Can AI Voice Agents Fix Customer Service?
Brian Kenny says a customer once cancelled a clinic appointment with Laila, Momntum's AI agent, because her dog was ill. The lady rings back in about a couple of months and the first thing Laila says is, how's your dog?
In this episode · 18 chapters
- What the mission is
- Has customer service got worse?
- Emotional intelligence
- The relationship model
- Throwing it to a frontier model
- They will not take on a business unless
- What contact types
- Miami, and Spanish
- The real reason
- The crazy experiment
- Accents, dialect and swearing
- Why Laila?
- The thought experiment
- Pricing
- The Gartner number
- Should our kids study it?
- A personal question
- The EU AI Act
Humans should be explorers and adventurers and continually challenging and pushing themselves to go further. And technology is probably one of the biggest aids in doing that. It unlocks rooms that we’ve never been in.
Brian Kenny, CEO and founder of Momntum, built Laila, an AI agent that handles phone calls, WhatsApp and Instagram messages for businesses and remembers each customer. He explains why Momntum won't take on a business unless Laila can connect to its booking or CRM systems.
Kenny sets out how he measures Laila's success rate, why she is not ready for bank transfers or medical records, and how she speaks Spanish to clinic customers in Miami. He leaves a question: should an AI agent always introduce itself as one?
Brian Kenny, CEO and founder of Momntum, argues on the Thinking On Paper podcast that AI voice agents only fix customer service when they are built around a lasting relationship with each customer, not bolted onto existing call-centre processes. Momntum's AI agent, Laila, handles phone calls, WhatsApp, Instagram and Facebook Messenger, and remembers returning customers.
Can AI voice agents fix customer service?
Only when they are built around a lasting relationship with each customer, Brian Kenny argues. Bolting an agent onto existing call-centre processes does not fix the service.
What does omnichannel support mean for Laila?
Laila works as true omnichannel support, meaning one continuous conversation that can move between phone, chat and messaging apps. Kenny says she recognises returning customers by name and, if a call runs long, offers to hang up and carry on over WhatsApp.
What is the Momntum relationship language model?
Momntum keeps customer knowledge in its own relationship language model (RLM), a proprietary layer that learns who each customer is. Kenny says it sends only parts of a conversation to outside models such as OpenAI's and Anthropic's, under agreements that they cannot train on the data and must delete each thread.
How successful is Laila at handling customer conversations?
Kenny reports an 86 per cent success rate, measured as the share of conversations that end without the customer asking for a human. He says Momntum is tracking a 4,000 per cent return on investment, with Laila generating over a million euros in bookings for clients every two months.
How does Laila work for medical spas?
Momntum's first market was medical spas, where Laila books, cancels and reschedules appointments across clinics in the UK, Ireland and the US, and speaks Spanish to Miami customers. Kenny's example of emotional memory: a woman cancelled because her dog was ill, and when she rang back months later Laila's first words were “how's your dog?”
Which tasks does Momntum keep Laila away from?
Kenny keeps Laila away from high-risk tasks such as bank transfers and medical records, where a hallucination, meaning a confident but false AI answer, would be costly. He also says Momntum will not take on a business unless Laila can connect to its back-end systems, so she can resolve requests rather than promise a callback.
Where Mark and Kenny disagree
Kenny asks whether an AI agent should always say it is an AI, and Mark first answers no, then reverses himself on air. Kenny says Laila introduces herself as an AI at present, and “Whenever you ask Laila, is she an AI? She'll of course say yes.”
Links from this episode
Questions
Can AI voice agents fix customer service?
Only when they are built around a lasting relationship with each customer, Brian Kenny argues. Bolting an agent onto existing call-centre processes does not fix the service.
Who is Brian Kenny?
Brian Kenny, CEO and founder of Momntum, built Laila, an AI agent that handles phone calls, WhatsApp and Instagram messages for businesses and remembers each customer.
What is Laila from Momntum?
Momntum's AI agent, Laila, handles phone calls, WhatsApp, Instagram and Facebook Messenger, and remembers returning customers.
When do AI voice agents improve customer service?
Brian Kenny, CEO and founder of Momntum, argues on the Thinking On Paper podcast that AI voice agents only fix customer service when they are built around a lasting relationship with each customer, not bolted onto existing call-centre processes.
Transcript
Mark Fielding disruptors and curious minds, welcome to Thinking on Paper where we explore your truth about the impact of technology on business, on culture, on humanity. I'm Mark, my co-host Jeremy is not here today, he is on a secret mission, but rest assured he will be next week, regular programming will be resumed then, until then you'll have to make do with me, but.
In exchange for your loyalty, we have an awesome guest. have a sublime show. We're talking about something very important. We're talking about customer service. We're talking about AI powered voice agents. We're talking about omni-channel messaging. We're talking about the hyper personalization of every business to customer exchange. We're talking about, I don't know, maybe our guest, Brian Kenny, CEO and founder of momentum can add some.
detailed to the numbers, but I got like 17 million people possibly in the world work in some kind of customer support. It's kind of the frontline of every business. And I think for a long time, we've heard that AI is going to revolutionize this customer touch point. But anyone who's experienced that, an AI chat bot in 2025, at worst it's...
It's infuriating at best it's passable. But you're helping to change that aren't you? Brian Kenny, welcome to Thinking on Paper. Thank you for thinking on paper with us today. And first question, I think I read on your Twitter that you're battling the caffeine demons that you're trying to cut down on your coffee. How's that working out?
Brian Kenny Not very well. No, I am. still on close to four cups of coffee a day, but trying to play this this mental battle of really nice, clear thinking versus being over caffeinated and getting a bit too much of the caffeine jitters. So, fighting that continually.
Mark Fielding It's a lifelong battle, the caffeine addiction to battle. Right. Right. Before we get into this Brian, let's tell us why. So you've got a history, a background in multiple businesses, not all of them connected to AI, not all of them connected to customer support. What brought you here on your journey? What was your aha moment? If you like.
What the mission is
Brian Kenny It's a fantastic question. I think to summarize it up, think the biggest problem that I've noticed in customer experience or customer service at the moment is if you ring up your broadband provider or your television provider or your bank or whatever it might be, it's always the same situation that I've encountered where it's like really long wait times. It's disconnected responses where
I could be ringing maybe two or three times because I got to reiterate the same problem to a different customer service agent. They don't know that I rang multiple times before they got to pull up my record. And it's all so very, very transactional. I think it's like kind of take a ticket, stand in line. Once we've dealt with your query, you're off my books.
We hope you get your problem solved, but we've no idea. It's each kind of agent is essentially just doing their role and trying to get through as much volume as possible and using things like CSAT scores or customer satisfaction kind of tables in order to determine have we done a great job or not. And I honestly think that, you know, in this modern age with the amazing technology that we have that there's such amazing potential and opportunity for building a customer experience platform from the ground up for the modern consumer. so this is essentially where Momntum has stepped in. Our goal is really to leverage artificial intelligence and AI to build a strong relationship between a business and its customer.
So we've been on this mission now since January. We've had some substantial kind of breakthroughs along the way. But the goal that we've set out to do is to really strengthen that bond between the business and the customer and focus intently on lifetime value as in customer satisfaction over long periods of time rather than transactional on the phone, deal with the query and move on.
Brian Kenny style of customer service.
Has customer service got worse?
Mark Fielding I have the impression that customer service has deteriorated in the current age of AI as bad as it was when it was just speaking to a human over the phone. At least you felt like you were getting somewhere. You have the experience now isn't better on the whole. It feels more truncated. It feels slower. It feels like your problems aren't really being resolved. How?
How's the funding going for starters? think you've got some investment you're plowing ahead.
Brian Kenny Yeah, no, like you raised such a good point, which is, you know, with the evolution of AI, seems like everybody is deploying these solutions, but we're getting nowhere. And this is mainly because going backwards and like driving people crazy. And it's mainly around this, this idea where it's like an iterative approach as in this is how customer service is right now. And then we take the block of AI and we stick it on top and then we cross our fingers and we hope for the best.
Mark Fielding going backwards.
Brian Kenny And I truly believe that that's just not going to get us to where we need to be. So there's really fundamental things that we've done from the ground up as we build it completely for the modern consumer from the ground up. Number one, we got to be on every communication medium. So currently Laila which is our AI agent, she's on phone calls, she's on Instagram direct messages, she's on Facebook messenger, she's on WhatsApp.
She's on every communication medium that the modern consumer is. then doing really simple, omni-channel, but true omni-channel, like doing a really simplistic thing, is when you ring Laila and if a conversation, so first of all, Laila will recognize you straight away if you're a customer there and say, Brian, and know everything about previous calls and you know, your records and what you might be.
Brian Kenny talking about during that interaction. But then secondly, really helping to figure out how is the customer feeling on this call. So if the call goes on longer than, let's say, two minutes or three minutes, or what Laila would deem as it taking too long, you just say something really simple like, if you want this call that's going on a bit too long, we can just continue on WhatsApp because you've got other stuff that you've got to deal with.
If you want, hang up the call and I'll send you a WhatsApp and we can continue the conversation there. And there's all of these really, I would say simple ideas, but modern consumer experience ideas that puts the customer first and builds that bond together with the customer and the business.
Emotional intelligence
Mark Fielding emotional intelligence. I want to ask you a question about what you just said, because obviously ChatGPT5 has just come out. Everyone's experimenting it, see where it's weak, see where it's strong. So thinking on paper, we like to use the AI to curate all our episodes. We have transcripts for over 200 shows, different guests across different domain.
put them all into an open, I know we shouldn't be doing that. We should be using a decentralized AI platform on site, but you know, we do. And one of the things I find continuously with, I found it with chat GPT-233.5, the next five, it has, when you have that much data, that many shows, it gets confused between the shows. It can't keep track. It can't separate. It mixes up the guests. It mixes up the topics. When you ask it to join the dots between show,
50 and show 75, can't do that. How is your...
relationship, you call it relationship language model, solving that problem. If you have all these different customers and it can identify their past histories, how is that, how is your model working to make sure that it's picking the right person, it's getting the right history?
Brian Kenny Yeah, so we've noticed that too. We've noticed with even really large context windows, the larger the information set that you give to one of these frontier LLMs, the more Jenga tower wonky the kind of the response comes back. And so what you don't want to do is to have a large thread and then try to
Brian Kenny just continually add more and more to this thread with one of these external models. So what we've really tried to keep as proprietary as possible is, first of all, we're completely privacy focused first. So with any of the advancements that we make within our ORLM or anything around emotional intelligence, we want to really protect privacy and the responsibility that we have in building technology in AI. So a big part of this is you can think of Momntum almost like its own box. And within this box, have the regular kind of data silos, the business, the businesses, customers, information on customer records, CRM style attributes, purchases, all of these things.
The relationship model
Brian Kenny And then on top of that, have our RLM, our relationship language model that really gets to know who the person is that they're speaking with. that's overlaid against what the business is trying to achieve, as in who is the business, is it a coffee shop, is it a beauty spa, is it a hairdresser, these types of things. And then when a conversation starts,
We essentially only leverage external LLMs, know, like the, the llamas, the open AI, the anthropics in order to process parts of a conversation. we're kind of going out and back, but we have really specific agreements with the frontier models to say, first of all, you can't train any of your models based on the data that we sent. And then second of all, when we open a thread, leverage one of the
LLMs and come back, we dispose of that thread and they must completely delete all of this data from their servers. it keeps all of the proprietary tech and the relationship or the emotional and intelligent part of Laila within our silo. And it's really required for parts of our customer base and things like that too. But it also circumnavigates this large context just
Throwing it to a frontier model
Brian Kenny throw it over the fence to a frontier model, cross your fingers and hope for the best. So kind of moving away from that wrapper mentality, but more into what is the best way of dealing with the conversation that's at hand.
Mark Fielding That's really interesting. And I guess from a, when a business or when you approach a business or when they approach you, obviously the trust of their customers is paramount. And if they can be assured that customer data is not being used to train the language models of the big mega tech AI companies, they can sleep easier at night and their customers can sleep easier at night knowing that if I understand that correctly.
Brian Kenny Yeah, and it's so important. mean, in certain cases, you know, the best way of describing how we'll use an external or third party LLM is really, you know, we're not going to develop our own LLM at that scale because, we don't have tens of billions of dollars in order to do that. However, what we do have is custody of knowing that we're going to build a relationship between the customer and the business. And there's a lot of fiduciary responsibility that comes with that. We want to really make sure that we have solid guardrails in place. You alluded to this even at the very beginning, which is there's nothing worse than a customer service agent, be it human or AI, who can't actually solve the problem for you. that means that Laila, yeah, exactly.
Mark Fielding just hand it off.
They will not take on a business unless
Brian Kenny So for example, Laila, we won't take on a business unless we have the capability of interfacing with our clients backend or technology or something. As in, if it's a restaurant booking, we want to access to that booking system. If it's a home security system, we want access to place the leads or sales into their CRM. Otherwise,
It's kind of this whole thing where you're interacting with a bot and you're going blah, blah, blah. And the bot says, I'll have somebody come back to you. And it's just the worst experience ever. So making sure that we are really strong on taking the customer's query and solving it, whilst also being very responsible in how we deal with customer data and AI and the frontier of where we're going and what that means is really kind of at the heart of our company.
Mark Fielding two part follow on from that. So, which the first part should probably be the second part. Can you give a percentage or a number currently of how successful Laila is with resolving different queries versus having to hand it off? Is there any data on how successful it is at the moment on that?
Brian Kenny Yeah, so I think what I want to do is be as transparent about this figure as possible. So we're currently hitting about an 86 % success rate, which is mind blowing compared to some of the statistics that are out there in the market right now. But how is this number derived and offering as much transparency around that as possible? We have the capability within Laila that if a customer of a business says, I want to talk to a human as in
I want to just talk to somebody who's going to, you know, ring me back or whatever it might be. In that case, Laila will create what's called a support request and then the company is notified and they can get in contact with that customer. If we took a hundred conversations and one support request was created, then that would be a 99 % success rate for ourselves. As in Laila has handled 99 out of a hundred.
And that's the metric where we're getting to 86 % from at the moment.
What contact types
Mark Fielding Okay, what sort of contact types is that? Is that on? Where's my order or can I book a table or you overcharged me, undercharged me, I want my money back. What kind of interactions is it solving at 86 %?
Brian Kenny So this is the model. The first one that we kind of the business that we landed is in the medical spa space. So kind of Botox, lip fillers, aesthetic treatments and things like this. you can.
Mark Fielding Just what Jeremy needs when he comes back next week.
Brian Kenny Yeah, well, we'll put him in contact for sure. But you can imagine the complexity of booking an appointment within there. First of all, this client has multiple clinics, clinics in the UK, in Ireland and in the US as well. So first of all, we have the multilingual aspect as in
Miami, and Spanish
Brian Kenny A lot of the Miami based clinics will all speak Spanish and so Laila speaks direct Spanish to every customer there, which is just unlocked substantial potential. Number two is the requirement that the potential customer might have, as in what does it feel like? What is the service? Do I go in for a consultation and then what's the follow on treatments from there? And there's all the complexities around some of these treatments.
but Laila fluidly helps all of those customers in an emotional way to navigate all the questions that they have and then to book appointments, cancel appointments, reschedule them, tell the clinic that they're running late. pretty much all of the things that you could imagine a person on the phone within the clinic would have to deal with. and now it's gotten to a point where for that business, she is the number one.
a salesperson within the company and they think of her like an actual person, which is phenomenal. And we also on the customer side have people talking to Laila saying, can I please know who your manager is? Because I'd love to write like a glowing recommendation about the service that you're providing. It's just this mind blowing stuff that really has kind of shocked us of how powerful the emotional stuff is. We've one where a lady rang up because she needed to cancel her appointment.
because her dog had gotten ill. So Laila obviously cancels the appointment and says, I hope your dog is okay, et cetera. The lady rings back in about a couple of months and the first thing Laila says is, how's your dog? And it's just this whole way where an amazing bond is built between the customer and the business through the likes of Laila. We also have suffered the likes of Action 24, a home security alarm system company, capabilities of customers getting in contact, trying to figure out how to reset their alarms or change different zones. So Laila ingests all of the technical manuals and helps them navigate those queries, whilst also dealing with, I'd love to get an alarm for my business or for my home, lead management quotations, all of this kind of elements. All the way through to Perry, which is a
Brian Kenny hardware device that's coming out for perimenopausal ladies and helping navigate that whole space and all the different questions that customers may have. So it's a broad spectrum.
Mark Fielding There's so much going on there. When you're talking, I'm thinking about, are people being starved of conversation, human contact? There's this big narrative about social media is not social, that people have destroyed all their links to actual real life conversation and people don't listen to each other. Everyone's just waiting to talk and talk, speak about themselves. And this just idea of, even if it's just an AI listening, it almost feels like, part, big part of this is the AI listens and just when someone listens, you feel needed and wanted and special and that's why they want to write glowing reviews for you because yeah, thank you. Someone listened to me for once.
The real reason
Brian Kenny It's so true. Like you think that the real reason, like when I really think about it is we dug into the economics of the whole situation. And it turns out that any growing business that comes like this inflection point where it no longer becomes economically viable to keep customer service in-house. And so we look for external solutions, like contact centers, different pieces of software, all these different things to manage it. And
We decide like, it's got to be possible to flip that on its head where you think about the, you know, if, somebody who is passionate opened a coffee shop, tomorrow, they'd be so proud of their coffee shop and they'd know every customer that's a repeat customer and say, Brian, you want your filter coffee like normal hope all as well hows Momntum going, et cetera. That should be like the status quo for large.
enterprise customer service, like an intimate relationship between you and the business and for them to really know and value who you are as a single customer.
Mark Fielding It'd be like the anti Starbucks you don't, their tagline could be whatever the name of the cafe is. We don't need to write your name on a cup because we know your name. It could work. 700 languages I read, is that correct? Laila can speak 700 languages.
The crazy experiment
Brian Kenny Yeah, we did this crazy experiment as well where in that Med spa actually there's lots of different colleagues who speak languages from Greek to Spanish to German and everything in between. So we did a little kind of video where all of the different colleagues that could speak different languages were ringing up and asking Laila questions and she kind of nails it of the park every time.
Really difficult to test when you don't speak multiple languages, but a phenomenal experience when a customer gets to speak in the language that they are in. Even now our voice products, so when an Irish customer rings an Irish number, it's an Irish version of Laila that answers the call. The impact of that is just shot through the roof as well. But we need to do better on figuring out the localization. So how do we know?
Mark Fielding Yep, of course, yep.
Brian Kenny they speak German or what is the best way like we can reverse engineer phone numbers and things like that. But knowing what language to start with is, is one interesting one that the team is working on at the moment.
Accents, dialect and swearing
Mark Fielding accents, regional dialect, slang, swearing, any issues there?
Brian Kenny Emotional awareness is probably the biggest one. Where, you know, if I say, hey, I'm Brian versus, hey, I'm Brian, like, and there's all of these different ways that you can say the same thing in a different context. And then there is the regurgitation of that emotional response back. And this is where all of our proprietary tech comes in the most. If somebody is upset about a situation,
It's embracing that, it's trying to ensure that the person feels heard and understood rather than feeling like they're talking to a robot. So within the context, knowing where our ORLM interjects different emotional aspects to how to say the sentence is really important. Also, like the use of emojis and other different kind of gotchas.
Mark Fielding yeah, you need emojis for the young people, don't you? And for the LinkedIn crowd. Where does Laila fail? What is she? I said she. Already, have to mobile laser. Where does it, where does Laila fail? What is it not good at at the moment? Apart from that, you mentioned.
Brian Kenny Yeah, so I think it's not a question in my mind of what is she not good at? It's for me, what is she not ready for?
Mark Fielding You said she, you said she then.
Brian Kenny Yeah, I always do.
Why Laila?
Mark Fielding On an aside, why Laila? Where did that name come from? Does that mean anything for you and the team?
Brian Kenny I could tell you one of two stories. So in one story, we could say that the word AI is in it as in L-A-I-L-A, which, yeah, but that's actually something that came about after I picked the name. So I can't actually grab the attention for that one. No, it just, sounded like a really cool name to me at the beginning. And it kind of, it's, I don't know, it just sounds like a really nice person that you could trust. So I went for that.
Mark Fielding Yeah, it looks good on the logo.
Brian Kenny I've forgotten the other question.
Mark Fielding What does Laila not knock out the park? What does she struggle at? What does she fail at or not fail, but struggle at?
Brian Kenny It's yeah, so again, like it's, it's, it's really a question of what is she not ready for? and it's, there's, there's certain situations where you really want to ensure that the potentiality of hallucinations or anything not going completely 100 % correct, isn't, isn't brought into the mix. So for instance, bank transfers or medical records or any of these types of things. Anything where she's going to interact with something that has a really high risk factor. just don't think that it's not necessarily even a momentum problem. It's a AI at large and a responsible problem where we need to kind of watch as the evolution of this technology increases to see when is the right time to test on certain things. Even similar, the Tesla's testing their autonomous driving and all of this. There's so much rules and regulations and stuff that we really need to be diligent about, that there's certain cases that it's just not ready for there yet.
The thought experiment
Mark Fielding Understood. Okay, well, let's do a thinking on paper thought experiment because I really like the idea of somebody coming to the thinking on paper website thinking on paper.xyz for all our shows and book club episodes and having a pop up of me or Jeremy modeled on our mannerisms and so and our accent and our humor interacting with visitors to our website. Let's take it a little bit further.
for the businesses and brands listening to this, okay? And then we have a range of merchandise which we sell on the backend of the websites, t-shirts, hats, we organise events, that kind of thing. Walk us through how we, and by that, others would incorporate Laila or some momentum product into that amazing website which I'd made myself on Elementor and WordPress.
Brian Kenny For sure. I think, yeah, so like our target customer or ICP is really anything that's high customer touch and really strongly powered by technology. think about airlines, car rental, restaurants, booking hotels, all of these types of things that is a, you know, it's a really kind of a fluid conversation that tends to happen. looking for this restaurant at this time.
So there's all of that, but in order to get up and running with Momntum and to deploy it on, let's say, the Thinking on Paper website, it's first of all, just getting your account set up. It's connecting it to Elementor or to whatever your e-commerce platform is, even going back to, let's say, the coffee shop scenario.
And then there would be a little chat widget on the website. There'll be a WhatsApp number. We'd provide you with all of the kind of phone numbers where everything is ready. You can connect in your Instagram, you can connect in your Facebook, and then we can train Laila on exactly what we'd like her to accomplish. So as people start interacting, they might say, I'm really interested in some fruity coffee beans or whatever it might be.
And Leila will look through all of the e-commerce elements and products that are available and say, yeah, let's get some coffee beans and a hat and whatever it might be. Here's a cart. And even if you have, let's say an e-commerce store plus you're, you know, have a restaurant booking engine, she'll transact across both of those at the same time, which creates one single transaction, but, you know, reconciles it in different areas.
So you're up and running relatively quickly. She'll train on all of the information that you give her. But the real power comes when I go back and say, the fruity beans were lovely, but I'd love something that is a little bit more toned down. And she'll know all of my buying history and know how I felt about the last beans and whatever it might be, and recommend some other products that are within your e-commerce store.
Brian Kenny to try out or even if you have subscriptions and things like that as well.
Mark Fielding It's the hyper personalization of the internet of the buying experience of the brand customer experience that we've spoken about this quite a lot I'm thinking on paper this hyper personalized. So the brand you can personalize the brand voice. You can personalize the brand voice to every individual and create an individual experience. We've spoken about momentum. We've spoken about Laila I want to talk about what comes next and maybe if this is your personal views on this, how you see things evolving, agentic AI, crypto payments, AI is speaking to either, how will Laila fit in or how do you envisage it fitting into this agentic web where in fact there are no humans in the loop except the customer?
Brian Kenny Yeah, so I think one of the most interesting things that we're working on at the moment is, know, Laila is very customer focused at the moment. However, on the back end in the dashboards that we have for the business owner, it's all very, you know, tables and reports and charts and everything. You can see, for example, different cohorts of customers who are falling out of love.
or there isn't a strong relationship with the business. However, what we're playing with at the moment is a Laila, but for the business side, where Laila can interact. you can imagine the business owner in his or her car driving towards the office, rings up Laila and Laila is walking through all of the customer interactions that she had yesterday. You know, some customers really love this or I had this challenge that happened to me yesterday.
And some of the customers in Miami really struggled with X, Y or Z or whatever it might be. And I've also noticed Brian, that there's these 10 customers who are starting to fall out of, you know, a high relationship status with your business. What I'd recommend is that I write like a tailored offer to each one of those individual people based on the connection, the relationship that I have with these people to say, you know,
Brian, come back in for regular appointment and we're going to add in an extra treatment on the house or whatever it might be. So she'll write external outbound sides of messages to strengthen bonds that she's seeing dip off, but also to really bring business owners up to speed on how the customers are feeling about their business in phone calls, in WhatsApp, in whatever way that they would love as well.
Pricing
Mark Fielding Before we get into the big, deep philosophical questions on this, any data on the ROI, the return, like how much money this can, is, would, could save, make businesses?
Brian Kenny Yeah. So at the moment it's, it's, don't know if we price it wrong or whatever it be, but we're tracking at a 4,000 % return on investment. So this is, think Laila is doing over a million euros in bookings, know, in revenue generation for our clients every two months at the moment. And, you know, we only started in January. So
I think we're on track to hit six or seven million in revenue for our businesses at the moment. And then of that, the way that momentum charges our clients is number one, an active customer. how many active customers you have in momentum every 30 days. And number two, how many times Laila does something as in if she's booked an appointment, canceled an appointment or taken an action on behalf of your customers.
will charge for that also. So it's a usage based pricing model that's going in there. But the ROI has been insane.
Mark Fielding Will it do social media? Will it act as a social media manager as well?
Brian Kenny So we haven't figured that out fully yet. But I think our Northstar is building that relationship with the customer and the business. think social media advertising and Look-a-like audiences and all of these types of things are probably going to drop off in the future and be replaced by strong relationship bonds with customers and really relying further on the WAP kind of company style, the referral piece where people start to promote customers or sorry, businesses that are really performing well on customer service. think it's two out of three customers that have a bad experience with a business just won't come back. So the churn, I think it's something like 136 billion in North America alone.
is the churn rate that's happening each year. So this is all, in our opinion, completely avoidable churn if we just really double down on building relationships with those customers or helping our clients to.
The Gartner number
Mark Fielding Okay, thank you, Brian. So in a minute, we'll be asking the Kevin Kelly follow over question. a little bit of a, before we get into this, so a recent Deloitte survey found that employee resistance is one of the top barriers to AI adoption. Although on the same hand, on the same token as that, 65 % of companies are using AI in 2024, 25. And I think that's probably an underestimate.
Gartner says 10 % of all agent interactions will be automated by next year, 2026. 17 million customer service agents in the world, maybe more, maybe less, probably more. How do you approach and how should the AI industry at large address this pushback, address this problem? Like what do you say to the customer service agents who are listening to this saying, hold on, but that's my job, that's what I do.
I know it's a big question, what are your thoughts?
Brian Kenny I would say probably one of the most profound things that's ever happened in my life is that my parents bought a computer for the household when I was like six or seven years old. And computers back then were extremely expensive, but it became... I'm not giving that away. But it was a vehicle to really become technically literate.
Mark Fielding How old are you, Brian?
Brian Kenny from a young age. Now I've never been to college, but I've self taught myself to engineer, to design, to build products and build the first MVP of momentum. And I then look at, let's say places like Abu Dhabi. In Abu Dhabi, the government provides a chat GPT subscription to every citizen. And that again, it's just creating that AI and literacy within the people. I think
It's such an amazing tool. It's profound in its capabilities and what we can do because of it. To the point that the revolution that happened when we got computers at our fingertips is happening even again, but stronger and deeper. And we need to embrace that. It isn't about, you know, pausing time for how it stands right now and saying, okay,
What is the impact that it's going to have on the current workforce? But more looking at it as an evolutionary step where what is the current workforce now capable of doing by leveraging this phenomenal tool that's at our fingertips? And I think once we lean further into creativity and imagination and what humans are just substantially amazing at doing, it unlocks doors and potential of where we can go in the future.
Should our kids study it?
Mark Fielding So we should embrace and as a parent get our kids to study it, to use it, to try and this is about educating the populace that they can use it. It's not going anywhere. So they need to use it, embrace it and pivot. We're talking about pivoting, changing, augmenting. We speak a lot about augmentation of skills, augmentation of
Brian Kenny think there's always a fear initially. And to be honest, I'm probably one of the most laggard technology adopters that's out there in a way, because I don't want to change things too drastically too quickly. And it took me a while to really jump on the AI bandwagon. But now I have an 11 year old and an eight year old, and they both, you know, got full chat GPT subscriptions.
Lily, my eldest, has this strange concern about school where she's asked to do something, whether she's allowed to use ChatGPT to ask questions and to research and things like that. In our house, it is absolutely the way to go. To now, a point where she's starting to, you know, we went on holidays recently and she'll ask ChatGPT how should I pack my bags? Write me out a checklist and try to help me to not forget things. So it's amazing to see once people have leaned further into what this technology can unlock, where it can bring them and what kind of bandwidth they can leverage their mind over to chat GPT rather than being too worried about
Well, first of all, where are we going to go and what's the, you know, the fear side of it? I think we do need to be responsible about it. think it's really
It's really important that we get some sort of regulation around AI at some point in the future, that we really help to...
Brian Kenny I wouldn't say curtail, but push some sort of responsible attitude on commercial innovation moving maybe a little bit too quickly. Like if you have your open AIs and Anthropic and all of the people who are really pushing to go, fast can we get to AGI releasing models on a really fast cadence? And it becomes a commercial land grab question rather than a responsible one.
then I would like to see regulation placed around that and responsibility placed around that too.
Mark Fielding What would you like that to look like? that be a global AI strike force? Would it be country by country? how would, do you think that would work best?
Brian Kenny I don't know is my answer at the moment. mean, it's, it's, I've been keeping up to date on the EU AI act and what's going on in the U S and some of the other aspects that are coming out. in a way, some regulation can curtail, commercial viability and speed, in what we can do with the technology, but it's, it's a really, it's a tough, problem to solve. but I think we'll get there.
I've always seen the world as more of a global one, of which, you know, as amazing as different countries can be, I think we're all global citizens. I think AI in a way is a testament to that, that we need to figure this out as humans, as we say, in order to really double down on where it can bring us in the future.
A personal question
Mark Fielding Can I ask you a personal question? you might be able to help me on this. So my daughter's not quite as old as yours, but, she is a, she loves drawing like a lot of kids. She's really, she draws her own cartoons, animations. She draws still life. like any excuse she gets, she'll draw. And she hasn't used AI yet. And my initial thought would be that image generation would be a wall.
an image generator would be the best place to start. But then I think if I give her the keys to that, will that stop her drawing or will that encourage her to draw something different? It's a big battle in my head about whether to give her the keys to this or not. have you noticed any change in how they create?
Or do you think I will? Or do you think I'm worrying about something and there isn't a concern?
Brian Kenny No, I think it's such a valid, really, really strong point that a lot of people really need to think about at the moment, which is how much of your creativity and imagination are you handing over to an LLM in order and completely willing to take the LLM's output and call it, you know, that's the way that I'm going to go. I think that that's completely something that we need to be very cognizant and aware of.
Stifling creativity is something that I would completely push against. So when it comes to mass knowledge search and summarization and capabilities within AI to see things that you can't potentially see right now, to jog the memory to really understand and understand what's possible, AI is phenomenal at doing it.
You can't ask it to do your homework. know, you need to understand what it's responding and you need to have the competency to take it in, learn it, ask questions about it and understand it. And if there's concepts in there that you don't understand, then you can ask it to, you know, bring it down a few levels, you know, teach me like I'm I'm a labrador type of thing or whatever it might be.
I think it's phenomenal when you lean into it as a self-improvement or self-educational tool, but handing over your fiduciary or your responsibilities or your question or your work to it and saying, just go do it and I'll see you later, is not right path. You can see this with a lot of coding at the moment where we call it vibe coding, but if it's returning all of this code and you're not...
sense checking what it is or making sure that it's adding it in the right spot or it's the correct interpretation of what you wanted, then it's not the right path. You're going to end up with this whole spaghetti code base that isn't where you need it to be. So a great tool, a great addition, but it shouldn't substitute for your daughter's creativity, your drawings, your, you know, your kind of time to be creative.
Mark Fielding Wonderful answer. Thank you, Brian. Thank you for thinking on paper with us. I think you've left our audience with a lot of questions, some great ideas, any businesses, brands looking at this and wanting to update or upgrade or think about how AI is used in their workflow. I think you've given them some good ideas as well. I appreciate your time. got one last thing for the disrupters and curious minds out there. We've got a book club at Thinking on Paper.
where we read books slowly about technology and the impact. At the moment, we're reading Empire of AI by Karen Hao Dreams and Nightmares in Sam Altman's Open AI. Previous books include Nexus, Yuval Noah Harari, The Design of Everyday Things, Don Norman, Irreducible, Federico Faggin You can check all those out at thinkingonpaper.xyz. We record those live. So if you are reading Empire of AI, check out the website and you can join us.
And on that Brian, so Kevin Kelly left us a question that we now ask all our guests at the end of thinking on paper. It's a deep question. What should humans be and how does technology help us get there?
Brian Kenny Humans should be explorers and adventurers and continually challenging and pushing themselves to go further. And technology is probably one of the biggest aids in doing that. It unlocks rooms that we've never been in. It creates potential discoveries that we've never even known existed. And it constantly challenges us to get to the next rung in the ladder.
I meet that with excitement. I meet that with, wonder where we're going to be in 10, 20, 100, 1000 years. And as we grow as a species, embracing technology to be the aids in allowing us to discover more and not allowing, but empowering us to discover more is such a wonderful thing to have. And I'm excited to be alive right now.
Mark Fielding I love that analogy of unlocking rooms that we've never been in and then stepping into that room and letting your brain, your imagination see what's there. I love that. Thank you. Do you have a question that you'd like to leave for future guests of Thinking on Paper that we can add to our canon of questions?
Brian Kenny Yeah, I mean, not thought about too deeply, but I would say one question that I've often tried to noodle over myself is really should an AI agent acknowledge themselves or introduce themselves as an AI before you interact with them?
Mark Fielding No, is my off the cuff answer to that? think that personally...
long term. So does that mean in a hundred years, in a thousand years, if we're still around that you're having a conversation and the first thing they say is, by the way, I'm not human. Although maybe they should, I don't know actually. Maybe I've changed my mind straight away.
The EU AI Act
Brian Kenny It's crazy. it's, it's the EU AI Act is now saying that, that, you know, we should, or it's something that they're fighting for to be a part of the act that the, whenever you open up a live chat or WhatsApp or whatever, that it should say, hi, I'm Laila, an AI assistant. However, I don't know if that's the correct path. I mean, from, we've even shied away in momentum from.
you know, an AI kind of listing out bullet points of information, because a human would never do that. Like if you're looking for slots, it's going to be written in a text and sometimes poorly formatted and everything else. And I think we want it to be as human-like experience as possible. But it does leave us with this question of should Laila always introduce herself as an AI or, which is our current path or current situation.
Whenever you ask Laila, is she an AI? She'll of course say yes.
Mark Fielding Should have been a friend, great question. Listeners, let us know in the comments wherever you are consuming this podcast, YouTube, Spotify, Apple, let us know in the comments, they just come out with it straight away? I am Laila, I'm an AI agent, or should they not? Let us know. And on that note, I would say thank you, Brian. Be disruptive, stay curious, keep thinking on paper, and we'll see you next week when Jeremy will be back in the house without Botox.
Mark Fielding with news of his secret mission. Thank you very much.




