are realistically there. But once we are in this utopia then everyone has their own personalised AI that does everything for them, including telling you that you need to have a blonde because your blood sugar is low through you need to do your taxes within that ecosystem. Voice is one of the key connectors to everything but in that ecosystem voice so to understand my accent

Unknown Speaker : 14:57

What is your favourite skill? or chat experience.

Unknown Speaker : 15:03

What I use most is your MDS medical diagnosis. Right, especially in terms of Corona, they're just doing a great great stuff, literally you can get diagnosis to to minor things, or if not, you can get information about the conditions is not gonna tell you what some video is going to give you a few options. And you can read up about it. And it will also tell you in a non threatening way, when it's time to see a GP. I think they're just hugely powerful. And especially Just think about it in in the less developed areas. You have an access to a digital doctor. That's just incredible.

Unknown Speaker : 15:42

You have a brilliant job title, which is product whisperer. What does what does that mean?

Unknown Speaker : 15:49

Yeah, so I have I have a product leadership consultancy. And I'm focusing on three things, composition interfaces, as you probably gathered by now. product, scaling product teams, hiring product teams, and then helping them to work more efficiently. And I also focus on neurodiversity. So I am autistic. I have ADHD and dyspraxia. And I started creating videos on LinkedIn to help people understand my journey. And hopefully there'll be able, there'll be able to see some similarities, because everyone's affected by by diversity, and neuro diversity in one way, shape, or form. Either you're dyslexic, or someone else who you work with is and people just don't really know how to deal with each other. And people who are neurodiverse don't, often a lot of the times they just don't come out. Or even if they they say that look, I am autistic people don't know what to do with them. So I think it's hugely important for us to to try to educate people because there's nothing wrong with being autistic. What's wrong is being mistreated because you're autistic due to lack of knowledge. So I, I dedicated a large part of my professional career, to educate people and get the word out on on the actual benefits of neurodiversity in productive teams. I want to make a financial argument that that CEOs and business leaders understand that. If you want more ideas and better ideas, which we all know is the key to innovation, you need to hire neurodiverse talent in your product team. neuro diverse people's brains work differently, they will bound to come up with different ideas and different perspectives. And if we buy into the idea that product teams are trying to create empathy with their users, and also within their business to solve difficult challenges, then you do need different perspectives. And diversity helps

Unknown Speaker : 18:06

teams with more diverse input, get better outputs, right.

Unknown Speaker : 18:11

Yeah, exactly.

Unknown Speaker : 18:13

So are robots going to kill us?

Unknown Speaker : 18:16

I hope not. I don't think so. No. So I basically, AI is already powering us, right? All the algorithms that are logical, that are creating the engagements that that we just subject ourselves to using social media and all the different platforms, products and games that are already in there, they already controlling us more than we know. Countries being grown and opinions being changed by all of these aggressive campaigns that that can happen there. So AI is already controlling us. I personally think that a better way of looking at this is not us versus them. It's how can we work together? How can we make sure that AI and people enable each other and and get better things out of each other?

Unknown Speaker : 19:08

The Guardian had an op ed, done by an AI, and it was pretty damn good. It's been lightly edited by God in subs, it's probably the best piece of AI written content that I've seen. Scary is the wrong word. But it's it's getting quite professional.

Unknown Speaker : 19:24

I've seen a couple of natural language generation and it's gone a long way in the last three, four years. I must say,

Unknown Speaker : 19:33

go back to your robot answer. I feel a little bit better now. But I'm still a little bit scared.

Unknown Speaker : 19:38

Okay, is there anything I can do?

Unknown Speaker : 19:40

From your psychology background? Maybe we can have a chat afterwards. If a bot or skill or experience could do one thing, and one thing only what should it do?

Unknown Speaker : 19:50

I mean, I'd love if someone would take notes at my meetings that we do, like actual good notes from different people, properly organised higher. Bullet pointed, I'd love that.

Unknown Speaker : 20:02

What's the biggest challenge you come across when talking to clients or customers,

Unknown Speaker : 20:07

I always start with the education piece that there's this inflated expectations that everything is just gonna work the first time, you just don't get it right bang, and it's gonna be fantastic and more often is not. And like, we all have to take them on this journey, that it's going to be an iterative process where we are going to build and fail, and then at the end is going to be fantastic if we get there, but it's going to be a lot of iterations. And also getting users is still pretty damn hard. It I think that's one of the most bots are being killed by either not generating value or not being able to get users

Unknown Speaker : 20:49

to fundamental requirements. Yeah. Okay. What advice would you have someone coming into this whole AI space,

Unknown Speaker : 20:58

get a mentor or a coach, that's the single biggest differentiator that helped elevate my career. Good. Five years ago, I started having mentoring, coaching and it just transformed everything really my whole life. But also, you need to attend meetups like yours and mine, get into the community and talk to a lot of people. And also you just need to get stuck in and build a boat for yourself and and make make it something that people actually like and get some users. There's no better learning them from doing

Unknown Speaker : 21:35

sage advice, Martin sage advice. three tips for a brilliant conversational experience.

Unknown Speaker : 21:43

What are they? Interesting, so it needs to, it needs to be succinct, to the point. And you have to start with value. So people need to understand why they're talking to what the next one, I think it should be very, very easy. So if you can use buttons, any sort of visual stuff that fits in with a bot, or if you're designing for voice, it should be just a one or two or a yes or no. Make it easy for for people to interact and get their thing sorted, the quickest and easiest way they it can. People don't talk to both because they want to talk to both when they when they need to get something done. They want the outcome, easily accessible. well understood, and then quickly attained.

Unknown Speaker : 22:40

And the most difficult. question I'm going to ask is when we ask all our guests, how do you create a conversation worth millions?

Unknown Speaker : 22:50

Okay, so you need to keep your users in mind and giving them 24. Seven access to things they want to achieve things they shouldn't be waiting for. Is can can be the difference between between providing an excellent service creating a great connection or just simply being left behind.

Unknown Speaker : 23:14

Fascinating. Well, look, it's been great talking to you. Thanks for being very open, and giving us the benefit of all your expertise. Thanks again. Much appreciated you coming on the show. Martin Gaspar,

Unknown Speaker : 23:27

thank you so much. Thanks for having me. Such a good conversation.

Unknown Speaker : 23:32

Thanks for listening to the latest conversations worth millions podcast. For more information on emerging technology chat voice and conversational AI, head over to synthetic agency.com