Good morning, and welcome to Reverb. In a moment, we are going to dive into the future, talk about what is next, but before we do, I am going to roll it back. I am going to ask you to come with me to junior high and a science experiment. Let us see if I can do it. You hear that? That is 440 hertz of sound reverberating through this room. This is what we call reverb. It is a sound that travels, fills the room, connects us all, and right now, there are thousands of you online and many here in Raleigh listening to Reverb. But now for the real point of my experiment, let me try to align two tuning forks. You see that? The ball shows the second fork in motion. Sound waves traveling from one to the other because they are tuned to the same frequency. This is resonance.
This is physics, and it is very different from reverb. Reverb is that lingering sound that continues after the source stops. Resonance is whether anything or anyone is actually moved by it. Two perfectly tuned and positioned tuning forks resonating. Now, here is the thing. The communications industry has become very good at the first strike. Look over here. How cool is our demo? We are first. We are fastest. We are at the frontier. But also, do not ask too many questions. Volume, velocity, spectacle. Anyone can strike a tune in the era of super intelligence, but what matters is whether or not it resonates, whether it amplifies human creativity and capability, whether it arrives in tune with the person it was meant for, whether or not it leaves a lasting impression long after the initial call.
Whether it enhances a customer's experience or maybe it is going to turn into an escalation, what determines whether or not it resonates? Physics. Not the promise of infrastructure, but infrastructure itself. For over 25 years, while others were making noise, we have been building the thing that makes resonance possible, tuning every layer, refining the signal, earning every single millisecond. That is not a slick pitch. Just like the laws of physics cannot be bent, the same is true for communications infrastructure. It cannot be rented, cannot be prompted, or, these days, even vibe-coded overnight. AI made promises to enterprises, virtual agents that imitated human voices, conversations that tried to know your full history, customer experiences that hoped to keep up. But now we have entered a new era, an era that reaches audiences and breaks barriers well beyond what any of us thought previously imaginable.
The real question is, which partner can actually keep pace and promises, amplifying human potential for the future of communications with super intelligence? Our mission has always been about developing and delivering the power to communicate from a rock-solid foundation. It has always been about resonance, being aligned perfectly, tuned precisely, and ready to amplify communications for all of you, for all of our customers. Today, that mission is delivered in four distinct layers. Number one, infrastructure, every connection owned. We have got a global network that is architected and tuned precisely for your natural, ultra-low-latency conversations. Two, orchestration. Every single choice, yours. No single lock-in with a vendor. No DIY patchwork of communications. Complete control over your communications in the cloud. Three, performance. Every single millisecond counts. Every single millisecond earned. We are processing super intelligence right next to the call, so your latency budget benefits. Four, trust.
Every single call proven. In an agentic world, you need to know who or what is on the line, and the carrier layer authenticates that truth. Four layers of resonance developed to deliver human-to-human, human-to-agent, agent-to-human, and yes, even agent-to-agent conversations. Resonance starts with infrastructure, moves faster with performance, stays integrated with orchestration, and feels authentic with trust. Somewhere right now, while I am standing on the stage and you are here, a woman is calling a business, and something that isn't a person is answering. Guess what? It picks up on the very first ring. It knows her name. It sounds warm, even empathetic. It may even talk in the same accent or age group as she is in. She has a technical question. It's a new medical device.
She needs to ask a question about it, and in about 90 seconds, she will hang up, trusting that company a little more or a lot less. Today is all about resonance. Anyone can strike the first note, but what matters is whether it arrives, whether something or someone on the other end is moved by it. A year ago, AI came up in a lot of our customer conversations. Today, our conversations are reframed in this era of super intelligence, not merely imitation of human tasks, but amplification of creativity of folks like you and me. It's the content for everything we do, everything we say, and everything we build at Bandwidth.
Every enterprise we talk to, and I am talking from startups to Global 2000 to hyperscalers, and actually the biggest names in software and on Earth and in space, they are reaching the same conclusion, and they are reaching it very fast. Super intelligence is only as good as the communications platform it runs on, the communications platform it resonates on. This conclusion shifts how enterprises are buying. It changes what they prioritize and what we prioritize, and it determines who they partner with. I would like to venture to guess it is the reason you may be in this very room today. This is what today is all about, the intelligent communications infrastructure that we develop and deliver in-house. It's the core of our vision, and together, when your stack, your agent, your customer experience is tuned to that frequency, everything resonates.
Salesforce's Native Contact Center brings voice, digital channels, CRM data, and voice agents together in one system. Autonomous agents and live reps work side by side, solving issues faster across every channel, every case connected in one view, with seamless agentic handoffs from SI to humans. Intelligent agents need intelligent infrastructure, which is global, reliable, and secure. Salesforce's Native Contact Center runs on Bandwidth voice infrastructure. When enterprises are ready to expand globally, we are the first certified carrier for global communications in Salesforce's Native Contact Center. Direct control. Order, configure, manage every number, every port straight in the Bandwidth app. One point of contact for global reach. 25-plus countries, one network, no SBC required. Better quality, better reliability, ISO certified, and per-call data available in real time. Award-winning support 24/7 with a 95% CSAT rating. You are never doing this alone.
Migrate on your timeline in parallel with your existing setup. Software-controlled ports, zero interruption. Enterprise CX isn't built on promises. It's built on intelligent networks. It's built on Bandwidth.
Wow. Amazing. Hello, everyone online and here in Raleigh. I'm Devesh Agarwal, Chief Operating Officer at Bandwidth, and I'm joined by Ashish Seth, Vice President, Product Management, at Salesforce Service Cloud. Ashish, welcome to Reverb and welcome to our campus.
Thank you. Super thrilled to be here. Beautiful campus and a beautiful setting. Thank you.
Welcome. Ashish, let's jump right into it. You've had a big year introducing Salesforce's native Contact Center. You drove the strategy and question, what excites you the most about this release?
Yeah. I mean, I will start by saying it has been a very busy stretch at Salesforce for me and my team. I will start with the headline. Salesforce launched its own native Contact Center in March of 2026, and in six short months thereafter, we took it global. That contact center solution is now available in North America, in Europe, and in another week or two, we will be in Australia as well, and by middle of next year, pretty much covering the entire globe. That has been a journey in itself. To your question, what drove the strategy? We were super convinced at Salesforce that for us to succeed with agentic service, voice agents are critical. Guess what? Where do all these conversations happen? They start in the contact center.
The forcing function became that for us to succeed with agentic voice, we needed to build our native contact center and a native voice capability and infrastructure along with that. That is the forcing function. To the strategy, we said, "Okay, how can we do this better than others?" That is where we said our North Star is going to be: let us bring the best agents on a globally available platform that brings together data, context, and workflows all on one unified platform. With that sort of being the North Star, that became the strategy, and then when you execute to it and you launch the product, to your question, what makes me super excited, that strategy translates into product market fit, and we are seeing some amazing traction with customers. That is really getting me and my team super excited. Thanks to the partnership as well.
Yes.
Right. We are excited for you. Hey, you built this globally really fast timeline. How have things gone since launch?
The launch has been amazing. In fact, it has exceeded all our expectations. Since launch, we have added 1,000-plus customers. These customers include anywhere from what I call smaller contact centers to anywhere large contact centers with thousands of agents. As an example, customers like Canada Goose, I do not know if you are familiar with them, but they sell those parkas, right, up in Canada, and I guess that is where their major market is. Or companies like FanDuel, the sports betting platform. They have started to move their workloads onto our contact center. So, it has been an amazing journey, and we are seeing some amazing successes with customers as well. Nothing pleases a product manager more than a product market fit and customers actually using the product. That has been very encouraging.
Ashish, awesome. The brands you just listed and the pipeline we have seen. Salesforce Contact Center takes CRM data along with personalized SI data to handle complex conversation. I think you have raised the bar of what a contact center can do. This must require a solid voice infrastructure. How did we deliver?
Oh, I will say this, absolutely. They say your building is as strong as the foundation it is built on, right? The same analogy applies here. For us to have a contact center, you absolutely have to have a voice infrastructure sitting behind it that is scalable, reliable, trustworthy, and above all, the quality of service is paramount, especially in this era where, the agentic era, if you would call it. Here you have conversations between humans and agents, and agents to agents, and multi-party conversations. You can only have those conversations if you build it on a platform that delivers a very high quality of service. In fact, I will take one example, again, part of my North Star sort of approach, MOS scores, right?
Mean opinion score. Most of you are familiar with that. We said that the 90th percentile score should be 4.5 and above for us as a platform. Let us architect our infrastructure that way. Thanks to our partnership, working with you guys, our architecture collectively is now delivering that. Thank you.
You know, 4.5 on a scale of five, pretty good. Hey, for companies with global footprint, how does Bandwidth BYOC make Salesforce Contact Center a plug-and-play rather than a rip-and-replace project?
Now, great question. It starts with when we looked at our customer base globally, we found that our customers, especially the larger ones, have had some longstanding partnerships or relationships with carriers, right? Be it contractual or be it certain things that they're using it for, an enterprise-wide deployment. We quickly realized that we needed to, in order for them to be successful on our platform, we need to give them a very easy way to leverage that same infrastructure. So, we introduced this concept of bring your own carrier in conjunction with Bandwidth.
Yeah.
The headline there is, we wanted to make it configurational, not migration-oriented. So, what we really did was, when customers bring their workloads into our platform, they simply point their current infrastructure to our network, right? Again, built out behind the scenes. If I translate it into simple English, plain speak, our platform now allows you to log into the system, pick up the carrier that you're working with, and a few clicks and configuration steps, you can literally point that capability to us, so it becomes configuration rather than migration.
Makes sense. So much sense. Ashish, last question. You mentioned one team working together on this launch. What should we co-create next?
Oh, there's so much we can, right? I mean, so much of innovation is coming along that we just I mean, I'll say we could spend the entire day discussing. But few areas really are kind of driving the creative juices within Salesforce, which we would love to partner with you. The first is the rise of personal AI agents. I mean, the impact to contact centers is going to be huge. Now you have to kind of not only think about agents to humans, which we started to get used to, right? But now you've got agent to agent, and beyond that, we are seeing a world wherein you have multi-parties. There might be multiple agents and multiple humans all in one conversation. I mean, that day is not far off, right? And we need to be ready for it. So, that's an area of innovation, right?
That is on the innovation pillar. Quite frankly, the second pillar I feel is with all this happening at the fundamental, at the most core level, trust, reputation, reliability. I think the need for that is even more acute today. When you juxtapose the two, I would feel that our partnership should evolve to a point wherein customers look to us as our trusted, I guess, trusted partners who can deliver all these capabilities, rich functionality, but above all, on a reliable, scalable, and a trustworthy platform with Salesforce's name on it, and of course, with a partnership backing that.
Ashish, couldn't agree more on trust and the agent, and we are going to cover more on those two topic. We will take a few minutes on that, but thrilled to continue our partnership. A lot of growth for both of us. Excited to see what Salesforce does next and takes it next. Ashish, thank you. Thank you, Ashish.
Thank you. Thank you.
Thank you. Thank you. Speaking of what is next, the infrastructure behind it all. Let us take a look.
Whatever you are building, it sits on a network. The question is, which one? Bandwidth is built differently, owned and operated, not rented and relabeled. Here is what that means. Fewer hops means lower latency and better budgets at scale. 5X toll-free carrier redundancy with Call Assure means no single point of failure. HD Voice, private media paths, and edge native media processing means your calls sound the way they should. Patented anomaly detection means proactive responses to potential issues. Five nines of global core network reliability means you sleep at night. Owned and operated means business, because whatever you are building, your super intelligence included, is only as good as the infrastructure you are building upon.
All right. It is funny they asked the software guy to come up and talk about infrastructure, right? Why not somebody from the network team? Well, if you think about it makes a lot of sense. Whatever you are building, it is only as good as the infrastructure you are building it on. In the era of agent-enabled software engineering, our roadmaps have accelerated. Our commits, our PRs have doubled and tripled. It has never been easier to build and ship software than today, but it is still only as good as the infrastructure you are building it on. Here is the thing. While others are exiting the voice business, we are going the other direction. An owned and operated tier one global communication network that is architected at its core for reliability, quality, and observability. That is Bandwidth. Let us dig into what owned and operated really means across three key areas. First, global reliability.
We have unified nearly 70 countries on a single software layer, the Universal Platform. This year, we added capabilities across Peru, South Africa, Mexico, Brazil, and South Korea. We brought voice and messaging on the same telephone numbers to the U.K., Australia, and Mexico. Multimodal customer experiences on a single number. Bandwidth now gives you full stack PSTN replacement services, which means all of this in 43 countries on a single platform, achieving five nines global core network uptime. We are not done. In 2027, we are adding more coverage options across Malaysia, Philippines, Chile, Panama, Indonesia, Taiwan, and Japan. We are expanding our footprint so you can focus on what you do best, building software and delivering world-class customer experiences. Okay, so today, every regulated country you need coverage in typically requires a six-figure investment in licensing and compliance infrastructure. That does not include time, paperwork, research. You get it.
To make this easier, we are launching a new beta carrier of record. It is designed to enable our global voice resellers to enter new markets without all of the infrastructure headaches. We operate as the carrier of record for your end customers, helping you and your customers achieve local compliance. We also have another exciting new beta to announce, our SI porting agent. Our new agent works inside the Bandwidth app. It helps you translate a port rejection, checks it against your customer service record, and gives you control to correct and reschedule your phone number port, all without having to talk to a specialist. Two betas with the same idea. The best software on top of the best infrastructure is how you solve global reliability. Okay, next, let us look at what owned and operated means for quality.
Every extra hop, every added media anchor, every codec that needs transcoding, it all adds up. We prioritize the fewest hops with direct peering to all major wireless carriers. In fact, more than 93% of our inbound and outbound voice traffic is one hop or fewer in North America alone. But fewer hops is only half the story. We are changing how we handle the media itself. Let's talk about media transcoding. With AMR-WB, our network can skip transcoding on calls from a wireless carrier or an on-net number. This cuts an average of 200 milliseconds from your calls that you can use for SI workloads. We have started with AT&T, available today in beta, and we will be ready with Verizon and T-Mobile by the end of the year. Second, let's talk about media anchoring in-country.
We have deployed a model that keeps media anchored in-country, starting with Brazil and Chile, with more countries coming in 2027. This benefits regional calls as well by reducing latency in the surrounding countries. To wrap up what we mean by owned and operated, let's look at observability. We have taken complex network diagnostics and turned them into tools. PCAPs and call ladders, two of our most requested network diagnostics, are now available in the Bandwidth app and API. When you combine this with MOS, jitter, and latency insights already available in the app, you have full control and clear visibility into your calls. Okay, I started by talking about roadmaps accelerating, commits, and PRs doubling and tripling. Technology is moving quickly. Agents are building our software now. You and your agent need something to build on without sacrificing velocity.
That is why we launched Bandwidth Build, a zero-touch way for you and your agents to sign up for Bandwidth and use our CLI, our command line interface, to connect and make a call out of the box in five minutes. We have integration guides with leading AI platforms, a new developer migration toolkit to help you implement to our APIs, and even more features like a 10DLC messaging trial live now in Build. As you build, you can rest assured knowing that Bandwidth recently earned ISO/IEC 42001 certification. It is the global standard for responsible AI management, which means every capability we ship in the era of superintelligence is governed, audited, and accountable. It pains me to say this, but none of this, and this is a lot, would be possible with just software alone. It is only as good as the infrastructure that you are building on.
All right. Now let's get Judy, our Vice President of Account Management, out here, help me introduce a few of our customers and partners that are building amazing things. Come on out, Judy. Welcome.
All right, Jeff. All this talk about Bandwidth Build and voice AI running on the best communications infrastructure, and you know what? It's time. We need to show off the really cool things that our customers and partners are doing.
We do, Judy. You're right. I'm glad you're here. It's an amazing time to be a builder, and I can't wait to hear from our customers.
Well, today, we've got three innovators here with us to share their experience and expertise, so let's get right to it.
Let's get to it. Okay. First up, we've got Joe from Flip. Flip is a conversational AI platform that serves large brands. So think Fanatics, the NFL, GNC, Tory Burch, Under Armour.
Nice list.
Yeah. We are excited to have Joe here to share details of their switch to Bandwidth. Flip's approach is one that understands what the customer wants, so it applies the brand's own rules, takes action by canceling the order, modifying the shipment, scheduling the appointment. They are going deep with a verticalized strategy across retail, healthcare, and transportation, and we will hear why that is resonating with their customers.
Next up in our lineup, we have Eric from GeniusAI. They are a SaaS platform recognized by Forbes on the Cloud 100 list, and they serve over 100,000 salons and beauty businesses, facilitating appointment reminders, marketing campaigns, customer loyalty programs, and point-of-sale transactions. They have been building on Bandwidth for years, and now they have built something new, GeniusAI Reception. This thing answers calls and texts 24/7, including roughly 50% of the inquiries that arrive after hours. It functions like a real receptionist. It knows your calendar, your client history, and can book clients and answer questions in real time, and it all runs on Bandwidth. We will hear more from Eric about the value that this delivers for his customers in just a few minutes.
I can't wait. After that, we will round out the panel today with Spencer from Bland AI. Their name is ironic because their product is anything but bland.
Not Bland AI at all.
In fact, they've carved a lane in regulated industries like government and healthcare. They are the first voice AI company to become FedRAMP 20x Class A certified.
Awesome.
Meaning federal agencies can run voice AI built for high-secure, high-stakes work.
Their agents are amazing. I mean, real amazing. They actually sound like real people.
They do. I was just going to say they sound like real people. They go live almost instantly, and they get better on their own. Today, we are excited to announce that Bandwidth just released an integration with Bland, making it easier than ever to use Bland on Bandwidth.
This is really awesome news, Jeff. Great for our customers too. I think it is time to hear from them. Jeff, thank you so much for the intro. Thank you. See you later. We will talk to you in just a bit.
Thank you.
All right. Everyone, welcome to the stage, Spencer, Eric, and Joe. I am so very excited to be here with you today. We are going to just jump right into it, if you do not mind. This question is actually for all three of you. All right? But Eric, we are going to start with you, if you do not mind. What is one challenge that you faced and solved while building intelligence into live call flows today?
Well, first, thank you so much for having me, Judy.
Yes.
It is great to be here. GeniusAI Reception product is often that first interaction or maybe a critical continued experience that one of our users' clients has with that business. Tone and personality often matter as much as does it work. You heard David talking about that earlier.
Yeah.
To bring that experience to life quickly, we opted to buy rather than build. We layered voice capabilities directly onto that existing Bandwidth foundation, then we tested rigorously to make sure that the product is looking good with that small customer cohort before we launch. As adoption grows, we are going to offer even more flexibility to our users.
Does it work is not good enough. It actually has to be really good these days. All right, Joe, what about you and Flip?
Yeah. One of the biggest challenges is bridging the gap between a flashy demo and actual production, which usually takes months of painful implementation. We solved this by sticking to three different industries, retail, transportation, and healthcare, each with over 200 out-of-the-box integrations and intents. Our demos run on production systems, meaning that the customer is able to experience the real Flip. Partnering with Bandwidth lets us spin up numbers instantly without having to worry about capacity. Because demos run on actual production rails, it allows our customers to go live with zero engineering heavy lifting, and customers can just forward their calls to us and be live within an hour.
We all know how important it is to have that peace of mind that things can get going instantly and not having to worry about things like capacity. Spencer, what about you at Bland AI?
One challenge we've worked through is getting accurate answers when a customer's data is spread across different systems. We connect the agent to those systems, set guardrails around what it can say and what it can do, and use call reviews and testing to keep improving it. That all has to work on the phone infrastructure the customer already trusts, which is where our partnership with Bandwidth comes in.
What I'm hearing is, to get it right, it takes testing, scale, and the right foundation. And that foundation actually matters more than some people expect. So really, let's dig into why. Joe, across your customers' calls, what does call quality do to the agent experience, and how do you find the calls if something's not right?
Yeah. So there's two sides to call quality, what the customer hears and what Flip hears. The moment audio degrades, everything breaks down. The system misinterprets words, misses when a person stops talking, or can't handle natural interruptions. When a caller experiences awkward pauses or gets talked over, the customer loses trust. To control that, we host our own SBCs, media servers, and monitoring systems. We have full visibility of the call from start to finish. We track call quality with automated alerting for MOS scores, packet loss, premature barging, and latency. This comes with Bandwidth's network to help us keep the conversation seamless.
You know, like Joe said, latency matters because I think there's still some of that inherent distrust from our customers and their clients when a robot picks up the phone.
Right.
We need to build that resonance we were talking about earlier. Our goal isn't to mislead anyone, but we want to make it feel like a normal conversation. Latency and human-like speech matter, even including things like filler words like and.
Yeah.
It actually makes it seem more human. The Reception product from GeniusAI serves that salon owner on two sides. First, it handles the call, makes the booking without any need to be involved. Then second, it makes the in-person experience more, I guess, intentional, where that salon provider can focus fully on the person who's in front of them. It's a better customer experience, I think, for everyone, because it preserves that critical revenue in an industry where our research shows more than 60% of calls are missed, and those callers aren't calling back. All that puts a lot of pressure on that connectivity. Something like a latency spike can literally translate to lost clients, which means lost revenue. When the spikes do happen, we need to be able to track those issues. We need to investigate at the network level, the carrier level.
That's table stakes for us.
Listen, when the call feels off, trust breaks down pretty fast.
True.
Which really takes us to the next question. Spencer, can we talk about trust for a quick second?
Yes.
As virtual agents place and answer more calls, how do you help your customers make sure people trust them?
A really important question. We think about trust from both the caller side and the customer, the business side. For callers, that means a dependable experience, clear disclosure that they are speaking with AI, and respecting their consent and opt-out choices if they make them. For the businesses deploying it means being able to see what happened on a call and being able to audit it. We bring that visibility across Bland's models and owned AI infrastructure alongside Bandwidth's carrier network. We use call recordings and monitoring to support compliance. Security assurance like FedRAMP 20x Class A also matters a lot for government and healthcare customers.
Right. What I know is security, compliance, reliability, and trust is earned at every layer. Well, as we wrap up, I only have one more question for the group, and I am really excited to hear your answers because I am sure your customers are always asking for new things. Eric, tell me what your customers are asking for next.
For our customers, we are really thinking about all of these service providers being able to have a unified phone number experience. They want that single identity, and that comes for their communications with their clients. That is voice, that is text, that is more. I know it is obvious, but that is where we kind of bring this to life, and it speaks to that deepening partnership that we have been fortunate to have with Bandwidth and GeniusAI.
I am sure that is actually something that everyone in the room can share the same sentiment. Joe?
Yeah, like Eric said, more channels are always on our customers' minds, but additionally, what they want is deeper integrations into their own systems. The rule of thumb for us is simple. If a customer agent can do something in an internal tool, then the AI should be able to do it too. A lot of our customers have their own order management systems, so we are building custom connections to automate those workflows completely. That is a big part of what our customers are asking for next, connecting the AI deeper into their systems that they already use, so they can do more for their team, more for their customers.
Customers are asking us to go beyond just answering questions to completing tasks. That is booking appointments, resolving complex issues, and taking actions across their systems without sacrificing customer satisfaction or making the caller repeat themselves. Next is our new speech-to-speech model for more natural conversations, paired with enterprise control that customers need to confidently put their brand on the line.
We know customers hate repeating themselves. All right, guys, I sincerely appreciate your insights today. Thank you so much for joining me. Special thanks to Joe for traveling all the way across the pond to be here with us today. Let us give them a round of applause. I am genuinely very, very excited to see what those customers are asking for next so that we can continue our partnership and work with them. All right. As these guys have shared, it really is the infrastructure behind the intelligence that matters. This defines the customer experience, but now I actually want to bring out my partner, Caitlin Long, to the stage to share more about how we help businesses orchestrate it all. Caitlin, come on up. Hi, girl. Hey, girl.
Thanks, Judy. As you just saw, the industry has moved beyond the hype cycle and into real-world implementations, experimentation to production. This is a familiar story. All right, y'all, a little blood flow exercise, show of hands, who remembers any of these hype cycles? Digital transformation. Okay, there we go. Big data, ring a bell? There we go. Hands, hands. Mobility. Yes. Okay, thank you. We have been here before. Now, in the era of SI-powered communications, it is the same problem. Make it work with what you have already got and keep a close eye on efficiency, reliability, and ROI. That sounds so easy, right? Y'all know it is not. Recall David's story earlier about a woman calling support. Let us use her question to illustrate just how complex orchestrating this call is. Step one, her call hits the carrier. Two, the IVR.
Three, the CRM pulls up her account, and now, with SI on the call, steps four through 40 enter the chat. Recording starts listening, then transcription, QA, sentiment analysis, NPS. Do not forget compliance. You get it. You have got new agents in new apps, every one of them adding triple digits of latency, performing their own little side quests, and that is just one call in one country, likely with one provider. Add regions, regulations, carrier, languages. Every layer multiplies the one before it. At the end of the day, it is just one woman who needs a timely answer about a real problem. Now, do not get me wrong.
Every point solution has its purpose, and plenty of them are our partners, and we are so proud of that. But how do you make sure everything is tuned and orchestrated together? We call it Maestro. Maestro is our cloud-based orchestration platform.
It lets you connect, create, and compose communication workflows without vendor lock-in, simplify integrations with leading UCaaS, CCaaS, and authentication providers, add advanced capabilities like caller authentication, and bring your own virtual agents. Of course, orchestrate everything with a visual call flow builder. This year, we have added even more new capabilities and integrations. Let us recap. We launched audio conferencing for Microsoft Operator Connect, now available in 25+ countries. This lets Microsoft Teams customers choose Bandwidth as the certified carrier behind their dial-in and dial-out conferencing numbers, with more coverage and control and consistent policy management. Next, something worth repeating from the conversation earlier with Devesh and Ashish Seth. Bandwidth is now the first certified global carrier for Salesforce Contact Center, a CRM-native contact center that plugs right into Maestro. It is generally available in 30 countries across Europe, New Zealand, and Australia.
This is a big testament to the partnership with Salesforce and the future of SI communications. Maestro resonates with both analysts and customers. In fact, last week, we learned that CX Today named Bandwidth the winner of Best CPaaS Platform at the 2026 CX Awards. Yes, yes. So far, Maestro has given you the freedom to integrate your preferred tools into your stack with our owned and operated infrastructure across the globe. This year, we are taking Maestro to another level. We have invested in native capabilities that work alongside the best-in-class tools you are already using. We are closing two critical compliance gaps with our compliance use case products. First, 911 location for nomadic workers, and second, call compliance in the contact center.
First up, hot off the press, we are launching Emergency Location Manager, a tool that allows your employees to update their address as they work remotely, move between offices, and ensure that first responders can actually locate them at the time of that 911 call. This helps those enterprises stay compliant with U.S. regulations like RAY BAUM'S Act. The best part, it is native on Bandwidth's platform and does not require any additional third-party contracts. In fact, Bandwidth is the only tier one carrier with this capability. We are extending it to both our reseller and enterprise emergency services customers. The second compliance use case is a game-changer. Think about how you interact with virtual agents today. Transactional stuff, right? All day long. Have you ever actually tried to close an account with a virtual agent?
When the stakes are high, the call still escalates to a human. It is the same when the sentiment turns sour. The virtual agent hits a wall, and a person steps in. It is precisely at this handoff from super intelligence to human where the customer experience is won. Bandwidth Call Intelligence, it is our latest innovation with recording, transcription, redaction, and compliance monitoring built right into the orchestration layer. Let us check it out.
Every enterprise has conversations.
Could you read your credit card number for me?
Yes. That would be five. What is the reason for my conversation? It has to protect. Are you solving this the hard way? A redaction tool bolted into your contact center, another one for your IVR, and now an AI voice agent sitting outside both. Status quo, record the call, send it somewhere else, and hope nothing slips through. Every leg of the call, an increased risk. Now, with Bandwidth Call Intelligence, the call is secured from pickup to hangup. It records, it transcribes, and when something sensitive comes up
Could you read your credit card number for me?
Yes. That would be five. It's like it was never there. At the same time, it's vetting the call against your compliance rules and flagging anything that needs attention. Less compliance risk, less exposure in storage, less time on manual redaction. Bandwidth Call Intelligence, record, transcribe, redact at the network layer.
Y'all, I'm so excited to dive into this more, and even more excited to have Noemí, the Product Director behind Call Intelligence, on the stage with me. A little extra welcome, Noemí.
Thank you, everyone. Thank you, Caitlin. I'm very excited to be here.
Now, Noemí, you and the team have spent months building this, but what was the driving force behind this product?
We built it directly with our customers, side by side with their compliance and their operations teams.
I think you've brought a couple of examples that show off exactly how Call Intelligence can help those same compliance teams stay ahead by giving them the ability to configure their own redaction and alerting policies across the call flow.
Yeah. That's right. I'm going to show you all how it works.
Can't wait. Let's see it.
You can see here, this is a call from Fermata Financial, one of our fictional banks. Before we had Call Intelligence, the call arrives to one of Bandwidth numbers, and then it connects to an SI agent. Then the call can be handed over to a human agent whenever that is needed. That handoff in between the SI agent and the human agent is where compliance visibility usually breaks. Adding Call Intelligence just takes a few steps. You right-click to add a node, pick up Call Intelligence, drag it, connect it, and you're done.
A quick note here, because Bandwidth carries the call, it's captured on the network before it reaches any software, so that means no extra vendor and no integration project to plan around.
Yeah. None at all.
Love that. Maestro takes care of it, no matter what platform or contact center is connected.
Exactly. Our customer, Fermata, has everything recorded, transcribed, and redacted. Everything will be consolidated in one single record, stored in one single place. Now, let's move into my favorite part, the rules.
Y'all know Noemí loves a rule.
Names, date of birth, Social Security numbers, they all are going to be covered from day one under PII redaction.
Another point of complication, compliance teams, they all have their own policies or rules, right? Those can change.
Yeah. Every customer will be able to add their own rules based on the company's standard operating procedures. I'm just going to jump quickly to another call flow where we've already configured Fermata's policies, and you can see them all under this monitoring agent here on the right.
Right. In a sense, the team's knowledge base becomes a part of the call flow.
Exactly. If an agent breaks a rule, you will get an alert right away. You won't have to wait days later to get a report.
Do these alerts actually link back to the call flow so that the agent actually retains the context?
Yeah. Our customers will get a clear note of where and why the call has been flagged.
Awesome. Can you show us what that looks like in practice?
Yeah, of course. I'm just going to move into our insights page where we already have a few test calls.
Perfect.
I'm just going to navigate to recording. You can see here on the screen example where a customer has called Fermata Financial about a personal loan. The SI agent is answering first, and then the call gets handed over to a human agent. You can see the recording and the conversation word for word in the screen, and you can also see that the flagged moments are also marked in here.
Okay, perfect. Let's start with this first red flag up here in the right-hand corner. What happened here?
Basically, at this moment, the system was down, and the agent wrote down the customer's Social Security number.
Okay. Yep. See that right there in the transcript. I do.
Yeah. That action is breaking Fermata's policies based on their company's standard operating procedure. Their compliance team will want to know about that right away.
Yeah, based on this little frowny face, I'm going to assume that the customer wasn't super happy about that. I know that I'd be a little upset if someone was writing down my Social Security number by hand.
So basically, you can see that the customer hesitates, then the sentiment score is going down. Then the agent says, "Well, it's totally normal." That's the second yellow flag that you can see on the screen, is misleading information about how data is being handled.
This is really where you can take those insights and turn them into action with that human touch.
Exactly. It will tell your compliance team you need a conversation, not just a note in a file, whether that's coaching the agent or just calling the customer back.
Noemí, this is amazing. Thank you so much.
Thank you, everyone.
I am so proud of what Noemí and the team have built around Bandwidth Call Intelligence, taking those compliance rules and alerts and turning them into actionable insights, on the call the entire time, covering the entire orchestration journey, integrated with Maestro. That is what orchestration looks like at the infrastructure layer. Now, I want to invite Jason Sommerset onto the stage to preview an exciting performance upgrade coming soon. Jason, come on out.
Thanks, Caitlin. Quick calibration check. We have covered a lot of ground today. Layer one, owned infrastructure, quality and reliability across the globe. Layer two, orchestration, Maestro, end-to-end control, and the freedom to choose the platforms in your contact center stack. Those two layers are powerful on their own, but to make it all resonate, it comes down to another layer, performance. To us, performance means one thing, running the voice models where the phone call happens. I am excited to announce we have deployed GPU infrastructure in our U.S. data centers, and we are taking it global in 2027. Running our own GPU infrastructure gives us real advantages. We can optimize performance for the hardware we own. We can remove network hops to deliver real-world ultra-low latency speeds, and we can avoid the need to ship data to multiple vendors across the internet just for one call.
Zero cloud hops, zero cloud round trips. This is as close as you can get to bringing super intelligence to an enterprise phone call. Here's why this isn't just about infrastructure. It's about performance. This is really the awesome part. We've built our own streaming native speech-to-text model designed for enterprise phone calls. Let me say that again. We built it, and we built it from scratch. To learn more about it, let's hear from Mike Nelson, our principal engineer who led the effort.
Thanks, Jason. We get some wild phone calls, noisy rooms, busted cell phones, heavy accents, signals that fall apart halfway through a sentence, and most models were built for clean audio, so they stumble on calls like these, and they stumble unevenly, fast on one call and slow on the next. On a carrier network, the quality has to be there every time, and the latency has to hold steady. So in the lab, we set out to build our own model that could hold up on real phone calls. Designed and architected in-house and trained from scratch, we built the infrastructure to run it using state-of-the-art GPUs deployed on our network. The entire inference stack, kernels and all, in our own inference engine, built from the ground up for real-world communications. It runs on our own GPUs inside our network in the call path the audio already travels.
No extra hop and no one else's cloud. Let's dig in. There are a lot of ways to judge a transcript. Two of the most common, word error rate, any mistake from a typo or misspelling to a complete missed word, and the other is semantic word error rate, which asks whether the mistake changed what the person meant, a real error that would impact the understanding of what was spoken. We measure against three benchmarks. The first is our own, built from some of the most challenging phone audio we have. The second is OpenASR, where almost every model is non-streaming and can take its time to process the audio. The third is Pipecat AI's semantic word error rate benchmark, where real meaning-changing errors are counted on real conversational audio. On all three, we're holding our own on the frontier.
Today, we'll share our performance metrics to let you arrive at your own conclusion. All of the dots on our charts represent publicly available metrics from open source and commercial models, including ours. To us, streaming native means stable transcripts fast, while the words are being spoken. In our speech-to-text model, every word we emit is final. In our testing, words arrive in under 200 milliseconds on average. On Pipecat AI's semantic error rate benchmark, nothing is faster than us at the median or P95, while still being in the top four models for accuracy. Put accuracy and speed on the same chart, and Bandwidth's in-house model sits on the Pareto frontier. Nothing on that board is both faster and more accurate.
You can act on a word while the sentence is still happening, redact a card number as it is being read, or let an agent decide on a phrase instead of waiting on a paragraph. Those are not just clean room numbers. In our test, thousands of sessions on a single model instance with negligible impact to latency, even at P95. Hyper-optimized for Bandwidth's own edge GPU infrastructure means we are designed to deliver industry-leading latency at scale. Speech-to-text is only one of the many models under development here in the lab. It is one of a suite of models built and trained in-house on the same stack at the same speed. Real-time redaction, monitoring, and flexible detection tools that will help customers evaluate calls against their own policies, surfacing critical insights for whoever is on the call, human or SI, and more on the way.
These are the Bandwidth models powering our call intelligence solutions. Back to you, Jason.
Thanks, Mike. That is intelligence at its finest, architected and deployed on Bandwidth's own infrastructure to deliver ultra-low latency performance. We could only scratch the surface on the metrics and the benchmark data during today's keynote. I want to invite all of you to a live virtual deep dive on Monday, November 16th at 2:00 P.M. Eastern. I hope you can join us. Our team will share more about the model's performance and how it stacks up against alternatives. Even while launching our own native streaming speech-to-text models, our vision is still an open platform, just like we have created with Maestro. Owning the GPUs is what makes that possible. We can host the best open source and commercial models right next to the call alongside our own. Choice, flexibility, and control, now at the infrastructure layer for the highest performance SI in communications.
We are proud of what the Bandwidth Labs team has been able to do, and two of our team members are with me today to demo our work. First, let us help me welcome Dane Linsky to the stage. Backstage, we have Jake Gregory. You are not going to see him today, but you will be able to hear him because he is going to be on the other end of the phone when I place a live call. Let us revisit what Noemi showed you earlier, where Fermata Financial's team learns about a call, learns from that call after it is over. This time, we are going to call in while the model listens live to see what it can do before the customer hangs up. It will be the same bank.
Jake is going to act as the agent for Fermata, and I am going to be the customer, and I am going to ask a question.
Dane will walk us through what is happening live on the screen. Three things to watch for: how fast the words appear, whether anything sensitive lands as text, and what the system can do while the call is still going on. To prove this, I am going to need some help from the audience here. I need to come up with a new email address. Let us do this Mad Libs style. From over here, I am going to need an adjective. Maybe shout out a couple adjectives. What is an adjective we got? Handsome. Handsome. I like that one. Handsome. This side, an animal. We got handsome, we need an animal. Kangaroo. Kangaroo. Handsome kangaroo, and let us throw a couple digits on that. 22. 22? 22. Handsome, I got to get this in my head. Handsomekangaroo22@hotmail.com.
That will be my email address for this demonstration. Dane, do we have any live calls running right now?
Nothing yet, Jason. Once you start the call, it will show up here.
I will get this call up and running here.
Thank you for calling Fermata Financial. This is Jake. How can I help you?
Hey, Jake. I'm hoping you could help me here. This morning, I sent a wire transfer to my contractor, and it doesn't look like it's gone out.
Yeah, I understand, and I'm happy to help you look into this. To get started, can I get your name and the email address on the account?
Sure. My name is Jason Sommerset, and the email address is handsomekangaroo22@hotmail.com.
All right. Give me one second, Jason. I'm looking up your account details now.
While Jake looks up his account, let's review. Mike showed you the benchmarks, but this is what 200-millisecond streaming latency actually looks like in the real world. Now, let's focus on the streaming redaction. The email address, we certainly could not have predicted that, but here it is with a red tag. Noemí showed you what gets redacted. This is when, before it lands anywhere.
Jake, this is my second call about this wire this morning. My contractor's waiting on $5,000, and if this doesn't go out, I'm seriously thinking about closing my account.
Oof, that doesn't sound good. Let's take a look at the live sentiment analysis. Live sentiment updates with Jason's last spoken turn. Caller sentiment keeps track of the call as a whole. As we listen, pay attention to the updates happening.
Jake, I'm seriously getting testy here. This better go quickly.
I understand this is frustrating, and I'm working as fast as I can. I just need a little more time to figure out what's going on here.
This call is not going too well for Jake. In the insights section, we can see that the escalation risk is increasing, showcasing Jason's frustration. With the customer on the line, we have actionable insights ready in milliseconds.
Jake, actually, I'm sorry for coming in so hot. The reality is, as long as this goes out today, this is all I really need.
It's all good, Jason. Don't worry about it. Okay, so I see the wire for $5,000 from this morning. It looks like it got flagged for potential fraud, and it's pending a review. I'm releasing it now, and the transaction will be completed immediately.
Okay, this is perfect, Jake. I'm really glad this is getting fixed. Appreciate your help.
During the call, a few violations popped up in the compliance window. While Jake is an awesome software developer, working for Fermata is just not his expertise. Fermata actually requires a home address for account verification, something that Jake just never asked for.
Hey, Dane, the raw speed of this is truly impressive.
Thank you, Jason. We are super excited by the performance, and we can't wait to get this into our customers' hands. You just watched a live call on the Bandwidth network get transcribed and redacted as it was spoken, a median of 200 milliseconds from Jason's mouth to written word with state-of-the-art word and semantic error rates. Here's why that's possible. We own the network the call travels on, we own the GPUs sitting right next to it, and we built the model running on them.
Noemí's call intelligence demo showed you how compliance alerts can arrive in time to do something about it. With a model this close to the call, in time can even mean while the customer's still on the call. Fermata's playbooks were built around what happens after the call. What would you change if you could act during the call? All of this is possible because we own everything from the voice packets to the GPU to the model, and we optimize the whole stack together, and that's what truly sets us apart. Well, that's the preview, and we're excited about what's to come. Dane, thank you for joining me on stage. Everyone, please give Jake and Dane a hand. Everything we just showed you is built for the next frontier of communications, the era of super intelligence, where human cognition isn't just imitated, it's amplified.
Bandwidth Labs is laser-focused on what comes next and how we can build high-performance, trusted experiences. With GPUs in our voice network running our own speech-to-text model, call intelligence is even more powerful. We took the hops out, so recording, transcription, and redaction can stay in one place. That's not an optimization. It's a new architecture for voice networks built for the SI era. You could actually try our model for yourself today at labs.bandwidth.com. Building trust isn't just about the models, though, and that's why I would like to invite Seth Ray onto the stage to share more about what's next for trusted communications at Bandwidth. Seth?
Thanks, Jason. Watching that demo backstage got me stoked for the future of trusted customer experiences. That future has one underlying problem: Can you trust it? The trust gap doesn't just exist, it gets amplified in the era of super intelligence. Let's hear it from one of our customers, Quiq, who is pioneering conversational trust at the leading edge of customer engagement.
In your personal life, if you want to communicate, you pick up your phone and you probably send a text message. We're not trying to deliver 100% AI-driven conversations. We're trying to deliver conversations that actually resolve people's issues. Customers come to Quiq to help improve their customer experience, and they do that with AI as well as humans. No matter where the customer is, whether they're on phone, email, in a messaging channel on their mobile device, the Quiq platform can facilitate that conversation. We are focused on building the next generation customer experience based on messaging channels. Let's talk about IHG as an example. The two benefits of RCS are the branding plus the rich messaging. If you search on, say, IHG customer support, you'll see a messaging button.
They can also directly open a conversation in RCS to have an interactive two-way with either an AI agent or a human agent. They have seen traffic that would have otherwise come in on other channels starting to originate from the messaging conversations. If you're actually trying to book a hotel, when you get a message that has the logo and the color at the top, as opposed to just a generic phone number where it's like, "Who's this?" The level of trust associated with actually seeing a branded interaction, it's significantly better than just a random phone number. RCS, SMS, a lot of times those technologies get viewed as commodities. In our case, each one of our customers is a unique implementation. The use cases that we solve for customers are different every time.
Bandwidth has been a good partner to be able to work alongside with and navigate the compliance issues, not just the U.S., but across the world. As a communication platform, the fundamental thing that we need to get right is flawless communication. Working with Bandwidth has given us that high reliability and seamless delivery of service to trust to work with to deploy new technology. It has to be built on a solid foundation, and if the foundation isn't 100% reliable, then it doesn't matter how good the application is.
What Michael just shared shows how trust can be built from verified channels like RCS. Reinforcing trust in communications is the North Star for real customer engagement, but it's also more than just customer engagement. Super intelligence has given bad actors new tools, new strategies, and more ways to erode trust. Communications that used to be inherently trusted are now met with skepticism by the general public. "Don't pick up that call. Don't respond to that message. And if they're not in my contacts, I'm not engaging." This is an industry-wide problem, and we addressed this head-on last year with Number Reputation Management. It's our way to mitigate your legitimate calls from getting this label to spam. It's available for all of our customers, whether you're using Bandwidth phone numbers or those from another provider.
Today, Number Reputation Management works on numbers in the U.S., with support for the U.K. coming soon. Nobody answers when the screen says spam likely, so that's where we started. But a clean number is only the beginning. Getting rid of spam mislabeling doesn't mean your calls will be answered. Eight out of 10 people still don't answer a call from a number they don't recognize. That statistic hasn't changed. What has changed is what you can do about it. By partnering with industry leaders, Bandwidth has come up with a three-layer solution for effective, authenticated, branded calling. Prove it to you, protect your name, present your brand. First up, we have layer one, Identity Authentication. When your enterprise places an outbound call through Bandwidth, we send an API signal to our authentication partners to generate a short-lived verification token.
That token travels with the call to the terminating carrier, AT&T, Verizon, or T-Mobile, where it's validated in real time. If the token checks out, your brand display is delivered. If it doesn't check out, your brand won't appear. This proves your brand and reputation at the network level before the consumer's phone rings. Next, we have layer two, identity defense. The first layer we just talked about prevents your brand display from being associated with spoofed calls. But what if you could just block spoofed calls from the very beginning? Identity defense goes a step further to prevent fraudsters from placing a call using your number, your trusted enterprise's number, to impersonate your brand to consumers. It does this by blocking those fraudulent calls at the network level before they reach a consumer's device. This protects your brand. Finally, we have layer three, Identity Presentation.
This puts your logo, your brand name, and your reason for calling on the consumer's phone before they answer. The call is verified, the brand is visible, and this closes the gap between unknown caller and answered call. This presents your brand. This is what trust looks like at Bandwidth. Prove it, protect it, present it. A layered approach to a complex problem. Everything we've talked about today only matters in action, and the stakes are especially high in the healthcare space. Our next guest, Ryan Norton, deals with this every day. He leads the care center at LifeMD, a telehealth company serving more than 600,000 patients across 50 states. When LifeMD reaches out to a patient, that connection matters. But first, patients need to know who's calling. They need to recognize the call is legit, not spam, and feel confident picking up the phone.
To solve this, Ryan's team modernized. They moved to Zoom Contact Center with Maestro and added Number Reputation Management to help protect their outbound calls from spam mislabeling, improving customer connect rates. Let's welcome Ryan Norton, Senior Director of the care center at LifeMD. Welcome, Ryan.
Hi. Good to see you, Seth. Thanks for having me.
Ryan, thank you for being here. Welcome to Raleigh.
Excited to be here.
Ryan, at LifeMD, you deliberately call your team the care center, not a call center. Why is that distinction so crucial, and how does it set the tone for patient trust from the very first interaction?
We use the term care center really to keep our priority top of mind. We are there to provide elite care, right? We want to provide that consistently. Oftentimes, our frontline team members are going to be the ones who set the tone for that entire patient experience. So you will see we use another word for our greeting. We will say, "Welcome to LifeMD," rather than, "Thank you for calling." We do not want to thank a patient for having an issue that required them to call us, but we do want to welcome them, right? We want to welcome them and any concerns that we can address to help elevate their experience with LifeMD.
We know that when a patient reaches out to us, they are calling in having already overcome several hurdles, right? A lot of times there are cost concerns, potentially a bad experience with another provider. Whatever it took, we know that it requires them to provide effort, and also to overcome some of the hurdles that come with everyday life. The best way to keep a patient really is to care for them, and that is why we are the care center.
That is great. I love how that small distinction just creates that trusted experience from the jump.
Yeah.
Can you share more about that real business problem, though, that exists when a call gets mislabeled as spam or sometimes even blocked?
Well, I would be one of the first ones to tell you that if I see a number on my phone and I do not know it, I am probably not picking up, but certainly will not pick up, I can guarantee it, if I see spam or fraud likely. Our patients, we know they want these calls, right? They are expecting us to call, right? A lot of times these are provider follow-ups, maybe specifically requested by the patient or potentially an appointment. Anytime that they miss that call, ultimately, that is something that will delay their care, right? Potentially break their care plan, actually. We have patients that are calling in, and those same-day reschedules are really difficult. We are a lot about convenience, and oftentimes these patients are trying to fit in a consultation during a lunch break or during a 15-minute break outside of work.
Again, anytime that they miss that initial consult, there is a disruption in their life and in their care, and we are going to be the one that is going to accept responsibility for that, right? That patient will assign that to our team, and really, that is going to damage our relationship with that patient and really our credibility overall. We also know that our data bears out, any patient who does not have their consult within that initial 30-day window, they are about two times as likely to cancel subscription services with our team. Those patients, we want to have them seen and prescribed so they can ultimately start that health journey and get where they are looking to go.
That is great. Yeah, I cannot imagine a less trusted experience than seeing that spam likely.
Right
When you are a patient in that scenario, how did you guys solve that problem and that sit rate issue in trying to create a more trusted experience?
Solve is a word maybe I won't use, because we don't want to approach this in a way that we think we've arrived. We're always looking for an edge. We want to get 1% better, always in pursuit of that being better. I'll tell you that a big change came from when we began looking at our number reputation in a way that was not so reactive. For a while, addressing it after a number's been flagged is a really big issue, and it was really slow. Since we've joined the NRM platform, we've really transitioned our approach around it and really have all our numbers in there, and it's really become a foundational kind of component for our outbound dialing. A big lift has come from the automated remediation.
There's no longer a bottleneck, such as myself, potentially waiting to manually queue up this remediation process, and that has major ripple effects. We have our marketing team where they will send out the number that we'll be calling from to a patient of ours. That patient, we hope, will save that phone number in their phone, that contact number in their phone, and anytime that we have a spam designation, and we have to switch that phone number out, we're talking about having to ask patients to update contacts, potentially updating marketing flows. Just a whole number of things that come about that really this automated remediation process and NRM overall have really helped solve for us.
That's great. It's good to hear. You said solve there at the end.
Oh, yeah.
All right, let's shift gears a little bit. Thanks for the NRM piece. Today, everyone is exploring what automation looks like. With over 70% of your interactions happening digitally, how are you thinking about trust as you scale virtual agents alongside human agents?
Well, AI right now is a catchy term, right? Some folks do one of these, and others may roll their eyes. We know that AI is going to largely shape our future. We're really hesitant about just throwing out the old just to be in with the new, right? We've spent a lot of time, a lot of effort really refining our approach with our human agents, and we really view our virtual agents as an extension of that care team that we have. That means we're going to leverage the same verification techniques, right, the same touch points. We want to use AI as a way to enhance the accessibility and the convenience for our patients without sacrificing the confidence that they have in the outcome and just the trust in our company overall.
Yeah. That's great. Okay, last question for you, Ryan. It's the big one. For contact center and care leaders, in your case, preparing for their next phase of growth, what is your golden rule for preserving trust at scale?
Golden rule. I'd say you've got to be great at the fundamentals, right? You've got to do the basics exceptionally well. Tremendous execution on the blocking and tackling, because anytime that you layer something on top, such as AI, on top of a foundation that may have cracks, you're really going to see those start to present themselves in a much more meaningful way when you layer AI on top of it. I'll say, in short, just build on top of a foundation that you trust, not one that you hope you can trust.
Yeah. Love it. Ryan, absolute pleasure talking to you today, man. Thank you for making the journey and appreciate you.
Yes, sir. Thank you.
From the network to the call itself, trust takes accountability at every layer. You saw it today in the infrastructure we own, in the orchestration that keeps control in your hands, and in the models running right beside your calls. Because in the end, when the phone rings, there's a real person making a decision about whether to answer or not. Trust is what earns that yes. Now, to bring us home, I want to welcome back David Morken.
Thank you, Seth. We started today with a simple illustration, two tuning forks perfectly in sync. I struck one, the other one resonated. The power of physics. One note, 440 hertz. In your business, you don't just run on one note. You run a complex melody of communications channels getting amplified by super intelligence. You have voice, messaging, RCS, WhatsApp, human agents, and SI agents, and every one of them has a different frequency, and every one of them has to resonate. In physics, there's another word for the range of frequencies a signal can transmit or process effectively. Bandwidth. It's literally our name. It's part of our DNA. While you're asking yourself an important question, "Do I have the bandwidth I need to develop and deliver experiences for the next frontier of communications?" Remember, you are only as good as the infrastructure you run on.
Here, every connection, owned. Every choice, yours. Every millisecond, earned. Every call, proven. I hope we struck a chord together today, and whether it resonates is up to you and what you do next. We would invite you to build the future on Bandwidth. Let it resonate everywhere. Thank you for being a part of Reverb 2026. Let's roll.