Full demo of our biggest launch ever: AI Agent, Goals, LLM actions 

We just launched the biggest product release in Customer.io's history—eight new capabilities centered around an AI Agent that builds campaigns from a single prompt, measures their impact, and gets smarter the more you use it.

In this launch event, our team demos the AI Agent live—building a campaign from scratch in real time—and walks through every major feature in this release:

We built a single conversational agent directly into Customer.io that can do anything you can do in the platform. Build campaigns, audit existing workflows, analyze performance, draft content, and get recommendations grounded in your actual data. It remembers past conversations, accepts file uploads, and learns your preferences over time.

Call any large language model directly inside a campaign workflow. Score leads, generate personalized email content, or make routing decisions for every individual person passing through a journey — no webhooks, no external tools. Every paid account gets 100,000 AI credits to experiment.

Define the outcomes that matter (purchases, upgrades, activations) and see which campaigns and broadcasts are actually driving them. Attribution is counted by people, not events, and deduplicated across campaigns — so you see the real picture, not inflated numbers.

We've added two new channels that extend your reach into 180+ countries. WhatsApp has over 3 billion monthly active users; LINE is the dominant messaging app in Japan, Thailand, Taiwan, and Indonesia. Both integrate directly into your existing workflow builder.

Hit ⌘K to instantly find any campaign, broadcast, segment, or attribute across your entire workspace. Complete UI overhaul: We've redesigned the entire interface for speed and clarity, including a new home dashboard, updated navigation, smoother campaign builder interactions, and dark mode email previews for Outlook and Gmail.

MCP server upgrades (coming soon): Our MCP server is getting the full power of the Agent, so you can create and edit campaigns from Claude, Cursor, or any MCP client—with access to the Agent's memory and stored context.

CLI and Slack integration (coming soon): A command-line interface that lets LLMs interact with Customer.io more naturally than a REST API, plus an early-stage Slack bot that brings your customer data into the conversations where your team already works. Every channel connected. Every outcome measured. Every conversation making the next one smarter. The more you use it, the more powerful it gets.

Speakers:

• Naomi West — Sr. Product Marketing Manager
• Avi Goldman — Technical Director, Engineering
• Glenn Cornell — Sr. Product Manager
• Savannah White — Product Marketing Manager
• Nuno Sousa — Software Engineer

Naomi (00:03)

Hello, hello everyone. Hopefully everyone's tuning in. No technical difficulties, knock on wood. Welcome to "This Changes Everything." I'm not gonna do the classic like, oh, just wait a minute or two for people to trickle in. We're gonna get started right away. But welcome. I'm so excited to be here today. This is such a big launch for Customer.io, but also end users like myself. I'm a marketer, but also an end user of Customer.io. We have just released really the best AI features you've ever seen in life. We're excited to demo them live here with you today, share how you can get involved with feedback, talk about some of the things that are coming next. Of course, hear a bunch of compliments about how much you love these features. But of course, we're here to answer questions. It will be a live demo, so who knows what that will come with. But I'm excited for it. The team is excited for it. And thank you, everyone, for tuning in.

This morning or evening, wherever you're coming from. You might be a current user of Customer.io. Maybe you're like a week in, maybe you're a month in, maybe you've been here for years. You might be a partner of ours working to be here today to learn about what's launched so that you can best enable your own clients. You might have just heard about Customer.io on LinkedIn and not have any idea what we do here. We are a customer engagement platform.

And if you're watching as a competitor of ours, that's okay. I do it too, don't worry. I'm excited for everyone to be here today and hopefully learn about what we've just launched. So welcome.

Naomi (01:42)

So this was a really big release for us at Customer.io. We have a lot to cover today. I want to spend as much time as possible live demoing some of these features, which I will welcome Glenn up to the stage shortly to do so. But I want to start off with some admin, the classic. Will this be recorded? Yes, it will be. There will be time at the end for Q&A and comments if you have any of those. In general, we'll be demoing the new features that we just launched, including our agent — or your agent as soon as you start using it — goals, LLM Actions, WhatsApp and LINE, so these two new global messaging channels. Then we'll talk about the new UI, UX, and some hidden wonders like new features that came with this launch.

In addition to myself, Naomi — if you didn't know, I am a marketer here. I'm also an end user. I come from a background being a career email and lifecycle marketer. I was an end user of Customer.io before I joined. So it's such a fun job now to not only use the product, but then give feedback and help shape it for people like us to do better at our jobs. But this is the lineup of other speakers you will hear from today.

They are the strategy, some of the very few brains behind some of the features that we launched here. And they'll kind of be some of the experts that will help answer questions. And of course, I do have some colleagues in the chat as well that will be moderating. So there's no such thing as a dumb question. If you have a question — someone else — looks like there's almost 600 people here today. Most likely someone else in the room shares the same question. So please feel free to use the questions tab just at the bottom of Livestorm here and pop in questions as you go. I'll bring them up at the end or one of my colleagues will answer them live, which is a fun part of having a webinar be live.

Naomi West (03:44)

So starting with a general call out here, every time I do a webinar, conference, or an event online, this is the question that is asked the most. I keep a tally how many times this is asked just for fun. And it will be recorded. It's definitely going to be recorded. When will you get access to the recording? Soon. We will send out an email after this. Ben, thank you. Yes, here's my first one. Will it be recorded? Yes, it will be recorded. Feel free to pile on in the chat, in the comments. If someone asks this question, you can assure them that this will be recorded.

All right. So I want to start with some FAQs, some call outs. I'll shortly bring Avi onto the stage to talk about how we really thought about building the agent. What was some of our thinking behind it? What type of customer experience did we want to create and enable? But first I want to go over some general call outs. The first being that three of the features that we launched as part of this mega launch are in beta.

So we're actively shaping them. We're listening to feedback. If you have a request, send it in. I see firsthand every day customer requests come in and engineers pick them up within minutes of them coming in, whether it's being tagged on social or sending in an email or DMing me or someone else that works at Customer.io. The team loves this type of feedback. So if you see a bug or an oddity, please flag it. It could be something that's really super quick to fix to make your experience much better.

And we want this type of feedback when it comes to the agent, LLM Actions, goals. Just please know that these are in beta and we're actively still polishing and perfecting them.

Naomi West (05:34)

The second call out — it's just such an exciting time. I think to be a marketer or a product developer or really anyone in the industry, it's just such a period of rapid learning in our industry, right? I really do consider myself a lifecycle marketer. So even though my title is product marketer, I'm actively building emails and in-apps and customer communication touchpoints. I built the launch email myself, but I'm also an end user of the product that they're building. And with all of these new features, I'm learning about them so that I can talk about them as a marketer with all of you, but I'm also learning how to apply them myself. So when we release these features and I have people reply back to me via email with a use case, I'm often in awe of how you all as customers are using the features yourself.

So talking about use cases and how we're integrating AI into our own workflows and tech stacks — it's the best way to discover new areas of opportunity. And I want to call this out: as consumers, as marketers, we're all learning in this together, but same for the product side too at Customer.io. We're learning how you want to use these features, how you want to integrate them. So yeah, it's just a fun time. If you have learning you want to share, if you have a question you want to share, please do so. We're all learning about these features together.

And then the final call out is that one of our new features, LLM Actions, has come with a newly introduced credit model. So all paid users of Customer.io on a paid plan have received a sample allotment of a hundred thousand AI credits to experiment with. We'll be explaining LLM Actions and the potential use cases for them in a couple of slides. So if you're like, I don't know what this is, don't worry. We will inspire you shortly.

And then my fourth call out that I didn't list here, but it did come to me as I was falling asleep last night. This is a live demo. Okay. So if there's any oddities, please be patient with us because a live demo, you just never know. I think it'll go really smoothly, right? As one does. But important for me to clarify that.

Naomi West (07:55)

Moving on, talking about the launch story. So this launch has a handful of individual features that when used in silos add value, sure. But when combined — combined all of them together — they create this really fantastic story of a platform working together. And when I initially kicked off, all right, we're going to do a big launch, how do we tell this story? It came really clearly in the form of this three-pillar story, starting with the launch of the agent.

We are introducing with the launch of the agent intelligence that compounds. We're introducing this new way of working where a platform learns your business specifically. For example, in your workspace settings, you do have a business context hub. If you haven't visited before, go into your workspace settings and then click Business Context where you can go in and fine-tune really the foundation of how AI is going to craft content around your brand.

Naomi West (08:55)

You will have the ability to add in your voice and tone, your business, just some general background on what your business is and how you want to speak to your audience so that our AI can learn your business and your audience. Which leads us to this intelligence layer. We're looking at your event and attribute names to understand what context you have on your customers. We're looking at your goals. What are your conversion events? Let's start with the outcome — what are you trying to achieve as a marketer, product person, and engineer? And we will dive into goals in a second here.

Secondly, this launch extends your reach through the new addition of WhatsApp and LINE. WhatsApp has over 2 billion users globally, which is more than I can comprehend. But it's such a huge channel in the same way that SMS emerged what feels like five years ago as this new hot channel for marketers to start to lean into. And I feel that way about WhatsApp. I'm seeing it everywhere. I don't see as much of LINE personally, but it does have a huge reach in markets like Japan. And we see this through our own customers here at Customer.io. So really fantastic channels that we're excited to introduce.

And they're the next evolution of extending your team's reach into global audiences or new communication avenues wherever your customers hang out. In this final phase, the platform story that can be told with the release of these new features is that we're creating a new layer of clarity. As an email marketer myself, a career email marketer, I've seen such an evolution of my job's worth being tied to clicks and opens, for example.

You know, sure, I can lightly look at a conversion event, but I've always wanted to see the larger reach and the larger value of my impact. Like what are all of my touchpoints doing to contribute to a singular goal? A singular goal, meaning — me being a primarily email marketer in the SaaS space — I'm trying to increase subscription count. I want people to subscribe to my products, whether it's Netflix or Customer.io. I'm also focusing on win-back, trying to bring back churned customers into my business.

And it's not often a singular touchpoint that reaches those goals, right? So I have a whole toolkit of campaigns and broadcasts contributing to these goals of mine. And now with this release of Goals, I can see how all of those touchpoints are laddering up to that one North Star metric. So it's a total win for me and I'm sure the rest of the team. If you see one of these features today and you're like, that's amazing, I'm going to integrate that —

Naomi West (11:46)

— specific feature, but it's an even larger win if you pick up the features collectively and are able to see the value of this intelligence that compounds and the fact that you can reach more people and that you have more clarity into your work.

Are you tired of me talking about AI yet? You guys are like, enough, take me to the demo. I'm almost done. I'm not done. I still have more. I know we will get into the demos, but I do think it's really important to talk through our philosophy of why we're building things. We're really intentional here. What I've noticed — the engineers and the product managers and the leadership team — they're intentional about how we build AI.

We're not just slapping AI onto our platform to say, hey, we have AI now. We really are thinking through what we want to build, how we want to integrate it. And it can be thought through our philosophy in these four quadrants. So ideally, AI earns its place by learning your business and the context accumulates over time because it collects all of this background information. And then it really grows with you over time.

So the first philosophy is that we want to create a world where context compounds. And the second quadrant is that we want to create a world where speed with control drives progress, meaning that Customer.io's AI handles repetitive work while you remain in control. You're viewing outputs, confirming impact before anything goes out, leading us to the fact that we believe AI belongs in the work — not beside it. So it's deeply ingrained and this intelligence should live inside your decisions and your logic and your execution so that you as a marketer spend less time stitching things together and more time actually thinking through what's next.

And finally, we really believe that sophistication should be accessible. If you can describe what you want to do, you should be able to build it regardless of your tenure, your role, or your technical background. One thing that I'm absolutely loving about this new world of AI is it's unlocking a lot of really cool scenarios and opportunities for me as someone who identifies as a non-technical marketer.

So now that you have that foundation, I'm going to pass it over to Avi, who's one of our talented engineers here, but a key thinker in building the agent. He's not only strategizing the why behind the agent, but he's actively building it. He's going to talk through why one agent — a singular agent — and also outline some things to come.

Avi (14:38)

Yeah. Hey everybody. So yeah, like Naomi said, I just want to give a quick behind-the-scenes peek at how we're thinking about AI and agents at Customer.io. I think we've seen a bunch of these launch, and especially in our industry, we see this kind of segmentation — like a copywriting agent and a creator agent and a deliverability agent. And while we played with that idea, the philosophy we landed on is that having a single agent actually creates a lot more value and power.

Think about how you approach your own work where you create an email, and then you analyze that email, and you're going to worry about the deliverability, and you're going to think about how that fits in with the entire customer lifecycle and how all the data comes together. You are one person, and creating those boundaries actually doesn't provide any additional value. It actually makes it harder where you have to do handoffs between these different tools. So we really wanted to focus on providing this one agent, this one entry point, and make that a really smooth experience.

The other important idea was making it so that the agent could do everything that you could do through the product. So instead of trying to rebuild our product with agents in mind, we said, what if we give the agent access to the product — give it the full power that you already have through the UI. And that's what we did. And so you can ask the agent to do basically anything you're able to do in your workspace. It has that capability and is ready to help you with that.

That's just a really important idea. As we're looking at everything, remember that the agent has all of that behind the scenes. And the last thing is memory. So building context — Naomi talked about the business context and how you can fill in details around what your business is, your brand, your tone. But also talk with the agent. It has a memory. So you can say, hey, remember what we were working on yesterday? Let's keep working on that today. And it will know that. It will remember that. And it can learn to work with you better.

And so it's really not just a chatbot. It really is one of those full agentic systems ready to go, built out for you. So all of that comes together with this idea that your workflow lives in a bunch of different places and we're building so that AI is there with you whenever you're doing work. The agent is with you in-product. It's with you when you're building a campaign or building an email, with you when you're reviewing a profile.

Avi (16:57)

And then when you move over to Claude or Cursor or ChatGPT, we have an MCP server and we're working on that. Nuno will talk more about that — so that your AI is with you when you're in those external tools and Customer.io kind of moves with you. And if you're using Claude Code or you're building a full agentic platform internally, whatever you're doing — we're looking at, and Nuno is going to talk a lot more about this — an Agent CLI so that we can meet you even there, so that all of the power of Customer.io is able to be used with AI at any point within your workflow.

And I'm going to invite Glenn up to give us a demo.

Glenn (17:45)

Awesome. Let me go ahead and share our screen here. Hey y'all, so here we have the homepage. A lot of buildup here — I'm going to be talking through the agent functionality. This here is a new account. We've already pulled in some data, but it's mostly going to be working towards building new things into this account.

We've added this Agent panel. So you can see all of your past chats, kind of give you a focused view of your conversations. But you'll see, and I'll pull this up as we go, there's also the sidebar experience. There's full awareness of what screen you're on so that you can actually be building while you're talking to the agent as you work.

A couple of things I want to point out. First off, this is all part of a made-up website, a made-up business. Over here in Context, this is where you'll see a quick description of what we're working with. Lenora MetaSense is a generative AI creative platform for marketers that creates content. I'm going to be working towards a campaign and a goal area of this business.

Go ahead and get started. I'm just going to show you a super simple use case here. Let's just say I want the agent to remember that this quarter, credit pack purchases — that's the big KPI.

I'd say that usually when you're getting started with the agent, having these kinds of conversations about what you're focused on, what might have changed, it can just be helpful in getting more personalized responses. So in this case here, it's just taken a quick note. To make sure that OK, going forward, here's my Q2 2026 KPI. Now hopefully this will play out later in the demo as I ask for other things. We'll see, because demoing a live agent can be fun. It remembers what you've done in the past, so you can't really practice too much. You gotta do it first time through.

So mix in a little bit of dry humor as we go, just to liven things up a bit. I personally do enjoy just making my agent personal to me. It's everyone's agent — for every user, it actually remembers you as a user. So it'll behave a little bit differently for you, however you want it to.

So go ahead and add: while you're at it, my birthday is November 18th. What can you do with that? So it's kind of telling me some different ways that it can take that. It could remember it personally, build something — campaigns around birthdays. It can even set a routine. So it'll send you a little birthday email on your birthday if you'd like. So it's remembered my birthday. Cool.

Glenn (20:34)

But let's get into some actual use cases here. I'm going to start a new chat, just give it a fresh look. Here I'm going to say: I want to grow revenue for my existing users. What would you recommend based on my data?

Now this is a pretty generic one. Again, this is a generic account, so if you all have some better specific use cases that you've done that have gone more in depth or worked out really well for you, drop those and share with some other people in the chat. In this case, when you're asking something like this, it's going to try to be consultative. It's going to try to look at your data and it might ask a few questions back and forth about what it is you want to do.

It's also probably going to steer towards content creation here because we don't have a ton of content in this account. And it's honestly always kind of interesting to see where it goes. One other thing I'll say when you're having these early interactions with the agent — it's really good to treat it like a new employee on your team. You might have certain preferences. It might lean towards a certain direction. If you don't agree with where it went, just let it know, and it will remember that.

So this is the fun part about having a totally fresh demo environment. Zero active campaigns — but it's pointing out here that an expansion revenue campaign looks like the biggest impact, and we do have one active campaign that it wants us to audit. I'll come back to that one later, but let's lean towards this expansion revenue campaign. One of my favorite prompts to give back is just: yeah, let's do it.

So it's basically saying it's going to build out this campaign. Now, it might go a little bit back and forth asking some questions about my preferences. It's going to look across the data that it has access to. Maybe there's certain events that would be a good trigger here. Do I want it to be segment-triggered or event-triggered? And it's giving me some examples of what it's found. I think credit pack purchased would be a good one because we mentioned that earlier. And what upsell actions would the campaign focus on? Yeah, let's do pack purchases.

I'm just going to stick with one for now and submit. So if you use other tools like Claude Code or other LLM-based tools, sometimes they will go into a planning mode here. This is like planning, and we're getting coordinated to make sure that it's not going to build the wrong thing. Alright, confirming this all. Looks good.

Glenn (23:13)

So one part here is that a campaign can take a little bit longer. Let's see what we got. Which event triggers the campaign? Let's do when folks' credits are low. That way, we'll reach out to them to purchase more.

So as mentioned earlier, it's going to pull things like your global styles if you have them. It's going to be checking business context for the language to use. When it creates a campaign, it can go pretty deep as far as creating the emails for that campaign and the copy for that campaign. Once the campaign's been created, I do like to go in and work with it, talk back and forth to do some of those edits. And sometimes you'll just get a recommendation that you want to make the change yourself.

You can just go in — you don't always need the agent to make the update. Now it's going into the actual creation flow. Now I'm going to go ahead and approve all changes. It could take two to three minutes for it to actually create the entire campaign and show it to me. I went ahead and just created a campaign ahead of time that we can go take a look at.

Glenn (24:21)

Let's pop over into campaigns. This was brought up earlier — take a look at the workflow here. Now a couple things: if you open the agent, it'll pull you right back up on the last conversation. I'm going to start a new chat because we're looking at a new campaign. I'm just going to say: take a look at this campaign and give me some recommendations for how I can improve it.

So this is a basic type of prompt that I like to give. It'll try and go into consultative mode. It'll look at the campaign's performance — in this case, it's not a lot because again, these are all test people. So my metrics are going to be pretty low. It's just going to look through a couple of things. Look through general setup — are there things that are clearly missing? It'll help with analytics. It's going to look at the decisions you've made and whether there's additional logic we should add or additional steps we can take to make this work. And it's going to look at the actual sends — do you have massive drop-offs? Lack of opens? Are there bounces? Just to try and make sure we're improving the fundamentals.

So it's rightfully noticed a critical issue here — my emails are not performing well. Again, one of the funny parts of having a demo account. But let's see — no conversion goal set. That's actually a really good catch. I could say "could you set the conversion goal for me," but actually, I want to do this one myself. It should be updated to credit pack purchased.

So I'm just going to go over here and click this, and let's just go ahead and set the goal. Let's make the event credit pack purchased. Cool.

So yeah, you don't always have to have the agent make the updates for you. Sometimes it's just a nice reminder. Again, kind of like how I work with the team — they might point out a few things, and I go and do it myself.

Glenn (26:34)

Let me close this and take a quick look at the other campaign we have going. And as you can see, I've kind of become a multi-tabber where I do often have the agent running in the background, going and working on other things. You can see the progress that it's making — it's working through each of the templates and building out the emails for that campaign.

While I'm here, I'll just point out a couple other things about the agent. If you go to the homepage, there's a "what should you work on" section. It knows what you've set up so far. And so these recommendations are going to be based on your account. If you haven't set up any segments ever, it's going to recommend segments. If you have tons and tons of segments, it'll recommend auditing those segments. I would definitely say as a little trick — jump over to this section, take a look at what it's recommending, and give it a shot.

Another little trick here: file upload is quite nice. If you have things like a style guide, just drop it in — any information or context that you might give a new employee, provide that context to the agent. It organizes these files, it remembers them, and it uses them when necessary. You've also got the Business Context section to go in and specifically give it instructions about your business. And if you have any other type of context you want to provide, you can drop that in as well.

Let me demo a couple of other features here, but I'll cut so that we can shift over. We'll talk about LLM Actions next.

Glenn (28:26)

Next, we're going to talk about LLM Actions. With the campaigns earlier, we've been going through the agent, but we've also done a few things here that don't require chatting with an agent to get that AI experience. So the way that LLM Actions works — it's kind of like taking your campaign, taking that workflow, and you add a specific step where an LLM or an agent is essentially interviewing every single person that's going through that campaign.

Let me open this up so you can see. This allows you to write a prompt that everyone going through the campaign will then be evaluated against, and that prompt will add journey attributes for each individual person based on the criteria that you've set. So I added a little sticky here so you can see: this workflow is identifying free users that are heavy users. They're not yet paying us — maybe they're trialing, maybe they're using some free credits that we've provided — and then this will send a personalized upgrade message for those folks that are highly engaged.

Now, I think a pretty good use case that matches both how marketers should be thinking and how you should be using LLMs is that you want to put most of your emphasis and effort on some of your best high-performing people, the ones that are most likely to make the purchase. So one example of an LLM action that we've added here is essentially a lead score.

So I'll jump in here and give you a look at what this looks like. You write up a prompt for this action. In this case, this is a lead qualification action. And it's going to look through every single person that goes through this campaign. It's going to look across this information here, and I've given it a guide on how to score each individual lead. It creates scoring, and then every single person that goes through this action is going to come out with a score. There's going to be the qualification itself, and then there's going to be some reasoning. And that's because I've added the outputs as the following fields.

Now, the funny part about this feature is I actually didn't work on this feature. I got introduced to it the same time you all did. And so the way that I built this was I pulled up the agent myself. You can even see my history here — I went through "what are some good LLM Action use cases" and then actually had the agent work with me to build this out.

One of the things that can be really intimidating at first is just seeing this prompt. Like, how would I know to write this prompt? You can just tell the agent, hey, I'm building out a workflow, here's the goal, and I would like to score everyone that goes through. The scoring came up because I asked the agent for good use cases of LLM Actions. But then the actual prompt writing — again, you can use the agent in this way to help assist you. It has access to our documentation, and it also has access to how Customer.io works.

Glenn (31:34)

There's a few other things I can point out here. We have fallback values that still exist, if you want those fallbacks in the event that you don't have a specific output for an individual person. And there's a lot of other ways you can adjust this. We have Gemini 2.5 Flash — it's a cheaper model. I'm using this at the first step. And maybe later on when I'm doing personalization, I might use a more expensive model when there are fewer people going through. So I'm just trying to be tactical in how I'm using my tokens here.

So back to the workflow — that was the lead scoring. And then I've got a branch here that's going to say, hey, everyone whose score is above 50, let's put them over into this side. And then everyone who scores below 50, we're not going to mess with them. We're just going to assume that maybe they're not getting as much value. Maybe I'll do something later with that.

So over here, we have another LLM action and another prompt that was helpfully created by the agent for me that essentially says, now you are a marketing specialist who's going to write content. And it's going to go through and look at every single individual person. It's going to pull these attributes from those people. And it's got instructions to write out email subject lines and email bodies. And for every single person that goes through, you're going to get that journey attribute added. And when the journey attribute is added to that individual person, when they receive an email, this will be inserted in that email.

So it allows you to take every single person that goes through your campaign — once they hit this step, stop, interview them, learn a bit more about them, understand the right email to send, and then send a personalized email to every individual. This is really fun to play with. It's also extremely flexible, as you can see. I would recommend going in, trying out the agent on the consultative side and saying, hey, could you take a look at this campaign I've created? Can you give me some recommendations on LLM-based actions I can add to this campaign? Start playing around, see what works, see what doesn't, and give us feedback. As I mentioned, this is in beta.

Glenn (33:32)

We think it's going to be really powerful for a lot of different use cases. So you can see that then you would add the "send personalized email" and all I have here is the journey email body. This would be a plain text email. I'll go through and just look — I had this sent out just yesterday. You can see it went through. This person was added to the true column. It ran the LLM action and created a personalized email. A real marketer would probably polish everything up a bit more, but hey — a personal email that's basically speaking to that person in a way that you really could never do if you were just blasting out a message, even with standard personalization. Cool, that is it for LLM Actions.

And I will go ahead and jump straight into our next feature, which is Goals.

Glenn (34:30)

So taking a quick breath so folks can transition. Now Goals is totally different, but it's also using our new AI functionality. The way to think about Goals is: when you come into your day, your week, your month, you probably have these high-level things that you are trying to move. Maybe it's the thing that would get you a promotion. Maybe it's the thing that your company is currently focused on this quarter. Goals tend to be those numbers that you most want to move, that all of your other campaigns and broadcasts are hopefully working up towards. So it's not an open or a click. It's more usually close to revenue or a lifecycle stage change — going from a trial customer to a paid customer.

So when you create a goal, there are a couple of ways that you can do this. We've gone ahead and created some recommendations here. One of the fun parts about live demoing is we actually have some updates going through today. So this might look a little different even an hour from now. When it generates these recommendations, it's looking across the best practices that we've provided for Goals. It's also looking at your data. And so it's going to pick a specific attribute or a specific event that you're working on. And it will notice — this looks like a goal.

I'm going to go ahead and just do it from scratch so you can see how this looks from a raw perspective. But if you chose one of the recommendations, it'll just jump in, pick the attribute, give you some suggestions. You can work from there.

So going from here, I'm going to use that credit pack purchased event that we talked about before. I'm starting with an event. And because this is an event, I can take an event attribute where the revenue itself is stored — in this case, amount paid. And let's say that this is going to be euros.

So once I've done this, we're going to look across every single one of those events that's been generated, see how many times that's happened. And then I added the revenue element to this on April 7th, as you can see, but it'll actually pull in the revenue that's converting with each of these goal conversions. So I have a big picture here: there's 230,000 euros that have been generated from this goal across 10,000 different people.

Glenn (36:46)

You can play around with these numbers here. You can look at this based on count, cumulative, and more. But what's really most helpful and effective is getting the specific campaigns and broadcasts that you're already working on and seeing how much they are contributing towards that goal. I could suggest sources — again, this is the piece that we're actually actively working on right now. So in an hour from now, it'll probably take that campaign we just created, notice that it has this exact goal in it, and then recommend that campaign.

So I'll go ahead and just take a look here. There's one, a few others — I'll just grab some. Now there are no actual sends here that are going out to real customers because, again, this is a demo account. So lots of zeros. But for you, if you go in and you're taking campaigns that are actively sending, this will take the attribution of every single one of those campaigns and actually break it down by the individual messages. So you can see which messages and which campaigns are driving these goals, what percentage of the time that's happened, and then a breakdown that shows you how much revenue is being generated by each message.

Another cool part of this is the edit settings. You can see all of the different types of channels that we provide here at Customer.io — email, in-app, SMS, WhatsApp, and LINE. And for each of these, you can choose how attribution is defined. If you say, nah, I don't want to count an email unless they actually open it, go for it. For in-app, if you only want to count it when they open it, go for it. We've also just recently included LINE and WhatsApp to support those new channel releases.

Save settings. And now I have a customized view of how often folks are purchasing credit packs, which campaigns and broadcasts are driving them, and how much revenue is coming from those messages. There's a lot more that we're working on for Goals. You won't see milestones here yet, but that's something we're planning to add so that you can actually start to measure how these goals are progressing and get some deeper analytics. But that's some of the stuff coming in the future. I think that's all for me for now on Goals. Thanks, everybody, and I'll pass it back to Naomi.

Savannah (38:59)

Hey everyone. My name's Savannah. Thank you, Glenn, for going through all of those updates. I'm really excited to be here today to talk about two new channels that we've recently added. So before I get into that, really quickly — I want to say the features you just saw are really, really powerful. And there's a lot of potential there. But they're only as powerful as far as they can reach. And so that's why I'm so excited to share that we have expanded your reach to these two new channels: WhatsApp and LINE.

And for many of you already running email, SMS, push, and in-app in Customer.io — adding these new channels, LINE and WhatsApp, can be seamless to your existing workflows. There isn't a new platform to learn. It's just two new places to send to from the one you're already in. Let's take a deeper look.

So if you haven't heard of WhatsApp — you likely have. It's considered to be the most popular messaging app in the world. They have over 3 billion monthly active users across 180 countries. And with open rates north of 90%, that is some serious reach. And if you haven't heard of LINE, LINE is the dominant messaging app in Japan, Thailand, Taiwan, and Indonesia. Markets where email and SMS simply don't perform the same way as they would in other countries around the world.

And so for global teams, or really anyone with customers in these regions — these really aren't edge channels. They're the future. They're where your customers actually are. They're the apps your customers are using in their regional areas of the world. And now we've made it possible for you to seamlessly connect with those users on their preferred channel right alongside your current workflows. So you get that unmatched global reach. You get to meet customers where they are on the apps that they want to use. And you get to do it all from the same platform that you already know and love.

Setup is lightweight by design. For WhatsApp, there's a native self-service flow. You connect your Meta account, you sync your approved templates, and then you can start sending inside Customer.io. And for LINE, you can connect it through the integration steps, and it appears right alongside all of your other channels. And once you've connected both those channels, they live in the same workflow builder you already know and use. You can drag and drop, you can personalize with Liquid, no separate tool, no disconnected workflows.

Savannah (41:20)

In fact, let me show you. I'm going to hop over into Customer.io and show you a LINE message in a campaign, and then we'll build a WhatsApp message together. Let me share my screen.

All right. So here we have a welcome campaign set up for Lenora. You already met Lenora, our fictional AI company. Lenora sells to customers all over the world, and different markets, as we all know, prefer different channels. On the right-hand side, you'll see all the channels you have at your disposal — email, in-app, push, SMS, and now WhatsApp and LINE.

And you can drag these new channels into your campaign builder like you would any other channel in Customer.io. I'm not going to be dragging and dropping them in because I've already set this campaign up, but you can do that in your own builder. You can see how easy it is to build out your channel logic by region. I've kept it pretty simple for this demonstration, but we can take a look here and see that for my customers in England, I'm sending them a WhatsApp message because that's the top app that they use there. For our Japanese customers, we'll send them a LINE message because that's the top messaging app in Japan. And then for our US-based customers, we will use SMS.

Perfect. So let's go ahead and take a look at these messages. We'll poke into our LINE message. You'll see here I've already built it out. You'll notice it supports Liquid personalization like it would in any other channel. We'll see our first name, plan name, plan credits in the preview here. It looks great.

And let's go ahead and build that WhatsApp message together. We'll edit our content. So WhatsApp is actually a little bit of a different process than some other channels. You can't always send freeform blocks of text. You have to use pre-approved message templates. But once those templates are approved and in sync, you can simply drop in your variables. You can preview the message and then you can start sending.

Savannah (43:30)

Let's go ahead and add some personalization — my name, the plan name, and then some plan credits. And we'll see that that populates exactly like I want it to. I can see the name, Pro plan, my credits. That looks great. So I'm happy with that. We'll save our changes. And then I will pop over into my WhatsApp.

Okay, so this is my real WhatsApp. I use it to communicate with my family based in England. And yeah, that's what the message looks like in my WhatsApp. I see that it came through exactly like I wanted. It's my name, my Pro plan, my credits. I'm happy with that. So this is exactly how your message could look in your customers' WhatsApp accounts.

Let me stop sharing this screen and pop back over to the slides. Before I pass it off, I just want to say one more thing. Now that you've seen these channels in action, it's clear that this intelligence and this reach is really, really powerful. And with WhatsApp and LINE, that's potentially 180-plus countries and maybe even billions of interactions that are waiting for you to take advantage of. These are channels your customers are already living in and using daily, and you can reach them from the same platform that you're already using to send your other messages.

And with that, I will pass it over to Nuno.

Nuno (45:17)

Thank you. Hey everyone, I'm Nuno. I'm a software engineer at Customer.io. I'm so excited to show you three things we have today about AI and agents at Customer.io. Starting off right away — the energy in the chat is amazing. You guys have been asking for this.

So the MCP server — we launched that last year. We learned a ton from it. It allows you to read information, allows your MCP clients — your Cursor, your Claude Desktop, whatever you're using — to retrieve information from Customer.io. But it was really limited in its ability to make changes in Customer.io, in your workspace. And so now that we have all this power for the agent, we want to bring that to the MCP server as well.

So the MCP server will get all of this power, everything we've talked about today. It'll be able to create campaigns and segments and broadcasts and edit them. It'll even be able to fetch the stored information from your agent. So your agent has memory, just like Avi mentioned. Your agent has user preferences. Your agent can store files that it might need for some requests you might make. And so the MCP server will be able to reach into that data as well.

That'll enable some key use cases where you can start your work through the agent if you want, and then just continue on in an MCP session and just say, hey, remember this campaign we were working on? Let's keep it going. And finally, you'll also be able to use routines. Routines is a concept we've introduced to the agent — you can ask the agent to set something up for a weekly schedule, for example. And you can ask it to send you a report of a campaign analysis or anything you might want to do. And you could do that as well with MCP. You can set up some of those routines. So that's number one. That's coming very soon.

Nuno (47:14)

So CLI. Yes, we will have a CLI.

Bear with me — why would we want a CLI? We've already got MCP, you've already got the agent, and we have a REST API, right? So you could ask your Claude or your ChatGPT to just use our API and interact with Customer.io. But there are some key differences with introducing a CLI. Here's a quick example.

The CLI is much more natural for an LLM to explore as opposed to a REST API. This is what we found — LLMs are actually very, very good at using the terminal. The frontier models have been especially trained for terminal use. And so we found that giving it a CLI where it can just do --help and figure out how the CLI works is actually very successful.

This CLI will allow you to create a campaign or fetch the schema for a campaign. So that's one key difference — why this is actually better for an LLM as opposed to a REST API. With a REST API, if you ask an LLM to go to our API and create a campaign, it might just completely make up the payload, right? It doesn't know the payload. But with a CLI, it can discover the payload. It can do cio schema campaigns and it'll find out the payload and then submit the correct request. So it's really making your AI interactions much more successful.

And then the last thing is skills. It's not just about knowing the correct payload. To be able to use Customer.io effectively, there's some things you might want to know — such as before creating a campaign, you might want to create a segment.

Nuno (49:41)

And so we will ship some skills alongside the CIO CLI. Your agent will be able to simply ask, how do I do this? How do I create a campaign? It'll retrieve that skill and then follow that playbook, so to speak. So it'll make these interactions much more successful.

And by the way, this isn't just for agents or AI. It's actually very cool to use for humans — much better than trying to do curl requests directly. So you could use this in a script that you're writing. It's not just to give to an AI. And by the way, our agent will use it too internally. The way that our agent will communicate with your workspace is through the CLI. So you can be sure that anything the agent can do, the CLI will be able to do as well.

And then finally, the third thing I want to cover — we're exploring a Slack integration. We're still in the very early phases. We're exploring what it means to have a Slack bot in your Slack workspace that could talk to your Customer.io data. The way this would look is you would install the bot onto your Slack workspace, and then you could just @ the bot — @Customer.io — and maybe ask it some questions. Some of the things we're thinking through: maybe those long threads that you've been having debugging an issue — why didn't this person receive this email? You might just ask the bot, hey, why didn't this person receive the email? And it'll see the whole thread, figure out what you're asking, and then respond. There's a lot to figure out here. This is very early on.

But we do want to bring AI and these LLMs and the power of the agent wherever you work. Slack is certainly a big part of that. So yeah, those are the three things I wanted to show off today. There's a lot more coming for the agents. We're actively improving it. I want to pass it back to Naomi now. Thanks, Naomi.

Naomi West (51:59)

Thanks, Nuno. I hope everyone liked the demo, hearing from Nuno, Savannah, Avi. Certainly a small launch for us. We didn't launch much at all. But it's been a really fun past couple of weeks. So the features that we just walked through — clearly a heavy focus on AI — but this launch also included quite a few features that hopefully will just let you work faster every single day that maybe aren't as AI-forward as the rest of the headliners.

The first feature is actually my favorite. Easily the best feature — and just realizing now my GIFs don't work — Message Invaders. So let me actually just — I can't have you all not understand Message Invaders. I'm going to share my screen really quickly and jump over to it because it's my favorite feature.

We're in Lenora. Down here in the right-hand corner, you're going to see Message Invaders. Really, all you have to do is hit play. And while the agent does your regular day job for you, you can challenge yourself and your colleagues to a high score. So I've spent probably way too much time on this. We don't need to talk about it any further. But it's my favorite feature as part of this launch.

High scores — if anyone wants to share them, let me know. If you want to try and beat my high score of 147,000, you're welcome to.

Naomi West (53:58)

Next feature — the homepage. Back to serious new features. We've released a new homepage. It comes with visibility into things that were recently being worked on by myself. This is a huge win. I no longer have to go into the broadcasts first or the email designs that I was working on — I can clearly see what I was working on in my last session and jump right back in there. Another one of my favorite features of the new homepage is I can see what other people are working on, which I am always curious about. That's all I'll say about that. It's really helpful.

Moving on — we have the Performance Dashboard. You now have a new performance dashboard, workspace performance. If you toggle your settings button in the top right of your navigation, you will see Workspace Performance and you should ideally see a little healthy status. You can click on that and jump in, get an overview of your workspace performance. I really like the clear optimization boxes here. It tells me that I have the opportunity to archive a thousand unused segments. We won't talk about that. But just a really great insight into what's working in my account and areas of opportunity to clean things up.

Moving along — Universal Search. This is a really nice quality of life improvement in my opinion. In the same way that I use the "recently worked on" section on the homepage, I can also jump into universal search and quickly search something like onboarding, broadcast, mega launch email, and have it pop up.

Naomi West (55:23)

These are ordered in a manner that makes it easy for me to jump into what I was recently working on. And profile search is coming soon. So if you're currently in there and throw in an email address and you don't see a profile matching that specific email address, it is coming soon. It's being worked on and will be added at some point.

Moving on. Okay, so this is a random one, but it's always my opinion with product launches that sometimes the smallest ones are the best ones. And I have a really big space in my heart for Design Studio. So I wanted to shout out this one: if you're currently using Design Studio to build your emails, when you go to preview an email, you can now preview your email with forced dark mode specifically for Outlook on web. And you can also preview your email with forced dark mode specifically for Google, which includes Gmail and Android apps.

We all know that dark mode is a pain for us as email marketers. It's great as a consumer. I love it as a consumer. You will not catch me after sunset without dark mode enabled on my phone. But hopefully this preview will give you as a marketer or developer or designer a little bit more insight into what your email looks like. And I love the suggestion — we'd love to be able to export dark mode screenshots. Content squad, if you're here with me today, make it happen. You know what I'm saying?

Naomi West (57:00)

So where do you learn more? That kind of wraps up our demo, looking at what we've just launched. Hopefully you all like it. I know we're really tight on time, so I don't think we'll have time to answer all of the questions — apologies. But what I will say is feel free, you can email me directly. If you got an email from me for this launch and you're like, I'm frustrated that I didn't get my question answered — you can email me directly. That's fine. I actually am desperate to receive emails.

Email me directly, I'll find an answer for you. We also have some of our content team in here, some of the marketers that will be helping me put together probably an FAQ of all the questions that were asked and we'll send that out at some point. I'm committing to things that are not in my power. They could be in my power. Yes, there we go — Molly, thank you. 100% an FAQ blog post coming soon.

Anyways, back to this. If you want to learn more — if you're a current customer, please reach out to your CSM. If you want to chat about specific applications of some of these features, you can also reach out to product@customer.io (mailto:product@customer.io) to give feedback. Or if you're like, I want the Slack integration — join the club. But this is probably a great avenue for you to request early access. We love giving early access and we love working with people in the early days of the features.

If you are a partner, reach out to your partner manager or email partnerships@customer.io (mailto:partnerships@customer.io). And if you just showed up here, you have no idea what Customer.io is, you're not a partner, you're not a customer — welcome. You can definitely start a trial or you can book in for a demo. We will be sending a follow-up email with these links after the call. So you don't have to go fishing around for them on Google.

Naomi West (58:50)

Again, we love feedback. I have to reiterate this because I feel like most people don't take it seriously when I say this. I love when people reply to my emails. Okay, I don't like it when they're mean — and as the group of email marketers here, we all probably understand the mean reply. It's always interesting. They're always unique in their own fashion. But the team loves feedback, whether it's positively constructive or bluntly constructive, or just like, I love this feature. We all love it. So give us all the feedback.

Again, if you're enjoying these features, if you have requests, please share them with our team. So much so that the team is really, really down to take feedback. As an example, we were tagged late last week on X, I think on Thursday. I love this "100 out of 10" rating — it made my day. But we got a request: please, why can't the agent import a list for me? I can't upload a CSV file in the chat. And Monday, Tuesday this week, the engineers were like, yep, sounds good. They put their heads down and you can now upload CSV files into the chat. So really positive. I love it.

I wanted to finish this on three FAQs that we have gotten quite a bit. Things like, can the agent work with [blank]? LLM Actions, can it work with Goals? Can it work with my campaigns, broadcasts? The agent can do anything within the workspace that a current user can do, minus some destructive actions — without user confirmation. So every edit or deletion the AI wants to make must be approved by the user, either individually or with a broad approval.

The second biggest FAQ: will my data be used for training models? No.

And: can I disable AI in my account? Yes. Your admin can disable AI via the account settings. Hopefully, if you've been here for the past hour watching this webinar, nobody on your team wants to do that.

So just want to thank you all for joining today. This is where we would do Q&A, but the day must go on, sadly. But I will go through all of the questions with my team. We'll get some answers out. And again, feel free to email us at any time. Thank you so much everyone for joining us. Hopefully it was helpful.