Speaker 1 00:00:05 Hey there, thoughtful listener. Are you looking for introductions to partners, investors, influencers and clients? Well, I've had private conversations with over 2000 leaders asking them where their best business comes from. I've got a free video you can watch with no opt in required, where I'll share the exact steps necessary to be 100% inbound in your industry over the next 6 to 8 months, with no spam, no ads, and no sales. What I teach has worked for me for over 15 years and has helped me create eight figures in revenue for my own companies. Just head to up my influence comm and watch my free class on how to create endless high ticket sales appointments. Also, don't forget the thoughtful entrepreneur is always looking for great guests. Go to up my influence. Com and click on podcast. I'd love to have you. With us right now, it's Carrie Robinson. Carrie, you are the vice president of conversational AI with Waterfield Technologies. Your website is Waterfield Telecom. Carries a pleasure to have you.

Speaker 2 00:01:20 Thanks to us. Pleasure to be here.

Speaker 1 00:01:22 Well, I would love to hear more about the work that you are doing, the work that Waterfield is doing in the space of AI, because there's a lot of cool innovations happening right now.

Speaker 2 00:01:33 Yeah, a lot has changed. I like to say I got into AI when I was 9 or 10 years old. That was quite a long time ago now, a few decades, and been through quite a few cycles and big hype cycles in AI. And yeah, the last two years has been the biggest of those. But but the most meaningful. Finally, after 20, 30 years of being involved in this tech, we have AI that that really has a form of intelligence that we can really leverage. It's not human intelligence, exactly, but it's real intelligence. And it's exciting putting it to work.

Speaker 1 00:02:05 For someone that, you know, obviously this is what you do day in, day out. I think obviously we see the main tech headlines, we see, you know, obviously the tools that are very close to us, you know, ChatGPT and Gemini and so forth.

Speaker 1 00:02:22 but what are you most excited about that's been taking place over the past 6 to 12 months that maybe I, maybe we don't understand the true impact of this, this bubbling under this groundswell that's been happening.

Speaker 2 00:02:38 Yeah. I think I was at a, Amazon Web Services, leadership summit in London. And I think it really crystallized for me, because people are talking about this as just another tool and an important tool and, and very powerful. But but when I talked to some of the keynote presenters, they're about how are they discussing it? How are they managing up there? Like I'm just telling them it's another tool that that we can add to the repertoire. And I think that totally misses it. I think this is really, really something quite different. Like I said earlier on, before now, I mean, I've been working in AI or what we used to call AI for for decades. And and it was a slog, you know, we had to, to really work through and, and write down rules and scripts for everything that we wanted a system to do.

Speaker 2 00:03:23 And we had limiting limited machine learning capabilities like speech recognition and, and vision and, and those things that will say people like you bought something like this and and recommender engines and but they were they were hard yards. Whereas nowadays we've got AI models that can hold an actual conversation in chat and now in invoice with the new streaming models, it can coat, it can summarize, it can translate, and critically it can reason when we give it really, really quite tough math and science problems with the white train. These models can solve for them and that that isn't just a new tool. If you look around, if you look around, everything that we've created as human beings, we created using intelligence. That's what means that we cage the tiger rather than the tiger eats us. and having that on tap, having that with, with almost unlimited scalability and abundance, is going to change, of course, our working lives, but also our personal lives and our society. beyond recognition, I think over the next ten years or so.

Speaker 3 00:04:29 Yeah.

Speaker 1 00:04:29 So Waterfield, tech, works with contact center solutions or call centers, which, you know, we think about over overall operational requirements that some of these larger companies have. these are just huge operations and staffing, and there's obviously a lot of issues that go along with that. there is huge opportunity here because I think, you know, a lot of right now what we outsource to people likely could be, better resourced to AI technologies. And so I would imagine we're going to probably see some pretty big movements in this space over the next year or two. Do you mind maybe giving us a peek of, you know, kind of what Waterfield is, is doing right now?

Speaker 2 00:05:25 Oh, yeah. And, you know, there was a lot of hesitance because as we talked about earlier on, there's been lots of lots of hype, lots of false dawns in terms of the capability of AI. but I think a real watershed moment was when Klarna, a financial services company that handles, many, many millions of interactions, very, very large customer services team dealing with financial services with sometimes, you know, quite, individuals with with financial difficulties, you know, it's like almost like the hardest use case you can imagine in customer services to automate.

Speaker 2 00:06:02 And and they work very, very closely with OpenAI. I wanted their kind of lead partners, and they, released a case study a few months back where they said they had automated 70% of the interactions that previously handled by customer service agents. And critically, the customers were happier with the interactions they had with the AI than than they were with the interactions they had with humans and more less likely to call back, afterwards. So that was a real, I think, watershed moment for the capability of AI, in, in the contact center. And not a surprise to me. I think a lot of, a lot of people were surprised when that came out. But but when you really work closely with the tech and you realize what it's capable of, I think you also realize that contact center is the perfect place for it because it is intelligent. It is capable of doing some really quite amazing things. But it's not magic. It has its limitations. And the thing. Thing is, contact centers are literally built to higher junior intelligences.

Speaker 2 00:07:07 There's there's quite a transient workforce, a young workforce or part time workforce. And they hire these people in, they give them a few weeks training and then they get them on the phones and they have managers listening in. They have technology listening in. There's lots of support around them. If there's any problems, they can escalate to a more senior member of the team. And so it's the perfect environment to deploy generative AI because you've got lots of humans watching what's going on and picking up the pieces. If the junior agents who we used to use or now the generative AI chatbots and voice bots that we can use make a mistake. So you've got the perfect environment to harness intelligence, but have all the safe net around it in case something goes wrong.

Speaker 1 00:07:50 so who would be examples of who Waterfield is working with right now? And I'm just curious that, you know, if there's someone listening right now that, you know, they're in that space, you know, they're involved in the, you know, kind of that customer experience.

Speaker 1 00:08:08 you know, where Waterfield fits in here and exactly what that ends up looking like.

Speaker 2 00:08:14 Yeah, a really interesting question, because this is the there's going to be quite a lot of shaking up and thinking change in this area because you've got the likes of Waterfield Tech, which is a contact center, customer service focused integrator. So we resell contact center systems and we integrate them with with the the business systems. And then we integrate AI into that and then we can manage the whole thing. so you've got integrators like us, you've got the contact center vendors like Genesis and Avaya and Cisco, and the newer ones like Five Nines and Twilio and Amazon have their own. Microsoft just launched their own contact center. and then you've got the bpos the business process Outsourcers who essentially take all of that and then just sell it to businesses by the minute So, so sell, sell customer services by the minute. And the reality is that all three of these. These, partners have, have, have for many years, coexisted quite happily.

Speaker 2 00:09:08 But now with AI, it's kind of changed it all because the contact center vendors are trying to bring AI in. The AI vendors are trying to bring contact centers in. the Bpos realized that that they don't really need to hire humans anymore. They can do this with AI. So so the Chase's the space is really changing very rapidly, and there's lots of technological changes, but also lots of economic changes, as a result of that. So it's really kind of tough to navigate at the moment. and that's kind of the way we try to, describe the situation to people is that this is a journey that that we're all going on. And I literally like we most of us only really learned about this technology in the last 18 months or so. we were tracking it for 2 or 3 years. but I think, I think even the most, sort of those deepest into the research have been quite surprised at how capable these models have become as we scaled up their sizes. So it's a journey, and I think the key thing is not to wait because the, the impact of, scalable, cheap, intelligence that that never goes on holiday, never goes sic sticks to the rules if you do it the right way and program it the right way is just, just so powerful and and obviously, you know, we can see the power of that already in the contact center with, with with the impact of the likes of Klarna.

Speaker 2 00:10:31 we're deploying chatbots on people's websites. We're deploying voice bots to replace humans in the contact center. And we're creating Copilot. So, so, so AI that can can ride shotgun with human agents and, and help them make sure they've got the right documentation, the right information to help customers. so there's lots of different ways of deploying the technology. The technology is changing very rapidly. And so the key thing is, is get started, get started somewhere at a minimum. You and your team need to be using ChatGPT daily, or you might choose one of the other models. Because. Because only by using this tech do you start to get a real feel for what it's capable of. and my big tip is, is, don't just ask it questions about stuff that you don't really know very much about, actually use it to have a discussion in areas that you do understand well. so I'm very into AI. So I talked to AI about AI a lot, and I'm also quite into speaker building and audio.

Speaker 2 00:11:27 And so I'll talk to you about that a lot. and that's a really great way to, to learn what the models are good at and, and where they fall down. so that's kind of my number one tip is make sure you're using AI daily. and the second is, is use it to do analysis. So, so one of the great things in customer services is there's lots of data. There's lots of conversations happening all the time. and some of our biggest wins have come from using AI to analyze that data. And look for opportunities to improve customer service, improve the way agents are interacting with customers, but also look for opportunities to deploy automation. You know, we had one example where like we we analyzed 1600 conversations, like every like every word and every conversation. We analyzed it all, we categorized it all. And we identified 10% of the interactions of a particular type that could be automated. And now we're just beginning the process to to deploy that use case. But we'd have never found that before because we could have never looked at 1600 conversations.

Speaker 2 00:12:31 But but now you can do that. Any cost if you want a dollars of, of API usage.

Speaker 3 00:12:36 Yeah.

Speaker 1 00:12:37 I'm certain that you have likely seen carry some instances where I was implemented poorly, and there are consequences to that. I don't know if there's any that comes to mind. You don't have to call anyone out personally, but. Oh, no.

Speaker 2 00:12:54 This isn't quite famous ones, actually. So there's, there's one, DPD, which is a delivery firm, that operate in the UK and Europe. they deployed, a chat bot quite early, and they didn't do the work required to put what's called guardrails in. and so a disgruntled consumer deliberately led astray and, and got it to make up poems and songs about how bad DPD was. And of course, they went all over Twitter and social media and was very embarrassing. now that's that that's quite straightforward to fix. You can walk what's called guardrail, the model. So you give it very specific things. It can and can't say, and doing that right.

Speaker 2 00:13:34 will will avoid issues like that. There's other issues. There was one with Air Canada actually, where their chat bot, got inadvertently led astray by a customer, poor bereaved customer, needed to change their ticket. And, the chat bot made up, and it's called hallucinating. When a when a model make something up, it made up a false refund policy for bereaved passengers. Oh, no. And and then I kind of took that poor bereaved customer to court to, to claim that that the chatbot didn't really represent them. And absolutely correctly, the judge sided with the claimant and, and er Canada had to, honour the refund that its chatbot offered. So I mean, just a couple of examples that I think, I think, you know, tell us quite a lot. Firstly, your AI, your chatbot or your voice bot, it does represent your business. The courts will will side with claimants who, who, who want to uphold, you know what your bot said. and the second is, you know, it's really quick and easy to build a prototype, to build a really cool looking demo.

Speaker 2 00:14:37 We do it all the time, but actually doing the work to to evaluate does it work the right way all the time, and how do you make sure that that it doesn't go off topic and make sure it's got all the information it needs to answer questions. So I think that that's that's an intuition that, that you'll get as you start to work with modern jealousy by more is that it's been trained to help. I mean, the way it works is it was trained on the whole of the internet, and so it learned a lot of concept about language, but but also the things that that have been published on the internet, all sorts of science and languages and practical things too. but that wasn't useful until we did what's called fine tuning it to, to answer questions. It's called instruction tuning. So we teach it that when we answer it, ask a question, it should answer us. but through that process, you know, we've literally whipped it into submissions that whenever we ask it a question, it will use its knowledge to answer that.

Speaker 2 00:15:34 so if we ask it a question that it doesn't have the knowledge to answer, then it will do its best and it'll end up making something up. and so that's where the tools and techniques come in to make sure that the model always has the information it needs to answer the questions that your customers might have of it, but also you give out if it doesn't know the answer, it needs to say it doesn't know the answer and these techniques, and many others deployed appropriately, can help us wrestle this quite amazing technology to to actually do something useful. I think that is one of the hardest things is, is really recognizing what it's capable of and its limitations. And then in the process of going from from a cool demo to a production system, really putting a lot of work into the evaluation. And some of that evaluation is manual people manually interacting with the models. But but actually we found that a lot of benefit comes from automated evaluations. So you'll, you know, we'll have one chatbot paying a customer and another chatbot playing the business and have them interact.

Speaker 2 00:16:38 Then we'll have a third chatbot look at those conversations and say like, how good was it? And what did it? Did it stay on compliance? Did it get drawn off into conversations about, things that, that it really shouldn't do?

Speaker 3 00:16:50 Yeah.

Speaker 1 00:16:51 so your website, Waterfield Tech. Com to a friend that's been listening And, like, how do they begin this? Like when, when if they were to have a conversation with you, like what usually happens in that conversation?

Speaker 2 00:17:04 So, everyone wants to talk about Jenny because everyone bosses, everyone's boss is asking, like, are we using AI? How how can we leverage it? So we end up having lots of conversations, people just just wondering what is capable this technology is capable of? the thing that I think surprises a lot of people is really the quality of conversations it can have. So so we end up do a lot of demos, show people what's possible. and that's a benefit. But it's a trap too, because it shows what's possible.

Speaker 2 00:17:33 But, but, but as I said, the hard work is in getting from that demonstration from that prototype into production safely. and so that's actually a lot of the, the work that me and my team do is helping people understand the whole journey from, from finally finding the opportunity. Find the concept, picking a use case that could work, that can give, give value, and then doing all the work it takes to get that deployed safely to to what's called red team it to make sure, nothing goes wrong in production and do all that automated and, and manual evaluation as I said. So a lot of it is laying out that journey and then and then just just being by the side of our customers as they go through that journey. Because, you know, this is it is another source of intelligence. It's another it's part of the team. I think rather than think of AI as a particular tech that you might deploy, I think it's much, much more useful to think of it as, as, as, as a workforce, absolutely can deploy.

Speaker 2 00:18:28 And they need leading and managing, just like your, regular workforce does. Yeah.

Speaker 1 00:18:34 Well, Kerry Robinson again, the website Waterfield tech. Com, there's an orange button that says talk to us. click there and, grab some time together. but everything again you see consulting contact center applied I workforce optimization, managed services, product development. again you'll learn a little bit more about that at Waterfield Tech. Com. Carrie Robinson, you are the vice president of conversational AI. Very, very exciting space to be in. Carrie has been a great conversation. Thank you so much for joining us.

Speaker 2 00:19:08 I'm enjoyed it. Thanks, Josh.

Speaker 1 00:19:15 Thanks for listening to The Thoughtful Entrepreneur Show. If you are a thoughtful business owner or professional who would like to be on this daily program, please visit up my influence. Com and click on podcast. We believe that every person has a message that can positively impact the world. We love our community who listens and shares our program every day. Together we are empowering one another as thoughtful leaders.

Speaker 1 00:19:41 And as I mentioned at the beginning of this program, if you're looking for introductions to partners, investors, influencers, and clients, I have had private conversations with over 2000 leaders asking them where their best business comes from. I've got a free video that you can watch right now with no opt in or email required, where I'm going to share the exact steps necessary to be 100% inbound in your industry over the next 6 to 8 months, with no spam, no ads, and no sales. What I teach has worked for me for more than 15 years and has helped me create eight figures in revenue for my own companies. Just head to up my influence. Com and watch my free class on how to create endless high ticket sales appointments. Make sure to hit subscribe so that tomorrow morning. That's right, seven days a week you are going to be inspired and motivated to succeed. I promise to bring positivity and inspiration to you for around 15 minutes every single day. Thanks for listening and thank you for being a part of the Thoughtful Entrepreneur movement.