Envisioning a personal AI agent for every customer - Interview with Malte Kosub of Parloa

0:00So welcome to the next edition of the Punk CX podcast. With me today I have Malte Kosub, who is the CEO and co-founder of Parloa. Hey Malte, how are you doing? Very good. Thanks for inviting me, Adrian. You're very welcome. Now, for people that don't know you or don't know the Parloa, can you give us a bit of a thumbnail sketch about who you are and what you do? Yeah, happy to do that. So we are an agentic AI platform for customer service. So we help the largest enterprises out there to bring agentic AI into their customer communication, starting with customer service. Perfect. And so before we go, and I also need to congratulate you, apparently, because you recently announced a successful Series C funding round of$120 million on a$1 billion, if I'm right, kind of valuation, which officially makes you a unicorn.

0:55So you've got a big kind of horn coming out of the top of your head. And to be fair, there's not actually that many of them around in Europe. So congratulations on that. Thanks. Thanks a lot. a lot of hard work behind it the team is doing a fantastic job uh and yeah happy uh that we reached that that milestone but i always tell to the team it's still day one right so there are a lot of things in front of us excellent and so give me a bit i mean so there's there's a huge i mean i don't know if you're familiar with it but there's but there's a guy called scott brinker that has done a piece of work over the last sort of 12 or so years he talks about mapping them the martech landscape and he's been tracking kind of how many applications are in that martech landscape and it's gone from a few hundred a couple thousand to like over 12 000 over the last kind of 12 years or so so it's incredibly crowded out there particularly and particularly in the customer of service and experience sort of space and so i wanted to understand a little bit more about kind of parlo and and what makes you different and sort of like stand out what's driving a lot of the things that that's making you so successful yeah and there are a couple of parts that makes us special and i always tell to the team we need to be laser focused on the segment we are we're focusing on the first topic is that we are focusing on the large largest enterprises out there so the whole platform the whole company is built to serve the largest enterprises and there you need to do a lot of things differently than if you would focus your platform and your company to SMBs and mid-market companies, right?

2:40So that's the first part. The second part is, if you look at those large enterprises, still phone is by far the most relevant channel for them. The highest volume, the highest costs associated with it. So we have been focusing on voice and phone for the majority of our company history. We also enable customers to use chat and Messenger. However, we started with voice and all the others, most of them started with chat. So we are outperforming everyone when it comes to phone AI agents. Okay. And then if we take the third step, again, we're talking to those large enterprises. Most of them, they are global enterprise.

3:26enterprises with customers around the globe and as we are from europe we needed to enable our our customers to talk to their customers in a lot of different languages because that that's that is present in europe so we help those large enterprises to serve their customers in in a multiple of of geographies languages and so a pretty clear segment of of enterprise we're targeting which is it's interesting because if you think about some of the other companies in the in the space many of them originate out of the north america let's say exactly and they become english first as their approach so because you're coming at it from a multiplicity of languages because you understand that's the nature of of things i guess that's already kind of like hard wired or baked into kind of your approach you know that sort of multiple language support but But I guess it's also, you know, and it means that you don't actually have to go looking for patches or fill-ins or other kind of things because it's already kind of baked into what you do.

4:33Exactly. And people think that language support is solved in AI, but it is just not solved. And there are three layers when we look at our industry that are relevant to serve AI agents in a lot of different countries and languages. The first one is that you actually need to integrate into the phone infrastructure of those different countries. and own infrastructure is science in itself. Second thing is that there are different regulations in all those different regions. So you need to make sure that the platform is built for those regulations. The third layer is that the models in those different languages, they have different quality scores.

5:16So a lot of people think speech to text is solved in all of the different languages, but the quality is very bad especially if you look at smaller languages and if you look at things like alphanumeric numbers so we have fine-tuned a lot of those languages for our use cases for phone audio quality so enterprise come and then the languages are ready to be delivered on the highest level possible I mean I remember talking about low-resource languages and I for the last couple of years I was you know you were spotting this is a bit of a chink in the armor of some of these large LLMs and some of the countries which our would be classified as low resource languages weren't getting as you say the quality they would they would they would require I remember speaking to a company that was focusing primarily on the Nordic languages it's like danish and norwegian and swedish and finnish because they're just hard and there's not that many people so speaking them but because they were building models specifically for them then the quality right there are because then when you add in it's not just the complexity of the language but it's also some of the cultural differences and the way that people can speak and so i think that's really good and to fine tuning um to adjust for those those differences is incredibly valuable, particularly when it comes to driving different or better outcomes.

6:44Yeah, fully agree. Perfect. Now, we all know that this space is dominated by the customer service, customer experience space, particularly the technological space, is dominated by what comes out of North America and primarily what comes out of Silicon Valley, you know, in large part. And they drive a lot of the conversation and the rhetoric and the vision and things. But I know that you've got some interesting kind of thoughts around how you think customer support is going to evolve over the coming, say, five years. Now, if you want to go five years out, then knock yourself out, because I think that's just difficult in and of itself.

7:23Sometimes it's hard to predict what's going to happen next week, never mind next year. but i thought it'd be really interesting just to get like a very european sort of view on how you think things are going to evolve over the next sort of one kind of three and kind of five years and as a sub question to that is can we given that vision can we assume that that 2025 is year one okay yeah um happy to share some thoughts and um in general i think there are also a huge a lot of great great companies tech companies that are not originated from the bear area and we want to show the world that there will be a lot more coming in the future but let's let's let's shortly talk about how we believe customer experience will change going forward if we just look at 2025 that is still rather straightforward so you will have if we look at customer service we will have way more ai agents actually picking up the phone or answering a question in a chat and their quality of answers will be significantly better than what you know from old chatbots and voicebots and so that is the the straightforward first approach and those ai agents will solve end-to-end use cases they get more natural and so on i think if we go one step further we believe that actually customer experience as a whole will change because we believe every touch point will become conversational so every app every home page every single touch point between a customer and a company will become a conversation there will still be ui components obviously but the ui will also be driven by a conversation um and if we believe this is true we even go one step further and we believe there will be a personal AI agent per customer so Adrian imagine you're calling an airline right now you would probably wait in line for 15 or 20 minutes and if you are then connected to an agent you would talk to a different agent every single time right now imagine you're calling an airline and your personal AI agent picks up and says hey Adrian how are you have you decided if you want to upgrade to premium economy for your next flight from New York to Los Angeles?

9:48If yes, awesome. If no, just call me back whenever you want. And then 20 minutes later, you could call back. And after one second, the same voice, the same personal AI agent picks up and says, hey, Adrian, and have you decided? And if you don't want to call, you could also take out your phone and write a WhatsApp iMessage to say you want to upgrade. And if you don't want to call or write, maybe on Monday, this personal AI agent might call you and says, Adrian, there are just two seats available in premium economy, do you want to have one of those seats? Otherwise, I would give them to someone else.

10:20So we have a one-to-one relationship between you as a customer and a personal AI agent of that airline. And if that airline has 150 million customers, this airline will have 150 million personal AI agents. So we are creating an experience that is not just an efficiency driver for an enterprise. It's actually elevating the experience significantly because you create an experience that has never been possible before because it was way too expensive, right? To have one person for one customer. So that is what we believe will come true. And then if we go a level deeper, we also believe that there will be personal assistants like Siri, ChatGPT that will call or reach out to companies on behalf of the customer.

11:09But still, you need to have this personal AI agent of an enterprise interacting either with the customer or the personal assistant of the customer. right and so i guess my um my question on that i mean that you talk about conversational interface and because it always makes me think that we there was a bit there's been a big shift away from uh well big shift to conversational interfaces and my sort of view is that there was in that sort of like um space there was a danger that we were if you know the expression throwing the baby out with the bathwater, which sort of means we're throwing some of the good stuff that we knew worked just because it was getting displaced by this new technology.

11:56So for example, conversational interfaces, if they were largely text-based, let's say, what many companies were doing were saying, well, we were using some of our predictive analytics to understand the high volume requests that we're getting in. And we were presenting them to customers and their customers are going, ah, I can read that. It's really quick. I'll pick that one, that one, that, and it gets me rooted really quickly or into a sub chat, as it were. But then people have been moving away from that, but they're not realizing that actually it takes longer and it's harder to type something out rather than presenting somebody with options.

12:33And so how do you think that conversational AI interface is going to change over time to get that right blend? Because it should be about effort and efficiency. I fully agree. And I don't believe that it will just be a conversation. I believe that there will be visual elements where you can swipe and click. But the conversation will be driven by the experience will be driven by a conversation in combination with UI. So just imagine you're going to a UI in e-commerce store. What we envision is that the first thing you see is a speech bubble where you can talk or a chat widget or chat window like in ChatGPT.

13:16And you say, I need a trousers for an event on Saturday. Or I need the same trousers last time in blue. And then the next thing you see is images, options where you can swipe or click, right? and then you can maybe write or talk and say, no, do you have something from that brand or smaller or more blue or whatever. But it will be a combination of UI and clicking and a conversation. So I fully agree with you. Do you think that you talk about this kind of vision and how it's going to be primarily, it'll be multimodal, obviously. it will be you're talking about the enterprise as well I mean and then you're primarily focused on the enterprise as you can like to have said but I'm also kind of interested because you're coming from, I know you're founded Parlo was founded in Berlin and you've grown all over the place and Germany's like famous, the German economy is famous for the Mittelstand and And I wondered, do you see that kind of support is going to be, as it evolves, is going to be lumpy?

14:25I mean, are enterprises going to lead the way and then there might be an increasing gap between them and some of the smaller businesses? I mean, how do you see that kind of like working? Yeah, I would not differentiate between enterprise and mid-market. I would say if we look at both categories, we have innovators. Okay. And we have companies that just take way longer to adapt to change. And we see very large companies moving actually pretty fast because they want to report something to the board, to the public markets that they actually drive efficiency, they drive innovations with Agentec AI.

15:04But we also see mid-market companies that adapt the technology rather fast. Obviously, the depth of integrations are different. Maybe also the depth of individualization is different. because enterprises have more money, more resources to individualize the experience to their brand. But I wouldn't make a difference in how fast they move. Sometimes people say enterprises move way, way slower than mid-market companies. In that particular topic, we see that enterprises actually move pretty fast, because it is expected from the public markets that companies move, because it's just a drastic change in how you run a company.

15:46Yeah. I mean, just to kind of comment or question that, I mean, because I've seen, I mean, there is no doubt about the potential of the technology out there and its transformative impact on both, you know, executives and leaders and employees' ability to serve their customers kind of like better. And there is no doubt that customers could also benefit from this. So there could be improvement in outcomes all around the peaks. But there's also bits of research that suggests that many companies sort of struggle to get going, to get the most out of the technology or to deliver a return on investment.

16:30or they find it hard to get beyond the pilot or the proof of concept and then to implement it at scale. I mean, what are you seeing that enterprises are doing to accelerate themselves up that slope? Because there's many enterprises that are going on the slope and then they go, and then they stall, and then it takes a while to reset. And other ones just seem to be tearing up it, and they must be doing something different. Yeah. Yeah. So let's talk about the two situations where we see that actually the return of invest was not really there and pilots failed. Yeah. The one area is old voice for chatbot projects with old technology that still happen with old technology vendors where you're not you're not doing agentic first.

17:24So a lot of those older voicebot and chatbot providers, they talk about AI agents, they talk about agentic, but the core of their platform is not agentic first. So if you use the old technology, actually there are limits in what you can deliver. And still a lot of customer support, customer service departments, they struggle with their past experiences with voice bot and chat bot provides, it's not working, but this is a completely different game what we're talking about. That's the one part. The second part is agentic projects where the use case and the domain is not clearly defined. And that also happened.

18:08When we look at customer service right now and if they implement an agentic first solution, it's a pretty straightforward ROI case. And we're seeing massive value that we're delivering while increasing the customer satisfaction. So it's a win-win for everyone, right? And I think you just need to get going and you need to be agentic first because that's a completely different quality level. But does that require, therefore, they just need to be, to quote, you laser focused on what it is you want to do to start working on, to create a little, to create success. Yeah. And then to generate kind of momentum to scale from there.

18:51I mean I think what I'm trying to do is I'm trying to kind of understand what are the success factors if there are kind of commonalities between some of the people that are being able to implement some of this stuff and take an energetic sort of first approach at scale yeah so if we look at customer service again you need to be ready to integrate back end systems because if you just want to pick up the phone and do some FAQs or do a chatbot with FAQs yes you will you can actually already drive return and a positive return of invest. But if you really want to drive value, you need to start actually connecting the agents to your backend systems to drive end-to-end cases.

19:31You can start with exchanging your old IVR, then you do some FAQs, but you should also already start thinking of what are the use cases where I need an API that could be an authentication, an address change, where's my order. So very straightforward cases, but they are cases where you have a lot of volume and you can drive value pretty fast. Yeah, but they also can require you to make sure that you have your, I guess, your technological architecture in such a state that you can plug things together to allow the different agents as part of the agentic process to all speak to each other to deliver those kind of outcomes.

20:16And that is no small undertaking for many firms. I agree. You need to have your APIs and your backend connections ready. A lot of companies, however, for relevant use case, they have it ready. So a lot of companies, they have apps. So within their apps, they have already connected the most relevant APIs. So it is existing. It just needs to be connected to an agent that can talk to a customer. I mean, it would be great if you could try and kind of bring how you approach things to life. And if you can share any examples of companies or brands that you might have worked with, obviously, if you can name them, that's great.

20:53if you can't then that's fine just give us a examples of work because and it just sort of brings it then brings your approach to life and then you know around kind of what they were what challenges they were facing what they were what they were trying to do kind of what you did and kind of how they're kind of benefiting of that because then it just sort of goes oh this is the potential this is what people are doing this is kind of the the opportunity that's on offer yeah so we're a horizontal platform so we're working with companies from a lot of different industries and maybe to give you some some examples of our customers in in those different industries so one is retail so they retail customers they have stores uh around europe and one of the biggest challenges was that just in 50 of the cases someone pick up the phone because the people in the stores they need to talk to customers they are doing their work and so we implemented our ai agents and the ai agents are picking up in 100 of the cases right and they were able to to schedule meetings with the people in the stores and okay and what had a direct impact on their revenue because if no one picked up the customers called another store yeah um and that is They have hundreds and hundreds of stores, and that has a significant impact on their revenue, on their top line.

22:19That's one retail example. Well, I mean, I like that example because it reminds me of almost like a personal shopper type of appointment, right? And so you get to meet. But it also reminds me also of, there's a friend of mine runs this application where it's all about giving people with disabilities to facilitate access to kind of retail environments. But what it does is sort of pre-warns people that they're going to license to pick appointments so they get the service or the attention that they require. Because often or not, they might not show up or they might show up to a thing and they might not feel kind of the barriers.

22:54But if you can let somebody know that you're coming, then it makes it feel really positive. So it's a much more of a personalized or personal, rather, experience. So I mean, I like that one because it makes me think, oh, yeah, that makes sense. From both the personal shopping angle, but also from the access and accessibility angle. Yeah, fully agree. And that is one retail example, e-commerce example. We have a customer where people can call to make an order. And they grab hundreds of millions of revenue to that channel where people are calling. and talking to an AI agent and ordering something.

23:28Awesome. We have travel customers where people call because they want to cancel a booking. Pretty straightforward, very straightforward case. But there are significant amount of people calling and it's a straightforward use case, right? We have insurance companies where it's about claim management. So there's like 10, 12 questions. It's a very straightforward process, but an AI agent now picks up after one second, ask those questions, summarize everything, puts everything into the claim management system and then it's organized automatically and then the insurance can start picking things up. Or energy companies, one of our customers, they're driving all the meter reading submissions.

24:10They're saving more than two minutes on average per call because those relevant topics are being handled by the AI agent. So in all those different industries, is you find very repetitive use cases that the first line human customer support agents handled in the past, which our AI agents now can handle very smoothly and drive a lot of value for customers and companies. And just as a supplemental question on that, because you mentioned the human agents, I mean, I think this is also a really important kind of part of it. I mean, that, you know, it's it's that question that the the customer service space is going to change and the roles of agents are going to change i mean that how are you seeing kind of agents or companies that are employing your technology and redeploy their agents as it were kind of how does how's the nature of their kind of jobs kind of doing i mean are they are are they being the humans in the loop Are they being the overlords, as it were, making sure that the bots, any tweaks that need to be made, or making sure that the quality that they're delivering meets standards?

25:23I think the first topic to mention is that there's a huge lack of labor that wants to work in customer support. A lot of companies in the US, in the UK, in Germany, they have significant challenges of finding the right staff to actually do customer support. And for those that are working in customer support, their role will change. So they will focus more on very complex tasks, level two or level three topics. They will be elevated to be supervisors. So they will coach the AI agents to perform better. They step in and support the AI agent when there's a tough topic there to cancel. they will start training the agents for new use cases.

26:11So the role will change in the future of customs for agents. Yeah, no, it's great. I mean, I think that's the thing. It feels like the role, if people step into it, I mean, there's always going to be a role for people talking to people, I think, in certain situations. but there's also growth in that in that role whether that the the technology is is is presenting and and i would encourage everybody to kind of to lean into that to you know to learn more i mean so i think that's a it's a it provides a an interesting kind of outlook in terms of the the the future of customer experience and but also the the the idea that it's it's going to be voice is going to be at the heart of things.

26:59And that's fascinating because for the longest time, everyone was talking about voice is dead. It's all going to be text-based. But actually, voice over the last year or 18 months or so has made this roaring kind of comeback. And people are realizing and go, ah, people do talk to each other. Exactly. Or they talk to things and then things get done. They leave voice messages or send voice notes and things. And so we naturally talked to people and so it's it's good to see that people are folding in sophisticated voice strategies into into the into the mix so people get the help the right sort of help when they need it which is kind of great but walter they thank you for sharing some of those kind of reflections and thoughts with me i mean and and again congratulations on your success and and some of the things you've been doing with different kind of different kind of brands i wanted to ask you finally before we think about wrapping up is that is there anything else that you'd like to highlight or like to add that before i can ask you some quick fire questions so there are a lot of people in the industry talking about ai agents agentic ai and i think it's important to go one two three levels deeper what they're actually delivering from a technical perspective and really try to get into the detail what they're delivering and are they really delivering what you need from a technical perspective so you compare apples with apples and if you want to get a bit more insights into that feel free to reach out yeah no absolutely i think there is a uh to paraphrase what you're saying i think in a slightly more direct way i think there's lots of smoke and mirrors out there exactly i can say that more easily than you can and i think that's that that's true there's kind of people that are popping popping up and promising all sorts of different things all over the place and sometimes you have to be very careful about don't necessarily buy the sizzle make sure that there is exactly there's something that sits underneath it exactly so a few quick fire questions before we we wrap up so i always like to ask people you know i know we talked about a few things I always ask people to kind of like to boil it down and the way I've been doing that recently is to ask them to complete this sentence and the sentence is this Malta is if you want to improve your customer experience Malta says do this complete that sentence sir start thinking about easy use cases you can implement within two three four months and then make a decision to move on and implement it and not just talk about we need to implement ai issues but actually take easy use cases and start implementing and start learning from from that experience awesome thank you and a second question and it could not be um well it has to be a punk related question because it's a punk cx podcast and so what i wanted to ask you is like what company or brand do you think takes a punk approach to customer experience and why so i think there are a lot of great bands out there um maybe to to mention one brand um i also interact a lot with is booking.com okay i think they always think very very customer centric what is the best experience to get to the goal the customer wants and they've been doing that throughout their different touch points and they were very successful with it uh optimizing for the end customer yeah um yeah i mean i love that i i remember i remember talking to um a gentleman called stephan tomkey who is a big fan of an advocate for experimentation and creating a culture of experimentation and i know that booking.com are at the vanguard of that and they have this experimentation platform where people can run experiments and i also have the power anybody in the company has the power to shut an experiment down if you think it's going to conflict with another one exactly which i think is it's it's so democratic it's wonderful yeah and uh so i thought it's a it was a great example and if people haven't heard of the book then you should definitely check it out it's called experiment or there's two of them experimentation matters and experimentation works and my final question before we wrap up is that i've been asking people recently i mean it's been quite a few months now but i've been asking them recently to tell me a good news story because the world well it's just a bit weird right now and i'm trying to ask people to give them a good news story it could be personal it could be business it could be anything just to try and end us on well that's not to say that anything has been a real downer what we've been talking about but to end on a nice, positive, interesting, or exciting kind of like new story that you've seen in the last week, say?

32:05Can I talk about two? Yeah, of course. So one is more a personal related story from the business. We went live with a huge customer a couple of weeks ago and you need a couple of weeks to really gather lots of data and the outcomes this week, which they presented to the C-level was pretty interesting. The average handling time went down, the CSET went up, and actually the agents loved the AI agents handling the easy tasks. So everyone was happy. And there were a lot of people saying, ah, the agents will be disappointed or the customer will be disappointed because AI will engage or it will not drive value.

32:48And no, it's actually happening. And everyone is happy in the whole chain. that interact somehow. So that was my highlight. I think a second highlight, probably everyone knows Eric Schmidt. He recently had a TED talk talking about the future of agentic AI and AGI, super intelligence. And I think everyone should actively think about what is actually happening with the world and that we need, as humans, we need to make very decisive we need to make important decisions going forward to guarantee that what is happening actually serves as human species and we actually create a better world and we want to end on a positive note I think we can actually make it happen that it can benefit all of us but we need to make it happen and I would encourage everyone to watch that talk It's a great one.

33:49Awesome. Well, I will make sure I look that up and include a link in the highlights of the talk. But Malta, I just want to say thank you for, well, first of all, congratulations on the raise. That's kind of brilliant. And congratulations on your success so far and kind of leading the, you know, being part of the big charge around kind of European innovation and technology that's having an influence on the service and experience sort of space. And just want to say also, thank you for sharing your time and your insights with me today. That's been great. Thanks, Adrian. Thanks for the invitation.

34:26Thanks for the great questions. And looking forward to all of what is coming in the future with AI and improving customer experience. Awesome. Wow, what a great interview. I hope you enjoyed it. I know I did. Find out more about me and the work that I do at adrienswinsko.com. do leave a review on your favorite podcast platform and if you have any comments feedback or questions about the podcast then feel free to send me a message to podcast at adrianswinsco.com and do tune in again thanks very much