Explore the Next Wave of Customer Data

JOHN GRUNDY: Everybody — is it on? Here we go. Good afternoon, everybody. My name's John Grundy. I'm the senior director for product management here at Amperity. I've been with the company for about four months, so go easy on me. I wanted to see by a show of hands who's in the room. Those of you who have a marketing bent to your job, can you raise your hand? Okay, a lot of you. Those of you that have a technology bent to your job, raise your hand. Lots of other folks. And everybody else who didn't raise your hand, raise your hand now. Okay, good. It looks like a pretty healthy mix of marketing and technology folks. I'm here with my colleague Grau, and we're going to present some thoughts about the roadmap going forward — that's the main topic today. I hope that's why you're all here. Thumbs up. Okay, people nodding.

JOHN GRUNDY: The agenda is pretty straightforward. We're going to talk about the keynote a little bit and tie the product names to the things you saw this morning, so you'll understand which products play in which part of the keynote and how they look in the live world. Then we're going to talk briefly about our development themes and how we think about our development process. Then we'll tie the products back to those themes so you can see how they all relate together. And then we've got time for Q&A at the end. For my part, we have a half-hour break after this one, so if anybody wants to stay and ask more questions, Grau and I will be here. My boss Gry is in the corner over there, so between the three of us we can probably answer most of the questions you have. And we have a special guest here — Bryce, raise your hand — from our engineering team, who can also help with questions. So fire away as we go through. I want it to be as interactive as possible.

JOHN GRUNDY: So the keynote you'll remember from this morning — this is Susan and her husband Jack, and Jess was our marketing person. We're going to talk about their lives and what was happening in the story you heard this morning. There was a lapsed customer, Susan — a high-value customer and a champion. She had been on six trips already with Amp Adventures. She was excited again after seeing an ad on Pinterest or Instagram, went through and signed up to go on an adventure to Hawaii with her husband, and then bailed out at the last minute.

JOHN GRUNDY: That was the end of act one. Act two was about how they converted her. They sent her a text message with an offer in it. When she went back to the site, all of her information was there, and she was easily able to answer some questions via a bot and then purchase the trip. Jess was orchestrating all of this and tracking what was going on with the campaign to make sure it was going well. As part of that tracking and reporting, she built out reports and presentations in Co-pilot with Word and PowerPoint. That was the baseline story. Now I want to talk about how those things relate back to the products that Amperity is offering.

JOHN GRUNDY: As you all know, Amperity made its name on Stitch — 45 patents all in one area around how you create IDs. We are hands down the best company on the planet at this, and our patents prove that. On top of that, we now have a new AI product called Recommended Actions, which you saw in action this morning. Basically it goes into the dataset, looks for opportunities, and surfaces up an analysis. That analysis is based on 10 years of history from our analytics team going into our customer base, doing analytics on their data, and bringing insights back. If you're an Amperity customer, you may have had a visit from one of our CSMs or even our analytics team with some insights.

JOHN GRUNDY: That same knowledge and understanding from the last 10 years is what we've baked into the product called Recommended Actions. In the scenario we had, it surfaced an analytics finding: our champions — 250,000 of them — were churning. And Jess decided to do something about it. The Recommended Actions product has what they call a playbook: if you want to do something about the things we've found, here are some ideas. One of those was to run a campaign for these high-value people. She clicks on that, reviews a bunch of insights, and decides to run the campaign. What's happening in the background when she clicks "push this into a journey" is that the Recommended Actions product is actually pushing a prompt to our other AI product that we're introducing today.

JOHN GRUNDY: That product is the Customer Data Assistant, which has been out there for a little while. That AI uses the prompt to go build the journey. So now we're into the third product: Customer Data Assistant. It's an intelligence system that you can ask questions of, and that you can also build journeys with. The Customer Data Assistant uses our journey tool to build it out, and then Jess went in, looked at it, maybe made some tweaks, and launched the campaign. The campaign is what triggered the ad that Susan saw — that's act one. Susan saw the ad and clicked through. The last piece is site personalization. Susan went to the website, scrolled through, found the trip to Hawaii, and that whole experience — as Kate and Gregor were explaining — builds and builds the profile further as she clicks through the site. That's our site personalization product at work. Questions so far? Okay, thumbs up.

JOHN GRUNDY: Second step — the conversion. There's a lot going on in the background here. The piece I'm talking about now is predictive models. In this particular case, what we know from history and the predictive model is that Susan likes text messages — or responds best to them. Jack likes email. Since Susan was the one on the site, they wrote to Susan with a text message and not an email. You're using the predictive capabilities in conjunction with the journey to drive the process. There's a trigger because she puts things in her cart, there's a 15-minute wait, and then they send out the text message: "Hey, don't forget about us, here's an offer." She goes back to the site using site personalization, and all of her information is there — her trip, the dates she wanted, her loyalty points. In this particular case, the offer might include three times points if you sign up today, because we know Susan loves points. Then she talked to the bot — Fritz — and Fritz knew exactly where she was in her funnel because the site is tracking what's going on with our help.

JOHN GRUNDY: Fritz can talk to her and say, "Hey, I see you're looking at this. It seems like you're stuck — what questions can I answer to get you over the hump?" She asks a couple of questions, and then Fritz can actually book the trip for her. When the purchase happens, that triggers a confirmation — Jack gets an email with the confirmation information, Susan gets a text. And as soon as that happens, it triggers suppression, because you don't want to keep spending money advertising to someone who already bought your product. So there's a double trigger there, and that's part of the journey.

JOHN GRUNDY: Okay, last one — what's happening in the background. Jess is using Customer Data Assistant to check in whenever she wants: "How's my campaign going? What's the uplift? Am I getting to the right place?" And then she gets a call on Friday afternoon and needs to present on Monday morning. So she uses a third-party LLM — in this case, Microsoft Co-pilot — to create a document and a presentation for her monthly business review. All of that is using our MCP server. That's probably one of the most exciting things we announced today in my mind, because it opens up the ability for everyone in the company — even if they don't know what a CDP is — to use all the knowledge that exists inside of Amperity. And in the monthly business review, the IT person sitting in the back got worried about their amps usage. "Am I using too many amps? Is this campaign hurting me? What's happening?" So you can use an AI tool called AMP Insights to drill into your amps usage and understand what's happening in the moment.

JOHN GRUNDY: That's the end of the connection piece. What's important now are the product names and how they relate to what we've been discussing, because now we're going to talk about each of the products in more depth. I understand you guys just had a real-time presentation that went pretty well. Alright — development themes. There are five. First: agentic marketing. You heard all about this morning — right message, right moment. This is really about real time. How do we make the right message at the right moment happen? Then best of suite: this is really about a closed-loop system so that the system learns from itself. Then self-service: this could mean using our AI because it makes the system easier to use, but it could also be improvements to our own UI — we're going to put in-product guidance in there, for example. Use the AI to help you use our tools more easily and simply. And the final one: works where your data lives. This is about our efforts to integrate with our data lakehouse partners — Databricks, Snowflake, and others — and the connectors we have out into your dataset. This is how we're thinking about the world inside of Amperity, and how we bucket our efforts. There's money and engineering effort in each of these buckets.

JOHN GRUNDY: Let's talk about the first one or two, and then I'll hand off to my colleague. Pillar one: agentic marketing. The first product, which you heard about this morning, is Recommended Actions. Up here you can see segments of the customer base that have been defined. What you get in a snapshot is: what's happening with my segments? There's a line that goes up or down, and some numbers showing how it changed, so you get in one view exactly what happened — this is how my customer base is changing today. Then this analysis tab takes 10 years of our analytics team's knowledge and puts it into an AI system, so it can bring you back information. In this particular case, it's surfacing the lost champions issue that Susan was part of.

JOHN GRUNDY: You click the playbook, which says: here are some ideas about how you might be able to improve your growth or profitability, or both. The top one here is all about the lost champions. She clicks on that and gets a view showing some insights about who these people are, why we got into this situation, and the financial impact if you ran this campaign. She looks at it, says "okay, I'm going to make a lot of money back," and hits yes. That flips you over into the Customer Data Assistant. The Customer Data Assistant gives you the ability to query — you can ask it anything about your customer data and it will answer. But it also has the ability to do things like build journeys. As I mentioned before, the Recommended Actions product moves a prompt from itself to this product, and then the Customer Data Assistant builds a journey based on that prompt. So this is one agent working with another agent to build your journey. Clear? Okay, people are nodding.

JOHN GRUNDY: So that moves into the journey, and then she can look at the journey and decide whether she's going to approve it, tweak it, or whatever. This is our normal journey format here. On the left-hand side is the Customer Data Assistant — if you don't like what it built, you can tell it to change it, or you can go in and change it yourself if you want. You can do it all through the dialogue or through actual drag and drop.

JOHN GRUNDY: Like I said, for me one of the most exciting parts of today's announcements is the MCP server. For anybody that's not close to the server world: this is basically a go-between. You have all this knowledge in your CDP, and then you have these super powerful LLMs out there, and you really wish those LLMs could use all the knowledge you've learned about your customers. Now we have this box in between them that allows them to talk to each other. That's what MCP does — it's a translation layer that enables all the knowledge you have to be used in other places. Those other places could be anywhere. We saw the example today with Microsoft Co-pilot, but it could be any of the other LLMs that your companies are using today, accessed through this open protocol called the MCP server. Questions about that?

AUDIENCE MEMBER: What is the impact on your amps usage?

JOHN GRUNDY: You're going to use amps, because it's talking to your CDP. But it's not massive, because it's mostly using calculations you've already done. Most of the lift in Amperity — I think 20, 30, maybe even 40% of it — is the stitch: the actual creation of the IDs. And those are already created.

AUDIENCE MEMBER: So can you measure that separately? In a different database?

JOHN GRUNDY: That's the next slide. With AMP Insights, which is the next slide, you can now query and say: "How much of my usage is coming from this channel, or this connector, or the MCP server?" You can drill down and find out how much a given campaign or connector is driving. That's the whole idea behind AMP Insights — it gives you the power to know exactly what your system is doing and where you're using your amps. Make sense?

AUDIENCE MEMBER: We can see that today, but not specific to a campaign — we see it for segments, queries, or database regeneration.

JOHN GRUNDY: Yeah, it's evolving.

AUDIENCE MEMBER: Okay, fantastic.

AUDIENCE MEMBER: Quick question.

JOHN GRUNDY: Yep.

AUDIENCE MEMBER: So can this help give the explainability of a customer? For example, we've seen some use cases where the ID resolution does more over-clustering or under-clustering sometimes. If I ask a question — we're trying to build that in-house — can Customer Data Assistant explain why a fragment belongs to a particular profile?

JOHN GRUNDY: Like, why it's clustering one way or another? Yeah. I don't know the answer to that actually. Bryce, do you?

BRYCE COVERT: There are some tools in the MCP associated with Stitch where you can actually ask the question: "Can you explain to me how this clustering was done?" And it will go ahead and explain that. So to a degree, the answer is yes.

JOHN GRUNDY: Yeah. One of the more interesting things about MCP is that it has all the documentation about how the platform works, and it has access to all the data. So if you say "give me this data" and then you don't understand it, you can go back and say "can you explain this to me?" and it will use the documentation plus your data to show you why things are the way they are. Make sense? Okay. Anybody else? Alright.

JOHN GRUNDY: Predictive models. This was the text messaging scenario with Susan. We have a number of predictive models in place with more on the way. This is just one part of our AI strategy. If you think about what an LLM needs to do a really good job, it needs super solid data — and where do we get super solid data? Amperity. And then you need some ML or predictive capability: what is this person going to do next, what's their predicted lifetime value, how important are they to me? Those calculated fields are all happening with these predictive models. If you have those two things, you can do a lot with your customer base, because you know who they are and you probably know what's going to happen with them.

AUDIENCE MEMBER: Will these predictive models be on our enterprise data platform?

JOHN GRUNDY: The ones in green are already there, so you can turn them on today. There is also a resurfacing of them — we've moved them around inside the tool so they're easier to find. Those have been there for a little while and we'd love to show you how to set them up. The other ones are coming in a couple of months. Any other questions?

AUDIENCE MEMBER: This morning I saw a feature — "Bring Your Own Story."

JOHN GRUNDY: It's coming, yes. That's in the fifth pillar, about the lakehouse and how we integrate with it and connectors and so on. We're still in the first bucket, which is the AI part. All these things I've been talking about are different parts of our AI strategy — and how different users, from the IT team to the marketing team to everybody else in the company, can all leverage AI to make use of their Amperity data.

BRYCE COVERT: What about Bring Your Own Model?

JOHN GRUNDY: Bring Your Own Model is part of the presentation when we get to the lakehouse section.

BRYCE COVERT: Just on predictive models — that's something we're absolutely aiming to enable you guys on.

JOHN GRUNDY: Coming to a theater near you. Good question. Alright — AMP Insights. Someone was asking about this earlier — Kim, right? AMP Insights is a different analytics tool, AI-based, that is purely natural language. In this particular case, I'm the IT director, worried about my amps usage — "What's been my average usage for the last month?" And it says: 1,295, that's how many amps you've been using every day on average, and here's a breakdown of how it's split out. We have that today. But now you can ask the next level down — I don't have a slide for that, but you can keep drilling. You can say, "Show me a breakdown by connector of how my amps consumption is happening." If you want to see it in action, go to the demo station.

BRYCE COVERT: You can compare how many amps were used for running this campaign versus that one. You can literally ask all kinds of questions about the performance of the platform in natural language.

JOHN GRUNDY: Yep, it's awesome. We use it internally too. In fact, we started using it internally and realized there's so much value we can expose. For example: why did this customer spike yesterday? Oh, it's because they did this thing — okay, let's look into that. That's how it started.

AUDIENCE MEMBER: Can it help with estimates as well?

JOHN GRUNDY: It'll give you historic numbers — it doesn't have a forecasting capability, but you can say "show me my amps usage by quarter for the last year" and it'll show you the pattern. You can then make your own guess based on seasonality. It won't forecast for you, but it'll give you the data to help you forecast. Good. Any other questions about AI before we move on to real time?

JOHN GRUNDY: Okay. I know you guys just had a whole session on this — I wasn't able to join that one, I was busy talking to somebody. TK was running it. Alright. I wanted to explain a little bit about the technology side of real time. The core of it is what I've been calling a real-time engine. You imagine you have this huge investment you've made in your CDP — it does this amazing stitch process, and now you know your customers really, really well and you have this gold mine of information — but it's from yesterday. The way the stitch process works at most of our customers is it runs overnight, and then the marketing team comes in in the morning and starts using that data. But as soon as the day starts, it's outdated, because your customers are still buying things and doing things.

JOHN GRUNDY: Wouldn't it be nice if that dataset kept getting updated? That's what the real-time engine does. It takes the customer 360 that you've built and spent a lot of money on as the starting point for the day. Then during the day, streaming events come in from wherever — purchases in-store, wherever you're streaming from — and the engine makes decisions based on those events. If something happens that changes a loyalty status or segment, the engine makes those changes and you can take action. The action could be personalizing your website, or triggering suppression, email, text messages, or whatever. This is the combination of the big batch lift that everybody knows and loves, plus updates during the day that let you take action based on whatever just happened. It's today's intent combined with, in many of our customers' cases, 10 years of history. The best of both worlds. Is that clear?

JOHN GRUNDY: That's the concept. The products in this space are abandoned journey — our trigger activation product. If this happens, do this. Somebody purchases today and changes loyalty tiers — send them a welcome to their new tier. Or they bought something and you want to send a thank-you note. That's what trigger activation does. This is close to the journey they mentioned this morning: Susan put something in the cart, they waited 15 minutes, she didn't do anything, so they sent her a text message. You can see it here: something happened in the cart, wait for 15 minutes, talk to her again, wait for a day, and if she doesn't do anything there are branching choices below. That's the scenario Susan went through, set up by Jess. I think you went through this in the last session, so I won't spend a lot of time on it.

JOHN GRUNDY: Site personalization is the other one. This debugger we built shows how the product actually works. What you're seeing from this morning's presentation is that the information in the profile in your real-time engine is being transmitted to the website, and the website is using that information to personalize not only what's on the page, but also the special offer. In this particular case, Susan was booking economy, but we know she's a luxury traveler. So they offered her an upgrade to luxury travel, even though she was looking at economy. That's the kind of intimate knowledge of what a customer actually likes versus what they're currently looking at — it enables you to personalize past your regular product portfolio into specialty products that usually have a much higher margin.

AUDIENCE MEMBER: I'm just curious — this experience is a modal or banner pop-up type of experience. Is there a concept of in-page personalization as well?

JOHN GRUNDY: Oh yes, that was happening on top of that. When she came back to the site, clicking through from the text message, it was already pre-populated with the trip she wanted, the dates she wanted, already in the cart. All she had to do was click yes. That's because you can personalize based on the information you stored about what she'd already done — she wants to go to Hawaii, she's a luxury traveler, and so the page repopulates. And in addition to that, on top of that, you can also do these special upgrade offers. So you're not only capturing the business, but you can also do the upsell in the same motion — less friction, more movement.

AUDIENCE MEMBER: So the real-time profile and your stitch profile are both bringing data through an API call to the website?

JOHN GRUNDY: Right. Big picture: you have your customer 360, you've stitched your profiles and added all your attributes. When that finishes running, you dump a copy into the real-time engine, and as the day goes on you're adding to that profile. The website can then pull from that and personalize.

AUDIENCE MEMBER: Is that an API call from the website to the real-time engine?

JOHN GRUNDY: Right.

AUDIENCE MEMBER: Is there any latency from those calls?

JOHN GRUNDY: It's milliseconds, so no. And as you'll learn in one of our other slides, we just opened a data center in Australia, because real-time latency is a problem when you go across the ocean. We need to be in your region of the world, so we have data centers all over the world to service your customers wherever they are.

AUDIENCE MEMBER: Thank you.

JOHN GRUNDY: Other questions about real time before I go on? Alright — multi-stitch. This is the last slide in this section. Multi-stitch is the ability to have two graphs based on a single data source. This has been requested and used by our customer base for paid media versus things like loyalty, where exactness matters. For paid media, you want the biggest reach possible — you want to talk to the most customers because you're trying to drive the top of your funnel. We have cases where at some customers this number is like 300,000 to 400,000 people. Then the other graph is like 120,000 people because it's now deterministic and very tight. Your different marketing teams have different needs. I actually just finished a conversation about this exact thing before this session. The problem they brought to us was: "I've got this probabilistic graph that's very tight and we're very happy with it for this particular set of users — but I have another set that wants a totally different graph." Well, then you need to go back and talk about what ID means for each of those user sets, and you can use this capability to give each of them what they want. One is optimized for reach, one is optimized for precision. Okay — now I'm going to bring up my colleague to tell us about pillar three.

SPEAKER 6: Thank you John. Hello — can you hear me okay? I'm really excited to talk about our best of suite pillar and our self-service pillar. Best of suite refers to a full range of capabilities for the modern marketing team. One of the key capabilities I want to talk about today is measurement. Measurement closes the gap between the spend and what you can prove. Better data means better AI, better campaigns, and better optimization by our marketing teams.

SPEAKER 6: The first product is journey goals, which we launched in February. That now allows you to see which journey paths drive business outcomes, and you can optimize while the journey is still in flight — you no longer have to wait until the campaign is done. The next piece is closed-loop measurement. We added data pipelines for Google and Meta to bring data back into the platform automatically. We started with Google and Meta, and we're going to expand to more paid channels, and then add owned channels as well. Campaign optimization is something you should really lean into — this one sends offline conversion data back into your downstream ad platform so those platforms can optimize, and it really helps your return on ad spend. Those three capabilities are in the platform right now.

SPEAKER 6: And the next one we talked about earlier today — audience monetization. Your first-party data is one of your biggest assets. We now allow you to monetize first-party data in a globally compliant, completely privacy-safe manner without ever sharing any PII. While we've always supported in-house retail media networks, we now support a first-class integration into third-party marketplaces. The first is The Trade Desk. With that solution, you can now create custom audiences for one-to-one brand partnerships and monetize with a specific brand partner, or create syndicated audiences where you monetize in a public marketplace and really scale your monetization strategy.

SPEAKER 6: The next pillar is self-service — and I'm really passionate about this. To support every user in the platform, whether through AI or better UI, we really want to address self-service capabilities. A year and a half ago, we launched Quick Start. Some of you may have onboarded with it — it allows you to add data, resolve identities, check results, and generate customer profiles in day one. AMP Insights, which John talked about earlier, lets you go in and see your real AI consumption — that's a highly requested feature, and it's new as of our most recent release.

SPEAKER 6: On measurement optimization — I want to come back to that for a second. We are not just bringing data back in, we are going to make it really easy. We'll allow you to bring data back in with one click, and we're going to drastically improve the way you can set up conversions to send offline conversions back into the downstream platform. This is coming very soon.

SPEAKER 6: And lastly, I'm really excited to announce that we are launching Amperity Home — a new launchpad. When you log in tomorrow, you will see a new landing experience. For the first time, you can see what connectors are being added in real time, what capabilities are being added, and you get notified of releases right when they happen. You can align your organization around KPIs to make sure everyone is driving toward the same goals. AI is front and center — use it to ask questions about the platform, how to use features, explore, create segments, create journeys, right there. And tenant health: one of our most requested features. Every customer wants to know when something is wrong. At a glance, within five seconds, you'll see a tenant health score. And the getting started section shows you all available features grouped by foundational, building value, and mastering — showing you which features you're not using and what you should be exploring. If you mouse over, there's a help icon and you can use the AI to explore what a feature can do for you. It helps you get to value faster and helps new users get onboarded quicker. That's it for those two pillars. Back to John.

JOHN GRUNDY: Thanks. Alright, we've got one pillar left. Our friends in the back are already signaling that we're running out of time, so we might go a couple of minutes over and I'll apologize for that in advance. If you need to leave, it's okay. Works Where Your Data Lives — this is one people were asking about early on. This is really about our data lakehouse integration strategy and our connector strategy: how we connect with your data and make it as easy as possible to work with your existing data. First: data lakehouse. Bring Your Own Compute. You may remember that we launched Bring Your Own Storage for Databricks and Snowflake last fall. So if you want to use your storage capacity in Databricks or Snowflake, you can do that today with Amperity. Getting to full lakehouse integration takes time and lots of integration work. We started with storage, and now we're announcing compute. You can do all your stitch compute, all your queries, everything — using the compute from your existing data lakehouse. To show you that, we have a video.

SPEAKER [VIDEO]: Welcome to Amplify 2026. Amperity Bridge gave you the pipes for zero-copy data sharing. But true data sovereignty requires full residency. Many of you already rely on Bring Your Own Storage to keep files secure in your own cloud. Today we are thrilled to announce the missing half of that vision: Bring Your Own Compute, supporting both Databricks and Snowflake. Launching globally on June 1st, 2026, BYOC perfectly complements BYOS. Together they shift Amperity from a third-party vendor into a core internal enterprise capability — operating exactly like a tool your own team built, powered by our world-class AI.

SPEAKER [VIDEO]: Here on the Quick Start dashboard, the process begins with custom domain tables. As a table is authored to transform incoming data, the underlying SQL query executes transparently against your own Databricks serverless warehouse. This is pure zero-copy architecture. Amperity comes to your lakehouse; your data never leaves your cloud. You manage the encryption keys, network rules, and access logs, ensuring absolute data sovereignty without a separate vendor security audit. As custom queries are authored in Amperity, their execution is fully traceable directly within your Databricks query history.

SPEAKER [VIDEO]: Running these workloads inside your own account unlocks massive compute efficiency. You can utilize your pre-purchased cloud credits or reserved instances to power the processing, entirely eliminating brittle ETL pipelines and treating your lakehouse as the single source of truth. In the past, segmenting high-value customers meant relying on external data movement. Now, as a segment loads in Amperity, multiple concurrent queries are instantly dispatched to your Databricks SQL warehouse. This delivers intelligence at the source. You get immediate access to Amperity-derived insights, owning the compute resources that handle massive burst processing seamlessly. When identity resolution is required, a stitch job spins up a dedicated on-demand Spark cluster directly within Databricks.

SPEAKER [VIDEO]: Exploring the Databricks Spark UI reveals complete metadata transparency and auditability. Your data engineers have full visibility into CPU utilization, container memory, and JVM heap usage. They can see exactly how every customer record was transformed, keeping you audit-ready using your own cloud logs. Finally, the customer 360 database generation job coordinates parallel processes natively on your serverless warehouse. The cluster automatically scales as needed, building out comprehensive data models that instantly populate back into the Amperity application. Complete control over your data. Zero movement, unmatched efficiency. Bring Your Own Storage gave you control over your files. Now Bring Your Own Compute for Databricks and Snowflake completes the picture. Launching June 1st, 2026. Get ready to secure your ultimate data sovereignty.

JOHN GRUNDY: Alright. So that was Bring Your Own Compute. The message here is: this is a journey, and this is the next step. There's a lot of noise in the market about composable. If there's any question about Amperity's commitment to composability, this hopefully answers it — you can do it now with Databricks and Snowflake. There are announcements coming at the Databricks event in June and the Snowflake event to launch both of these products with the partners that helped us build them. This is a preview for you because you're our most important customers and you showed up today. The message is: composable is a big topic, especially in the IT world, and we are fully on board with that trend. Any questions or comments?

AUDIENCE MEMBER: I didn't hear — is BigQuery coming in the future?

JOHN GRUNDY: BigQuery is something we're working on. There are some technical challenges we're trying to work through with Google right now around package size and how we move data back and forth. We're still working on that one. Azure Fabric Bridge is also on our roadmap. We started with Databricks and Snowflake and we'll work through the others.

AUDIENCE MEMBER: One more question. Currently we're reverse-syncing into our data lake. Once this is enabled, do we no longer need to reverse the output?

JOHN GRUNDY: Correct — because it's still in your warehouse. It didn't move. That's the whole point. The storage is there and the compute is there. Other questions in the back?

AUDIENCE MEMBER: How does the zero-copy architecture work with the real-time base?

JOHN GRUNDY: Good question. The real-time architecture is not on Databricks — it's a separate architecture today. I talked about how you build the stitch process, then the customer 360, and then we move that into the real-time engine. That movement still has to happen, because in general the storage across the network isn't fast enough to deal with real-time. We're talking a couple hundred milliseconds to respond, and you can't make calls across the network and keep that timing up.

AUDIENCE MEMBER: That means it's not possible to use real-time with Bring Your Own Compute?

JOHN GRUNDY: It is possible, but the real-time won't be on the storage. It will be somewhere else.

AUDIENCE MEMBER: So the data is on our storage, and you make a copy of it for real time?

JOHN GRUNDY: In real time, there will be a copy, yes. This is for the core stitch product. The real-time product is newly built, and we want to make sure it works and you're all happy with it before we start putting it on the data lake. Crawl, walk, run. Other questions about that? Good questions. Alright.

JOHN GRUNDY: I think we've got two or three slides left. Data center — we talked about this one. Amperity runs a global business, and we have a big business in Australia. Because of this topic and the importance of our customers there, we launched a new data center in Sydney. This will help with local requirements but also with our latency issues and making sure the real-time experience in that part of the world is also great. And connectors — you're probably all familiar with our connectors. We have over a hundred. We're building more every day, and if you need more, let us know. We have a member of GRT's team whose sole job is to get connectors built.

AUDIENCE MEMBER: Are these real-time connectors?

JOHN GRUNDY: Some of them are and some aren't. If you want to jump on the real-time train, please come see us right away — we'd love to build some real-time connectors for you. We have a bunch of them but not all of them yet, since it just launched.

AUDIENCE MEMBER: I just want to confirm — we have Adobe connecting the web?

JOHN GRUNDY: Adobe is one of the ones we support today, yes, in real time.

JOHN GRUNDY: Alright — the road ahead. Today we talked about a lot of products. We've been here for almost an hour, and all these products are in various stages of being rolled out. This includes the Customer Data Assistant, the Recommended Actions product, the predictive products — which are already out there — the Amperity MCP Server (please let us know if you want to jump on that, we're looking for more people), AMP Insights — which should be live today or tomorrow if it isn't already — and the homepage, which I think is live right now. The new real-time products — abandoned journey and site personalization — are available if you want to work with us on those. Multi-stitch is already out. Audience monetization and paid media are also there. The Amperity homepage and the new data center are live.

JOHN GRUNDY: Next up, as you can imagine: more AI. There's a bunch of predictive work in progress. The AMP Agent that you heard about this morning is going to take a little while to fully build out, as Gregory was explaining, but we're looking for customers to help us define and build it out further. A number of questions have come up about our continued data lakehouse strategy — you see Databricks and Snowflake here, Azure Fabric is coming, we're going to figure out BigQuery, and Bring Your Own Model is also in the works. That's where we are today.

JOHN GRUNDY: If there's something you're interested in, my goal for this session was to get you excited about some of the stuff we've built. If you're excited, go talk to your account team, or to myself, Greggo, my colleague, or anyone here. If you want to get on the early access list to try some of these products, please let us know. If you want to help further as design partners — we usually partner with two or three customers to build out these products, because we don't want to do it in a dark hole, we want to do it with real people and real problems — we're always looking for partners. If something is interesting to you and you want to help us build the bus, we'd love to hear from you. And with that, we'll take questions. Oh — I'm also supposed to tell you to give us feedback on the session. Raul gets bonuses based on how many people submit session feedback, so do him a solid and help him out. Other questions? We had a lot of good questions today.

AUDIENCE MEMBER: Can you do a one-liner or short description of each of the products?

JOHN GRUNDY: Yeah, we can figure out how to do that. We have them, we just need to figure out where to put them. Had a question here.

AUDIENCE MEMBER: I was asking this question to people from this morning's session too, but I just want to understand — there was mention of real-time recognition launching in June or something. Is that related to some of these products?

JOHN GRUNDY: Yeah, the real-time engine is what does that translation. That's what we've been talking about today.

AUDIENCE MEMBER: I mean real recognition — like real-time stitch happening with data attributes.

JOHN GRUNDY: There's no real-time stitch coming. Stitch is a separate process in the batch world. There's a whole separate stack that does real time. I talked about how we move the customer 360 from the batch world to the real-time world, and then we interact with the streaming data coming in. That same data also goes through the batch world and overnight gets included in the batch side — that's how they stay synced over time.

AUDIENCE MEMBER: So if I as a customer create something new, does it create an Amperity ID on the fly with the real-time engine?

JOHN GRUNDY: Yes. The real-time engine makes Amperity IDs.

AUDIENCE MEMBER: And that's already launched?

JOHN GRUNDY: Pretty much — we have a couple of customers on it right now.

AUDIENCE MEMBER: On site personalization — I understand you're tying it back to the Amperity ID. When someone goes to the site, are there any conflicts with privacy settings? Do they need to be logged in?

JOHN GRUNDY: No. The question is really about what tag is coming from the site. With the customers that are testing or working with us on this right now, we're using the Adobe Experience Manager tag. We get a series of events on that tag, and once they log in, we know that this tag is associated with this ID — and then we merge that together. So you have a single profile on the real-time side.

AUDIENCE MEMBER: What if they don't log in?

JOHN GRUNDY: Then it creates an Amperity ID. As soon as you get a second event, it creates an AMP ID, and that ID exists for that session. If you come back tomorrow and log in, we can still associate the activity from yesterday with the login. We keep all that information per session — or whatever increment the tool you're using on the site side defines — and it has an ID on it.

AUDIENCE MEMBER: I have a follow-up on that. Say for example I come to the site and you create an anonymous Amperity ID. Tomorrow I become a known customer. What is the final AMP ID — does it change?

JOHN GRUNDY: No, because you logged in and it already has an ID. Bryce, do you want to take the technical details on this?

BRYCE COVERT: So the way it works is: if we can't figure out who you are, you'll get an Amperity ID from your anonymous session. Then the batch process that does Stitch overnight evaluates, for every data point we have, what's the best cluster and the best Amperity ID for it to be associated with. It might be that we think it's best connected with one graph one day, and then we get more information that says that data is better associated with a different graph. So that actually happens every day as part of the batch process. And the view of the profile is unified — so if you had some product categories you were interested in from your anonymous session, and some from your last purchase, we can use both of those data points in the attributes. They would be merged as part of the Stitch process.

AUDIENCE MEMBER: So it's merged in the batch process, not the real-time one?

BRYCE COVERT: Yeah. That technology is called Stable ID — it's actually one of our first patents. It uses the best Amperity ID, which is typically the longest-living one.

AUDIENCE MEMBER: And that matching — is that going to be available by June 1st?

BRYCE COVERT: Yes, I think June 1st is the date.

AUDIENCE MEMBER: Does that answer your question?

AUDIENCE MEMBER: Yeah, yeah. Okay, good.

AUDIENCE MEMBER: Do you have an issue with just collecting a lot of anonymous records from this technology?

BRYCE COVERT: No, we don't.

AUDIENCE MEMBER: It's not an issue?

BRYCE COVERT: Not in terms of storage capacity, no.

AUDIENCE MEMBER: Or just having this huge group of people that you can't activate on because you don't know anything about them?

BRYCE COVERT: Yeah, but if they show up again, sometimes you can use that information.

JOHN GRUNDY: Right — you can use that information to help personalize. Even if you're an unknown user, if you come back, we may be able to recognize you and start to personalize, even without knowing who you are. We can't text message you or email you or do anything like that — but we can help you on the site. We don't know what we don't know, but we can't create customer contact information out of thin air. Great set of questions. Anybody else? One in the back.

AUDIENCE MEMBER: You can't contact them, but you have the capability to create a segment. Can we not create a segment out of these real-time captures?

JOHN GRUNDY: You could. You could say: all the anonymous people that love adventure travel — make a segment out of that. And it'll put them all in a group. So yes, you could define a real-time segment that way.

AUDIENCE MEMBER: You can personalize based off that.

JOHN GRUNDY: Yeah. And you could then only show them adventure travel content if that's what was their bucket. So you can do that even with unknown people. It works more or less like with known people — you just can't do anything other than personalize the site until they tell you something else about themselves or log in.

AUDIENCE MEMBER: Currently we have a challenge to understand explainability of the profile — understanding why a particular fragment belongs to a certain profile and the strength of each connection in the graph.

JOHN GRUNDY: Ask the Customer Data Assistant — it will tell you how customer profiles are built.

AUDIENCE MEMBER: So it can show the graph and the strength of each connection?

JOHN GRUNDY: Right. There's a part of the tool where you can go to that customer and it will show you the six clusters that we think are all part of that person. It shows you the strength of each connection — this is very strong, we're very sure about this, we're fairly sure about this — and there's a graph in the tool that will show you that.

AUDIENCE MEMBER: Have you seen the stitch benchmarking at all? It's one of the capabilities to identify common issues where things are over-clustered or under-clustered. I believe that went GA about a year ago or six months ago. You're able to proactively dig into some of the problems you might run into — like why are these two things connected in the first place — and optimize to get your stitch results better and better by getting closer to baseline.

RAHUL SUNKAVALLI: All right, let's wrap up. I know some of you want to take your break. You have 15 minutes before the last session — please make sure you go back to where we had the keynote for the closing session. Paul from the NFL as well as Bridgette, our CMO, will be up there taking us to the finish line. Thank you.