How do you make sense of an industry that is changing at a pace few predicted, especially with SIGNAL London still fresh in our minds and Twilio unveiling the next stage of its vision for customer engagement? That question sits at the heart of today's conversation with Peter Bell, VP of Marketing for EMEA at Twilio, who joined me to unpack what the past year has taught both companies and consumers about AI's role in shaping modern experiences.
Peter begins by grounding everything in a single, striking shift. Only a year ago, AI-powered search barely registered in global traffic. Today it accounts for around a fifth of all searches. That leap signals a broader behavioral shift as consumers move instinctively toward conversational interfaces, which, in turn, leaves brands with a clear message. The clock has moved on. AI is no longer a nice-to-have. It is a direct response to how people now choose to discover, question, and buy.
Our conversation turns to the gap between customer expectations and the experiences they receive. Peter discusses why brands often struggle to integrate channels, data, and AI coherently. He explains how first party data has become the anchor for any serious AI strategy, why generic public models cannot solve brand-specific tasks, and why the most successful teams start with simple, tightly scoped problems. A password reset may not sound glamorous, yet it is the kind of focused use case that teaches teams how to govern data, automate safely, and build confidence in the process.
We also spend time on branded calling, RCS, and the evolution of voice. Peter breaks down what modern messaging now looks like and why trust sits at the center of every interaction. His explanation of Conversational Relay shows why natural voice exchanges finally feel within reach after years of frustration with rigid IVR systems. The thread running through all of this is clear. Consumers want speed and clarity, but they want reassurance too, and brands need to honor both sides of that equation.
Later in the conversation, Peter makes one of the episode's most compelling points. Brand visibility has become harder, not easier, because much of the early research now occurs within AI tools. Buyers form opinions long before they speak with a sales rep. That shift explains why so many B2B companies are returning to high-impact brand channels, whether that is F1 sponsorships or other standout moments that keep them in the initial consideration set.
We close with the topic that Peter believes will define the next stage of enterprise AI. Model Context Protocol. MCP has emerged as a quiet breakthrough, enabling LLMs to access data across CRM systems, files, and other software through a standard protocol. This removes one of the biggest blockers in AI projects: the practical challenge of connecting disparate data to a model built for a specific purpose. As Peter puts it, MCP gives companies a realistic way to make the special-purpose models that deliver reliable ROI.
It is a wide-ranging conversation shaped by SIGNAL London's announcements, the evolving customer journey, and a year in which AI moved from curiosity to expectation. I would love to know what part stood out most to you. Are you seeing the same shifts Peter describes in your own business, and how are you preparing for the year ahead?
Useful Links
Interact with the Inside the Conversational AI Revolution report.
Learn more about the Signal event
Connect with Peter Bell, VP of Marketing for EMEA at Twilio.
[00:00:04] - [Speaker 0]
Greetings and salutations, and welcome to another episode of the Tech Talks Daily Podcast. And today, I'm joined by Peter Bell. He's the VP of marketing for EMEA at Twilio. And in our conversation today, we're gonna explore how AI is reshaping customer engagement, where brands are getting it right, where they're getting it wrong, and what it really takes to keep interactions feeling human in this increasingly automated world. We've all got bad examples of what we don't want it to look like, but there are plenty of brands getting it right.
[00:00:44] - [Speaker 0]
And these are the stories I wanna hear more about and help you unlock your approach in 2026. And one of the big reasons I'm excited about bringing today's guest on is I was unfortunate to miss the recent Twilio signal event in London. But from people I know that attended, they were telling me that the energy in the room felt like a reminder of what happens when you bring builders, developers, and business leaders into one room to share the same curiosity. But one of the most valuable aspects of the event was how developers, business leaders, innovators from some of the most influential companies. What happens when they all come together for a day built around action rather than just theory?
[00:01:25] - [Speaker 0]
And one of the things that stood out to me is that very seldom do people from so many different areas get to collaborate this closely. And a combination of hands on building, open discussion, and just genuine curiosity, I think, can create a momentum that can carry through every session of an event. Kind of gathering that reminds you what happens when talent, intent, and opportunity share the same room. And I'd like to try and bring some of that to today's conversation. And from trust and first party data to the future of voice and branded messaging, Peter is gonna bring with him today a grounded and thoughtful perspective on what is changing inside the modern experience and what he's seeing and hearing with conversations with customers around the world.
[00:02:15] - [Speaker 0]
And we'll also be asking him what trends he will be paying close attention to as we move into the next phase of AI adoption and into a new year. Before introducing today's guest, I just wanna give a big thank you to my friends at Denodo who are helping enterprises make sense of the data world. For example, are you overwhelmed by data chaos? Because between AI hype and lake house sprawl and siloed systems, it can feel impossible to keep up. But Denodo is helping you cut through that noise.
[00:02:52] - [Speaker 0]
They unify your data across clouds, apps, and sources so you can power trustworthy AI and accelerate lakehouse optimization, delivering data products that scale self-service for every business unit. Now whether you are a CIO, architect, or analytics leader, Denodo will help you engage faster and deliver real results. And with their partners, you can also modernize without disruption. So if you're finally ready to make sense of the data world, visit donodo.com today. But now it's time for me to officially introduce you to today's guest.
[00:03:33] - [Speaker 0]
So a massive warm welcome to the show, Peter. Can you tell everyone listening a little about who you are and what you do?
[00:03:41] - [Speaker 1]
Yes. Of course. And thank you, Neil, for the invitation. My name is Peter Bell. I'm VP of marketing for EMEA for Twilio.
[00:03:48] - [Speaker 1]
I've been with Twilio around three years, which is just that anniversary is just coming up actually at the end of the year.
[00:03:53] - [Speaker 0]
Well, it's a pleasure to have you join me today. And I think many people at the moment still see AI as just another passing technology cycle even though we've been talking about it for three years now, and the technology dates back much, much longer. But Twilio describes it as a fundamental shift in how companies create and scale experiences. That's one of the things that I learned from my research before you came on the podcast today. But from your vantage point, where are you seeing the clearest structural changes already taking hold in things like customer service and brand engagement?
[00:04:27] - [Speaker 0]
Because I think as we get past that three year mark since ChatGPT dropped, we're now looking for ROI and measurable impact. So what are you seeing here?
[00:04:35] - [Speaker 1]
I think three things. The first is, look, whether we like it or not as brands, consumer behavior has changed. You know, if you go back even twelve months, the amount of search traffic which was being served by LLMs was was negligible, let's say, than 5%. I saw a number earlier this year. I think it was a Google number.
[00:04:57] - [Speaker 1]
It's around 20% now of the global search market. And in a market that big, you know, the billions and if not hundreds of billions of searches every day, to see a percentage point jump like that in in a twelve month period is is really unusual. So we've definitely seen the consumer jump in, and that means for brands, you know, this is this is not an if, it's a when. The second is, you know, there's real promise, and a lot of the conversation, a lot of the the demonstrations at the moment, including us, really focus on the customer service. We know that customer service, there's been a disjoint for many years between how brands self evaluate their how good they are at customer service and how consumers would evaluate them.
[00:05:44] - [Speaker 1]
You know, the two numbers are generally a long way apart with consumers being very critical of customer service, and AI offers a real opportunity for brands to close that that gap. And the third, because we're all very excited about ChatGPT and its brethren, but we've somewhat forgotten about machine learning
[00:06:02] - [Speaker 0]
Yeah.
[00:06:02] - [Speaker 1]
Which is obviously a branch of AI. And in the background, you know, there's all sorts of ML models working out, you know, what's the next best thing, how to nudge us into purchasing something, what should the next what should the price point for a particular product be over time. And that's kind of forgotten, but there's a there's a huge amount of of that work in the background.
[00:06:24] - [Speaker 0]
And at Twilio, you often talk about every interaction needing to feel seamless and most importantly, genuinely human. So how do you reconcile that ambition with the the reality that so many brands still struggle to join up channels, data, and AI in a way that feels coherent to the customer. We've I suspect we've all got bad examples of of that, but there's plenty of good too. So tell me more about that.
[00:06:49] - [Speaker 1]
And we have. And, you know, the certainly, we have high aspirations and also high expectations of the long term impact. And I forget the I think the old saying is, you know, we overestimate what happens in the short term and underestimate what happens in the long is going to happen in the long term. And I think this is a classic case of that. There's some there's some huge overstatements of where AI will be in the short term.
[00:07:15] - [Speaker 1]
But we do believe at Twilio in the long term that it is it is a game changer. And it's it probably won't happen as quick as we would all hope for, but it will happen.
[00:07:30] - [Speaker 0]
And your leadership has been vocal about the first party data being nonnegotiable and owned by the customer, not the model providers, which again is refreshing to hear. And in practical terms, where are organizations still getting this wrong, and what are the risks if they continue to rely on fragmented or poorly governed data? One of the reasons I ask that is every tech conference I've been to over the three months, the phrase I keep hearing is no AI. No data. No AI.
[00:07:58] - [Speaker 1]
Yes. And I would agree with that statement. And first of all, we should all regard our first party data as something precious and to be we are custodians of it, and we should treat it with care and respect, not just for our reputation and the trust of that consumers place in brands, but also there's quite a bit of regulation about this topic as well. So, you know, if you wanna stay on the right side of your customers, then you clearly have to protect your data. And the second thing, though, in terms of bringing AI to life and creating the type of customer experience that we all aspire to is the public LLMs are just trained on everything on the Internet, good and bad, acceptable, unacceptable.
[00:08:45] - [Speaker 1]
They have they basically, just as a search engine does, crawled everything that's out there, which actually provides great utility if you've forgotten something or you want to know when the public holidays in Germany are or if you're planning a trip, whatever it might be. But it's it's too general purpose to, like, solve a problem with. Because to solve a problem, you need very specific data to train the model on. I'll give an example. If you've got a model that is let's say you're a travel agent or, you know, is involved in the travel industry, you want to train your model just on travel because that's the that is the problem it's addressing.
[00:09:28] - [Speaker 1]
You know, it's when there's a flight this such a city, Is there a direct flight? Is and can I go indirect? What will what are the typical prices? It doesn't need to know about other things. It doesn't need to know about gaming or how to or a password reset or, you know, the price of copper.
[00:09:48] - [Speaker 1]
It doesn't need to know any of those things. And you give if you give it good and specific data, you'll get good and specific answers back. I think you'll get much less hallucination, I think, the the common phrase. And that's the work you have to put in. You have to make the data available and then train your own model on the data that for the problem you're you're actually solving.
[00:10:13] - [Speaker 0]
100% with you there. And something else we've gotta talk about when we talk about AI is trust because trust does feel fragile right now, especially with rising scams, SIM swaps, and consumer fatigue, etcetera. So how do you at Twilio balance that push for faster and more automated engagement with the need for security resilience and emotional reassurance? Because, again, these are things that often get missed, aren't they?
[00:10:40] - [Speaker 1]
Yeah. And this I would start by understand the problem you're solving. You know, there's a fairly high failure rate around AI projects, particularly the on the LLM side. And, like, one of two things happens. Either you set the problem too broad Mhmm.
[00:10:55] - [Speaker 1]
And it's just you're overambitious, or you don't really understand the problem. If let's take just a a really, really basic example. The process behind the password reset, you can and let's say you just simply want to automate that. Because a lot of reasons we as consumers wanna talk to brands is to get a very specific task completed. And we can understand if we understand the problem we're solving, the process that needs to exist behind it, and the data that's required in order to address that, then you're gonna take a lot of the risk and ambiguity out of the implementation.
[00:11:34] - [Speaker 1]
And ultimately, you also know which data is which data you're going to have to use. And you'll have policies and governance around that data, and you can make a decision. You know, password reset might you know, if you're for most companies, I think that's probably automatable. But if it's health data, you're gonna want to work with much, much greater care. You're gonna you know, it'll be a more complex process.
[00:11:57] - [Speaker 1]
There are probably more checks and balances. It may not even be suitable, quite frankly. It comes down to what is the type of data? What's the sensitivity to it? You know, what's that original problem that you're solving, a use case in tech talk?
[00:12:11] - [Speaker 1]
I think then you can navigate this. But being wildly ambitious, trying to create some agentic, solvable, catch all It's really hard to map out the data, the processes, and therefore the policies that sit over it.
[00:12:26] - [Speaker 0]
And if we all also look at branded calling and RCS, which is gaining ground in The UK and across EMEA, What does a a genuinely modern voice and message experience look like in practice for what you're seeing here? And and how should brands maybe rethink tone and timing in this new landscape? Because we are recording this conversation at that magical time of the year where we're starting to think about doing things differently as we approach 2026. But anything that you'd recommend here or advise or share any insights around?
[00:12:57] - [Speaker 1]
So I think there's first of all, let's visualize what it looks like because perhaps not everybody listening can picture what, you know, what what it what is branded calling, what is RCS. Yeah. Fundamentally, the experience you get is one which look at think a lot like Instagram in some ways. It's visual. You can present your product in a beautiful way.
[00:13:17] - [Speaker 1]
It's shoppable. It can be conversational, so you can get into a conversation with a service assistant either pre or post purchase, whatever the scenario. So it's very modern looking and all the things that you you would expect as a consumer. Therefore, you're in control of the presentation and the tone of voice. You've got much more flexibility as to how you show up.
[00:13:39] - [Speaker 1]
And the important thing is, you know, that brand that's presenting itself as that brand really is that brand. Mhmm. And in a world where trust is being eroded beak because of impersonation either by the consumer or by the brand, it's really important that I know this is the this is the brand I'm interacting with, and then that is authenticated. But conversely for the brand, they know they're doing they are talking to me, and they know and I'm not a fraudster if it's if it's an ecommerce transaction that it is indeed me who's shopping for this. So you inherently get to build far greater trust in there.
[00:14:17] - [Speaker 1]
It's a rich experience. You know, if you're a merchandiser, you can lay your product out beautifully. If it's customer service, you can conduct a then you can conduct a conversation over time when it's convenient. So just really modern and trusted. As for frequency, I'll just make a plea on befull on behalf of all marketers.
[00:14:38] - [Speaker 1]
I speak as a marketer to other marketers here. Do you think about the frequency of your of your communication? I probably don't need to hear from you daily, weekly, a push, perhaps once a month. And if you've got if you've got something really important to say, break that. But there's some really bad practice at especially at Christmas time when the frequency just seems to go up and up and up from brands.
[00:15:00] - [Speaker 0]
And voice AI is also something that's evolving from simple scripts to sophisticated and adaptive agents. I've got to ask, like, what excites you most about platforms like conversational relay? And, also, where do you still see caution or complexity that brands need to respect and avoid losing the voice of their brand as well?
[00:15:22] - [Speaker 1]
Yeah. And I think first of all, let me just take a moment to explain what conversation relay is. Yeah. It is a text to voice and voice to text tool. And most of the information we may be looking for is probably written down.
[00:15:37] - [Speaker 1]
It'll be held in a database somewhere in some system of records or maybe, you know, it'll be written down fundamentally. And yet the most natural way for us to interact is voice. We're human at the end of the day. As BT used to say a long time ago, it's good to talk, but that probably ages me. But it's a very it's our first form of communication.
[00:15:59] - [Speaker 1]
So that ability to simply talk to a brand and then for the technology in the background to go read and then turn that back into speech in a very natural way is really, really exciting. I think we've we've had aspirations around this in many false starts with voice interfaces over the years, and I think we finally have the tools to really make it natural sounding, free flowing, and just a very rewarding experience as opposed to let's just leave it at some very frustrating experiences in the past where where companies have tried to automate in this way the good old IVR has tested our patience and sanity over the years. And the sooner we can leave that is that behind and move to much more conversational. And I think conversational is the right word here. You know, voice should.
[00:17:00] - [Speaker 1]
It but it's a I think the word conversational is the one I that best captures what we're talking about and the excitement around it.
[00:17:11] - [Speaker 0]
And I'm cautious from those conversations you're having with your customers, etcetera, on that very topic. Is there any nervousness or caution about moving forward or or sacrificing their brand voice? Or are there any concerns there at all or or apprehensions before they come on board?
[00:17:29] - [Speaker 1]
Yeah. I mean, I you know, we know from the research the the research we've just done back something like half, I think it's 56% of of brands have projects underway. And that's I think that's reflected by the shift in consumer behavior and that very aggressive adoption in the search market. So it's clearly there's a lot that a lot but like everything, you've got to crawl, walk, run. And we're probably in the crawl phase right now.
[00:17:59] - [Speaker 1]
And with that, you're learning. And I think the key thing is that I would observe is just define I said it earlier, but define the problem. Yeah. But what is the what is the problem you're solving? That will set scope.
[00:18:14] - [Speaker 1]
It will highlight any concerns of a security. Can either put them aside or flag them legitimately. Like, let's not do this as our first project. Let's look elsewhere. And the risk of it running amok, therefore, is is much, much less than if you go for some huge ambitious agentic automation where you've got agents working on behalf of their agents, and it's all multithreaded, and you don't quite know what's happening.
[00:18:42] - [Speaker 1]
You know, we need time to I talked about LLMs earlier. Those specific purpose LLMs need to be built inside companies that need to be trained, that need to be tested, and they're the foundational piece that will take away risk.
[00:18:59] - [Speaker 0]
Yeah. I completely agree. I think three years ago, there was a lot of bandwagon jumping AI for AI's sake or tech for tech's sake. And I think right now, there is that big shift as well, which is just we start with problem first, tech second, not the other way around. And as we continue to look ahead and more organizations continue to inevitably adopt multi model AI strategies, what do you think becomes the defining factor between those brands that strengthen their relationships and those that quietly erode them through over automation or misplaced priorities.
[00:19:33] - [Speaker 0]
Do you think that gap will quickly widen?
[00:19:36] - [Speaker 1]
I do think it'll widen, and you can all you know, particularly, I work in b to b and so business to business marketing rather than business to consumer, and you can already see a shift in myself and my peers' distribution of marketing budgets. You're seeing a lot more money go into quite traditional brand channels. So go look at f one. You'll see lots of software companies on those cars. And it's not because every tech CEO is a f one fan.
[00:20:08] - [Speaker 1]
It's because the importance of brand has never been so it's so important in b to b marketing. And it's this collapse of the consumer of the customer journey where we used you know, we could reasonably plot out as marketers what your customer journey would look like as a buyer and where you would go, what that would typically look like, and it'd be multistage. That's just gone. Yeah. It's it you know, the first thing we do is go to our favorite AI tool and start chatting with it and building reports of the like that information you have to pay for.
[00:20:47] - [Speaker 1]
And we if you're educated, you know, not to trust every single thing in there and then the pro you know? But there's a huge amount of this done out of sight now, but mind. And, of course, it therefore, if you're not in that initial consideration set, you're you're never gonna get the call.
[00:21:03] - [Speaker 0]
Yeah.
[00:21:04] - [Speaker 1]
And can you know, you can only be in the consideration set if people know who you are. If I was to challenge you and don't worry. It's a rhetorical. But next time you replace your car, you probably already got three brands in mind. Yeah.
[00:21:18] - [Speaker 1]
The chance of another one breaking in is quite slim. Yeah. And it's no different regardless of b to b, b to c. So that brand building, it is a big shift there. And I think there'll be people who fundamentally get that, and then there'll be those that don't make the shift and don't make the investment.
[00:21:39] - [Speaker 1]
They'll become somewhat invisible. Then, of course, what is the experience you give? Because now it's all about what what is the next what does that in first interaction like? Because trust is something you know, trust don't ultimately and that first step is a a leap into the unknown. I don't know you.
[00:22:01] - [Speaker 1]
I'm prepared to enter into a transaction or a conversation with you. It needs to go well in those first steps. And that's where have you automated well? What do you put a human in the loop? Indeed, do you put a human out front first?
[00:22:17] - [Speaker 0]
And such a powerful point. And, obviously, I mentioned well, I mentioned a few moments ago that we are at the end of 2025. Gonna be a lot of people thinking about what they should be focusing on next year, what they're gonna be doing differently. From a personal point of view, you've been to a lot of tech conferences. You have the ear of customers located all around the world.
[00:22:36] - [Speaker 0]
You must get access to so many different trends around conversations and what businesses are aiming for. Anything in particular that excites you that you're gonna be watching closely next year, whether it be tech trends or the kind of problems that businesses are going after? Anything that you're particularly keeping an eye on next year?
[00:22:54] - [Speaker 1]
Yes. There is. It's gonna sound really nerdy, but it is model context protocol Right. Or MCP. And at the moment and we this touches on your early questions, which is fragmented data.
[00:23:11] - [Speaker 1]
How do I train that model? You're making you know, you've glossed over it. It's not as hard as it's it's not as easy as it sounds. This problem's been understood by the tech industry, and the MCP server standardizes the way that an LLM accesses data in the in the disparate systems. So regardless of whether it's in your CRM system, in on some server, in a particular file format, you, the LLM, can access all of these different systems through this standard protocol.
[00:23:48] - [Speaker 1]
And there's right now, there's an arms race of people building MCP servers either for their own for their own software. Let's say they're a CRM vendor. They could be a marketing automation vendor. But you've also got a lot of the people who do all of the kinda e t enterprise, ETL kind of which already addresses that problem of how do you get data to flow through your organization. So all these third parties are also building MCP servers for most of the common software out there.
[00:24:18] - [Speaker 1]
And it's a really exciting space. It's quite early days, but we're seeing rapid adoption. And this solves perhaps the single biggest barrier, which is how do you get the data accessible to the LLM so that you can build that special purpose model? And this does it. And it's it's a really important kind of unlock so that we can we can realize these aspirations and and the excitement around it.
[00:24:47] - [Speaker 0]
And as you said that, and especially around the adoption and the the interest there, I can hear people almost googling as we speak. Well, I think we'll probably divide the audience. Half will be Googling. The other will be talking to an AI agent to go a little bit more. But anyone listening wanting to connect with you or your team or just keep up to speed with some of the big announcements coming out of Twilio.
[00:25:10] - [Speaker 0]
I know you've had a a big conference recently, etcetera. So where would you like to point everyone listening?
[00:25:15] - [Speaker 1]
Honestly, it sounds a little glib, but the usual place is certainly the website. In fact, if you think this is all hot and you're listening to this, and it's like, well, Twilio, you would say this, you're you're invested. Go to our website, twilio.com, and it's the the AI revolution report. And on our website, you can chat with the report. You can read it if you wish, but you can chat with it.
[00:25:40] - [Speaker 1]
The first thing I asked it was, I can't be bothered to read this. Tell me the key point. That saved me downloading a PDF and never reading it. And I act and I'm obviously making fun of our own work, but we all do this. We we download stuff with great intent.
[00:25:54] - [Speaker 1]
So, yeah, go to twilio.com. If it's not still on the homepage, you're looking for inside of the conversational AI revolution, and just experience it for yourself.
[00:26:06] - [Speaker 0]
Oh, wow. I'm gonna be checking that out. I will link directly to it. I will find it and do just that. And it's interesting how so much has changed.
[00:26:14] - [Speaker 0]
Don't know if you saw in The UK a couple of days ago, there was the big budget over here, which was leaked early. I think it was, like, an hour earlier sort of thing. So I just took a copy and put it into chat GPT exactly the same with the what are the key points I should be aware of out of this? You know? And it's that how we digest information and and question information is changing so much, but more than anything, just thank you for coming on here and sharing your story and the work that you're doing.
[00:26:41] - [Speaker 0]
Thank you so much.
[00:26:42] - [Speaker 1]
That's alright, and and thank you for the opportunity.
[00:26:45] - [Speaker 0]
Wow. I really genuinely enjoyed my conversation with Peter there. I think his perspective on where customer engagement, AI, and brand trust are heading. And one moment that stayed with me was his take on model context protocol or MCP, And how it's much more than a trend and adoption increasing, it's more about how it can fundamentally change how large language models connect with real enterprise data across systems. And listening to Peter, I walked away inspired.
[00:27:20] - [Speaker 0]
It feels like one of those developments that is quietly reshaping how businesses build intelligent experiences without compromising control or trust and doing it all under the radar. So I, for one, whenever I hear MCP or model context protocol, I am gonna be thinking of Peter. And when you encounter it in the world, remember you heard it here on this podcast first. And if you're somewhat of an expert in this field or you've got a particular insight experience or expertise, consider this an open invitation for you to come and join me and dig a little bit deeper on MCP. I'd love to make that happen.
[00:27:57] - [Speaker 0]
And as always, I'd love to hear your thoughts on anything we covered in today's discussion. My username is just at neil c hues on LinkedIn x Instagram. Nice and easy to find there. And please reach out directly. Let me know what stood out for you.
[00:28:13] - [Speaker 0]
But more than anything, thank you for stopping by again, listening to tech talks daily. And if you've enjoyed yourself, why not leave a rating and review and join me again tomorrow? We'll do it all again. Hopefully, I will speak with you all again then. Bye for now.