0:00All right, welcome everybody. Great to have you all here on our MotherDuck webinar for today. We are talking all about how folks in sales and marketing are using AI in the real world and how they're using it to do data analysis to help turbocharge the job that they're working on. Applies to a lot of different folks, but I think sales and marketing, there's always an element of sales and always an element of marketing in any role.
0:24And I think this will be an interesting way to think about how to apply to things you're working on as well. So today we have a panel with us. We've got Ryan Boyd, co-founder here at MotherDuck, and we've got Logan Toski, one of our account executives. He's a team lead for our account executives. Wanted to spend a little bit of time talking about their career path, the things that they use AI for, so we kind of hear a little bit about how AI affects their roles.
0:49And then we're going to hear a bit about how MotherDuck intersects with that. A couple of different technologies that MotherDuck has that make that really easy and straightforward to where you don't have to be a technical person to accomplish really technical stuff. That's the magic of living in 2026, where you can get a lot of magical stuff done. So we'll also do this interactively, where we're going to use some of those features
1:11here on the call, and we're going to see some demos. So the second half of the call is really all about demos. You'll see what Ryan has built, you'll see what Logan's built, and we're going to build something together as well. During this whole time, it's a lot more fun if there's Q&A. So please don't hesitate to put your questions in the chat. We're going to answer questions all along the way. So no questions too big or too small, we'll do our best, and we'll definitely have designated time
1:35for Q&A at the end also. So put those in the chat as you think of that. So I wanted to first say welcome. So Ryan, could you introduce yourself a bit and tell us a bit about your career arc? How did you end up where you are? Sure, thanks, Alex. So yeah, Ryan Boyd, co-founder at MotherDuck, but I am not your traditional co-founder. I'm not leading engineering.
2:00I'm not leading product or the CEO. I lead marketing and developer relations. So my background actually comes from the developer relations space and really wanting to see marketers have greater empathy for the technical users of their products.
2:22So I probably have a little bit more engineering background than the average marketer, but I do really think that a lot of the stuff we're talking about here today can be built and used by the average marketer. And there are some folks on my team who have built stuff that I'm going to be showing
2:44that do not have an engineering background. So I'm excited to show that so you can see, it is useful business applications that we can build on top of this data and useful drill down into questions. And you can do that whether you have the engineering background or not. Fantastic, well, excited for that. Well, glad to have you for sure.
3:08And Logan, you want to say hello to everybody as well? Yeah, hello. Super excited to be here. Been at MotherDuck for almost a year now. I'm working on the accounts team. One of the first few reps here working with hundreds of our customers and now acting as a team lead here out of the Seattle office. And I've been in kind of tech sales since 2013. So quite a while and worked in kind of all areas
3:31from HR tech to MarTech, been at places, a bunch of startups as well as larger companies like HubSpot. And that's actually brought me to MotherDuck because I worked with a bunch of people and met a couple of those places that came here. We're like, wow, MotherDuck's doing some pretty great stuff. You should come over and build with us. So it's been a super exciting journey just a year in and yeah, for me, I'm excited to dive in
3:53because a lot of the kind of AI journey I've had has been accelerated, especially in the last six months. Everything has accelerated in AI in the last six months. So you're absolutely a good company there. Same for me as well. So yeah, my journey has accelerated quite a bit in that time. Well, fantastic. Well, why don't you ask us a couple other questions around that kind of AI side of things in your role a bit.
4:15So Logan, how has AI kind of started to enter into your role? What made you decide to jump in and kind of dig in really deep to the AI side? What was kind of the start of that journey? Yeah, I would say like outside of work, I'm not very techie actually. Like I, you know, spend a lot of time in nature. I like don't spend a lot of my mental energy in these places, but a couple of years ago
4:39when ChatGPT came out and people were interested, you know, I use it for all the things other people did. I made silly like birthday cards and tried to see if people thought that I wrote them or not, you know, that type of thing. But I didn't bring it into the work so much until I came here. And the first thing I started to do was using Claude to help save memories around formatting of all follow-up notes or just how I wanted things
5:03structured for kind of HubSpot notes. And it really started to act as like a really solid kind of assistant for me and made me a lot more efficient and consistent in how I was handling these things. It's also a really technical here, a lot more technical than other places that I've been. And so it was a really helpful at like, how do I ramp quickly into understanding kind of, you know, the data personas that we work with and what it's like to be at MotherDuck versus some other places.
5:27So that's mostly what I was using it for until about December when the models improved to the point where we felt it was necessary to relaunch remote MCP, which allowed me to start talking to the data directly. And then in the last few months with dives to start building kind of these lightweight, you know, self-serve analytics. So now it's, you know, it went from being a helpful assistant to kind of like my day in, day out,
5:51how I analyze my book. And, you know, I'll demo some of that in a bit. That's awesome. Yeah, I think there's definitely sort of the, there was the fun aspect of it for a while of just like, let's just tinker with it and kind of see what the heck this thing is. There's the, you know, kind of, let's use it to organize some things, kind of the basic office style, but now you're doing data analysis with it. And I think that's another level. The stuff you built is really awesome.
6:14I've been seeing the sneak peeks of it already. So psyched for folks to see that before too long. So, Ryan, I'm curious how you kind of have seen AI come into your journey and what was kind of the first kind of couple of things you started with on the AI side? So I would say I was somewhat cynical on the AI front for a while. You know, when we, many years ago, you know,
6:37I worked at Google, led developer relations for Google Cloud. And, you know, one of the products that we released was sort of an AI prediction service. And I was, you know, I was very frustrated by it because it got many things wrong. And fast forward, you know, a number of years. And last November, I actually gave a talk
7:00where I was super cynical on the ability of AI to do data analysis. I was using AI to work with content, to, you know, help me write and edit. And it was okay. But when it came to data analysis, it wasn't working very well. And I was very cynical in this talk.
7:22Three weeks later, you know, a new version of Claude was released, followed by new versions of all the other models. And all of a sudden, like, it was much better for data analysis. And that's where I find it, you know, most useful nowadays is sort of drilling down into data.
7:46Data that, you know, yes, it's always been there. Yes, I could always spend the time writing SQL queries in order to access it. But I never really wanted to spend that time doing that. And now it just becomes a lot easier to really kind of see what's going on with the business. And, you know, that's really what excites me.
8:10I've been in data for a long time now. And the increasing the accessibility of it has been what I've enjoyed. And not, you know, I can write SQL. And actually, Alex, seeing people like you who can write SQL really darn well, still being more effective by using some of the AI tooling is amazing to me.
8:35It's a big thing. I think a lot of it too is, yeah, the speed at which you can iterate, the speed at which you can look at things. And it means that you can do more analysis than before. So it's not necessarily just speed up what I was doing before. We can all be more curious. And also, I think I've never seen a data team that doesn't have a long list of backlogs of questions. I've never seen it. It's everywhere.
8:58And if you're on a different side of the fence, you know, we put up our little fences in companies, right? But if you're on another team and you're in sales or in marketing, it might be that your questions just aren't getting answered because you know if you submit the question, you won't get an answer for a week or longer. So it's a, yeah, go ahead. I'm just going to jump in here. Like I, you know, our data team,
9:21you know, data engineers, the X, as we call it, there's a gentleman named Nick and a gentleman named Rick. It's kind of funny. But our data team, the other day, I'm like, there's a bunch of HubSpot fields that are not in our data warehouse that I need access. And I start diving in and I start diving in
9:42and I came up with Claude with a list of 156 HubSpot fields that were not in our data warehouse. This huge ticket, right? So I filed a linear ticket, of course, using Claude to say I wanted these 156 HubSpot fields. Nick comes along. He has a single prompt that he uses to then put in the ticket ID and was like, process this ticket, handle this ticket.
10:07And, you know, I think within an hour or two, our data warehouse started getting populated with those HubSpot fields. So you would never have seen that in the past. You know, it's always been, you know, if I had filed a ticket 150 HubSpot fields in the past, like I'd be waiting months.
10:30And this literally, you know, I was able to create a prompt. He was able to create a prompt. He was able to do the code review, looked good, went to production, so. Yeah, I think it's absolutely a story of everybody getting on board. You know, everybody can do data analysis now, which is great. And I think that by and large, that democratization is huge and that's what we want to do, right? That our belief is that it's not that people
10:55are doing this analysis for the sake of the analysis. You're doing it because you want to make a better choice. And if we didn't do this analysis, which we didn't before, we'd probably just be going with our gut. And our gut's usually pretty good, but you know what? It's better if we, if our gut gets better over time, you know, we, you know, work to understand some data first a little bit. So thank you guys for sharing your background, sharing a little bit about kind of how AI has intersected.
11:20I think we'll jump deeper into how you're using it day to day today with the real demos, because that's, I think, really the fun part. I want to take just a couple minutes real quick just to do some overview of kind of what are the features that we're talking about here for MotherDuck. That way you guys can follow along, you can hop in, you can install it here today if you'd like, or at the very least, you'll know where to go right after this. So our goal at MotherDuck is to make it
11:44so that you can use your existing AI agent, you know, whichever one you like, and set it up with MotherDuck in five minutes. And I think that's even on the high side. I think you could do it in two, depending on how speedy you are at your just clicking through things. So what do you need? Well, you need a MotherDuck account. We have a seven day free trial. You can get that free trial without a credit card. You can hop in, you can register,
12:07and very easy to set up an account. And then we have our remote MCP. And that is kind of the layer that's powering a lot of the other features we'll talk about. So the remote MCP is using the model context protocol, which is just a way for agents to talk to other tools. So agents famously bad at math. Agents famously bad at, you know,
12:32and if you're bad at math, you're not gonna be great at actually analyzing data and calculating metrics and aggregates, but they're really good at using tools. And so what we do is we give the agents the tool in MotherDuck, which is an amazing analytical engine for doing those types of calculations. So as you ask questions of your data, your agent will be able to know, I have access to this tool. I can go talk to MotherDuck. And if your question sounds like something
12:57MotherDuck can answer, it will go ahead and check MotherDuck and use MotherDuck as a tool as a part of answering that for you. The best part about our remote MCP is it is incredibly easy to set up. There is no local setup whatsoever. All you need to do is tell your Claude account where your MotherDuck, that you wanna use MotherDuck and then log in.
13:21It's really that straightforward. So there's no installation, which is really what I think unlocked our team. So Logan, maybe as you heard about our remote MCP at first, what was your kind of initial experience with kind of getting started and getting it set up? Yeah, I think at that time, I think it was, I went into the Claude connectors library, found MotherDuck and added it. It was like two clicks.
13:45And I think maybe now it's even down to like one click. Like it could not have been simpler. And so that allowed me to start talking to the data and having Claude help me write SQL in maybe 20 seconds, so. Oh man, well, I'm overshooting it with five minutes. That's a great story. Well, thank you for that. You can time it as part of my demo, Alex, because I'm just gonna- Oh man, we're gonna speed run it? I love it, I love it. We're gonna do a speed run of this. Exactly.
14:10We'll see this in detail. We're an official connector in the cloud marketplace. So it's very easy to find us, especially if you're using Claude. It does work with other agents as well. We have some more detailed instructions in our docs page. So at that point, once you have this remote MCP, the first thing you can do with it is you can use it to talk to your data and you can get questions answered in natural language. And we'll see some of that practically. And that's really a great starting point because what it allows you to do
14:35is build up context for the model. So you say, hey model, hey Claude. It's a little more natural to say, hey cloud than it is to say, hey agent. But hey Claude, you know, I wanna explore this type of data set. I wanna do this, I wanna do that. And Claude will bring that into its context and it'll kind of read over that data and understand it at a high level. And then you can start to go deeper and ask more questions. And beyond that, beyond just the incredible power
14:59of being able to talk to your data in natural language, you know, this holy grail we've been after as an industry for 20 years, right? That was the first five minutes of the talk today, right? We can do that, that's great. The next thing we wanna do once we have that superpower is to then also look at our data and explore it and build interactive applications that use that data. And that feature in MotherDuck is called a dive. And all you have to do is once you've done some analysis with your agent,
15:24you ask for a MotherDuck dive. And the MCP server will teach your model, your agent how to do that. Be like, hey Claude, if the user asks for a dive, this is what you do. So what is a dive? A dive is a data visualization that you build in your AI agent chat. Any of the ones here are supported. So Claude, ChatGPT, Claude Code, cursor, a huge variety.
15:46Then not only are you just creating an analysis that's like a one-off analysis that you can do with Claude directly, this becomes a reusable, shareable, refreshable asset that your whole team can access. And it'll automatically pull in your MotherDuck data. So that's a key thing is that dives really take this from a single player activity that you do once to a multiplayer activity that's repeatable,
16:09part of your business process on a daily, weekly basis. Under the covers, what is a dive? This is not super critical to know, but basically it's a web app. You're building a webpage powered by React, and it uses SQL queries to talk to the database. But you don't have to write a single line of React, you don't have to write a single line of SQL, and you don't even have to read it. I've written both and read both a lot, and I've built up a level of trust
16:34where I don't need to do that anymore. It is crazy the world we live in, but that's the truth. And then the last piece is not only can you use this for your own work or for your own team's work as you share it, you can actually let your customers take advantage as well. And you can take a dive and embed it on any website that your company operates, and that way, you can very easily share this analysis with your customers.
16:57Typically, if you're using like an embedded BI platform, those are very pricey. This is a precious commodity, right? You want to expose your customers to their own datasets they can explore and they can learn. Now that's something where you get right out of the box with MotherDuck in a very straightforward way. What are some of the superpowers like why use dives compared to some of the other AI products out there?
17:20The key here is easy. That's our goal. We want everybody to be able to do this, not just the engineering team, not just the software developers. And we've seen that play out within our own company and with our customer base. So we've seen it. MotherDuck is also really great because we can handle kind of whatever data you throw at us. We are incredibly scalable. We can handle terabytes, terabytes, tens of terabytes, hundreds of terabytes, however much data you have,
17:44we can handle it very, very quickly, which also keeps it easy. You don't have to do any special hoop jumping. If you have a large dataset, we'll just chew right through it. We also have this superpower of dual execution. We might see in some of our demos today where you can actually have really fast interactivity where not only can you see your data, but you can explore it and manipulate it super quick. And then the MotherDuck architecture was built for humans,
18:08but actually is even better for robots. So each time you connect to MotherDuck, each person in your org can have their own isolated cluster, their own environment. So their own little sandbox to where they can do their own data analysis, which means that if I accidentally, together with Claude, write a giant query where I'm looking at huge amounts of data,
18:31doing crazy manipulations, no one else gets impacted. It is all, you're just working on your own data. You're just working in your own little environment. Last thing, if you're using Claude Code, if you wanna dive even deeper, there's a couple extra goodies there, but again, that's not necessary. We're gonna see how we can use this right in just the Claude web UI. Claude desktop works well as well, but there are some extra things
18:56if you are interested in doing it on the Claude Code side. I will say members of our marketing team have been using Claude Code. So this is not like Claude Code, that word code really means you need to write code or be a developer. It's become basically a brand name for a feature of Claude. Our product marketer, Garrett, like spends half the day talking to his Claude Code
19:19and it's become his new friend, I guess. It's another useful tool in the tool belt. So I would say, I would encourage anybody that would hesitate at all because it has the word code in it. That's another thing we can get up and running in just a couple minutes and then you'll be using it every day, no sweat. And I've seen it happen and I know you can do the same. I'm on the sales team.
19:42So I'm the most non-technical duck, if you will, as we refer to him here. So I was very hesitant to use Claude Code for a while. It was like, no, like it's got code in it. And then finally I sat down with Garrett and then I was like, oh, this is actually pretty simple. So most of what I built recently has been in Claude Code.
20:00With us releasing the Dives app viewer in the Claude chat, it has made kind of that local iteration something you can do just in the regular Claude experience as well and give something similar. So that's been really helpful. And also just say from my seat, when we released the remote MCP in December, I started building out a lot of visualizations that were artifacts, right?
20:24Like as Claude experiences them. And so I was able to use the power of MCP and MotherDuck to build, you know, these really great kind of dashboards, interactive experiences, but sharing them was a pain. And then every time it was if it was like two days later, and I wanted to like, okay, now update with new data, have to totally rebuild from scratch, and it was inconsistent in what it was rebuilding.
20:45And so having those chats with Jacob and others on the team was like, what if we, you know, could turn this into something? And, you know, so then moment I had Dives, it was like, oh, my gosh, this is immediately what I was looking for all along.
20:59So that's awesome. Well, that's a super encouraging story. I think, you know, I've had a similar experience as well. I hesitated on Claude Code longer than I should have. And I'm glad that I just jumped in.
21:12We like to say dove in when we're talking about Dives. Nice work there. Last couple things, then we're going to get to actually see some more fun stuff. So what is next for Dives? We're going to give you the ability to kind of give stamp of approval on Dives. So that way you as an organization can pick the ones you want to put the most emphasis on. Maybe the ones that you report out to your board of directors, etc. We want to give you the ability to kind of certify things.
21:36We're looking at doing write back capability. You could have a full app where you could edit some data. And then we're also going to get really granular with how you can do permissioning for these types as well.
21:46So that's what I had on the slide side. I wanted to go ahead and kick off a dive to go ahead and build one while we're chatting here. So Claude's going to work in the background, and then we're going to shift over to see some demos of what has already been created. So in this case, I did write the prompt ahead of time.
22:02I will paste it in and we'll talk about it while Claude thinks about it. So really, I just use a quick keyword in MotherDuck to kind of help it out. So it understands where to go to look for this data. And then just go say, hey, go explore this particular data set on ambient air quality. And I'd forgotten the name of the table. So I just said it's on this topic and it's going to be able to find it for us.
22:22Hey, go ahead and look online. I know this is a public data set. Learn about it what you can. And then I kind of ask it sort of a very high level purpose statement here.
22:34I want to be inspired. AI inspire me. That's kind of a silly thing, right? But if I put my goal into the prompt, there's a lot of times where if I don't explicitly ask for something, the model is able to still go in a direction to where I want.
22:48So that's one quick tip is actually give the high level goal, not just the very tactical goal that you want. So I want to do that. Then I also want to see the top success stories. What are some of the most positive things that have happened lately? And I'm giving some more details about what I want. And then after you do it, I want you to create a very interactive dive. And this is the keyword that says, OK, don't just do the data analysis with the MCP. Go ahead and create a dive.
23:14And I also am very picky about my charts. I said, hey, don't put any pie charts in there type of deal. Make sure that the charts look nice. And then I added some requests for interactivity. I want to be able to zoom in. I want to be able to click on any chart to go full screen. And then I want to be able to slice and dice things with filters.
23:33And and we're going to see as we come back that this is going to be able to do that. So I did have to allow it to do some things. And it's going to go ahead and chug away. And with that, I was going to say, well, it's doing that one other thing that I would say that's really helped me. So like Alex is really comfortable with visualizations and has a lot of things that he knows he wants to see in the data.
23:54I would say if you're not like that and let's say you're just good at when you see something, you know, you like it. The other tip is I often will just have Claude. I will tell Claude what I'm thinking that I would like to have a dive of right.
24:07Like, here's the goal I'm trying to accomplish. Now, ask me some questions to help me build out what would be a good visual. And it does a really good job at like, here's two or three different questions that I'll ask you to help figure out what you'd want to get out of it.
24:21And that often like helps me. I'm better at answering questions about what I might want to see than I am always at like coming up with it in a single shot. So don't be intimidated by like, can I come up with a perfect prompt like how Alex did?
24:33Yes, that is a limitation of the demo environment that we're in. Thank you very much for sharing how it really works in the real world there. So we'll let that we'll let that cook, as they say. And Logan would love to hear a little bit about what you've put together. Do you want to go ahead and share your screen? Yeah, let me share. I'll start by sharing the dive. Okay.
24:59And the backstory here on this dive is that, for me, my experience might be like a lot of others on kind of sales teams is that I typically have an access to whatever reporting was in our CRM we were using. So Salesforce, HubSpot, whatever it might be.
25:17And maybe one kind of other tool. And for most of my career, I worked really closely with our RevOps team on, hey, I saw this, can we tweak it to say this, or I need it to say this, or maybe I'd get in and try to get my hands dirty. But again, depending on the tool, the customizability, the editability, not always so easy to do.
25:37And so for me, once we had the MCP, and I could start asking questions of the data, then it was like, oh, I know the questions I've been wanting to ask. But the hurdle has been how easy was it to do it. So one of the first things that I started to do once we had the MCP, and then dives was, I wanted to be able to look at a book and understand usage, I wanted to be able to understand like how my customers were doing growth trends, declining trends.
26:04And there was no dashboard that we had that looked at this level of granularity prior. So this is a synthetic demo version of exactly what my experience would have been. And kind of what I wanted to see were I want to be able to see trends, right. So this is an example I built out where I've got multiple reps that I might look at across the team, I could kind of tier my accounts based on how we might tier them, what kind of stage they're at in the pipeline, trending, things like that.
26:34And all of this is fully interactive, right. So if what I'm trying to best understand are, let's say I want to look at folks who have grown the most in usage over the last seven days, I might sort it like this, right. And you can see I'm on like version three, right, it only took me a couple of prompts to be able to get to the point where I had something that was fully interactive in this way. I also knew that what I wanted was kind of a deep dive capability. So I could click on any account, and I could see more under the hood about what was happening.
27:03This type of kind of speed would not have been possible just with an artifact, right? Like, to be able to get this level of granularity, I needed to have my hands actually on the data, which we do here at MotherDuck. So now I can see a day by day breakdown of usage, I could see if anything was kind of spilling to disk. And these are all things that are helpful for me and how I'm managing the accounts. I could also again, sort by any other factor, sort by am I looking at just one of the reps books, you know, what it might be.
27:31We also have folks who buy what we think of as like an annual or capacity plan, right. And one of the things I'm trying to do to be good stewards for them is understand how are they doing for the kind of allotted capacity. So I built a capacity tracker, right, that just allows me to see each of the customers and get an understanding of like, how they're doing compared to kind of the capacity that they've signed for, are they in overages? Are those potentially new workloads they've brought on? And how can I support them?
27:59And again, nothing like this existed in our dashboarding prior. So dives allowed me to build this out, also allowed me to do things like see what kind of folks did we label as starters, but maybe are growing, right. So if I'm Casey, right, I might come in every morning or every few days and just say, Do I have any folks that I previously thought were just kind of a starter to your account, but actually have had a lot of recent growth, and I'd want to take a look at what's happening there. And maybe I'd, oh, okay, well, this one's interesting, like, their usage is up quite a bit over
28:29the last seven days, even though I'd label them a starter, maybe I should dig in and find out what's happening there. So this is where I've found dives to be super helpful. I always say, though, like, for me, the visual is the like, beginning of the sentence, it like helps me start to get eyes into something in a visual way. And I'm a visual learner. But it's not like, oh, that answered every question for me. It often just means I have more questions. So let me
28:59I'm going to add this other tab in here. So you can see. Okay, great. I think you might want to just do a share where you share the window instead of the tab. Oh, let me share that. Let me just share here. Sorry. All good.
29:23I think some of my favorite part about what you've built here is that these are things that folks in the sales domain know deep down of exactly how they want to operate. Like, you know that I want to look at people that I previously haven't really talked to a lot, I want to see if they are now, maybe in need of my help.
29:47You're tracking things in a way that's very specific to also what you've seen be most impactful for helping your customers, you know, based on the capacity limits or some other things. So because you're the expert in sales, you can build the best sales dashboard. And that has not always been true. Because the tools weren't there to catch up to team up with you. And so this basically, I mean, unlocks you as a sales expert, I think it's really cool.
30:12Yeah, Jacob has always said, like, those who have the best, like, context on what's happening in that area, like now have also the ability to go and build what they need. I've done this with like golf data recently and building out like better versions of, you know, golf leaderboards than what I was finding elsewhere in that type of thing. So
30:30cool. Okay, so now you can see this is a side by side. And the reason I bring this up is this is actually what my experience would be like, right. So again, if I'm thinking about that, I'm using this dive here. One of the first things I see is that this lumen fold account, I can see that, you know, usage is up, you know, 22% in the last seven days. Great. That sounds good.
30:52But I really want to investigate that further with the thought partner that is Claude. And this is literally the type of thing I might do on any given day when I'm seeing account trending. So the first thing I do is just like, hey, I just want to make sure you go find the right underlying data, right? So hey, go find it. It's in the sales demo dive database goes through it.
31:12Next thing is I'm asking it, hey, what is driving that lumen fold growth from the past week, right? Again, if I come here, like I know that their growth is here, I can see that in my dive. So that visual is guiding me to then want to ask this question.
31:28And it starts to analyze what's happening and takes a look. And yeah, it's a real spike. And it looks like it's coming from this. And what's interesting, the number of users haven't moved. And, and one thing that, you know, folks like Jacob have always said, it's like,
31:42Claude doesn't mind saying a lot of words, but like humans processing a lot of words is something that like, I see a lot in here. I'm like, oh, that all kind of makes sense to me. But like, maybe it's also more than I need. So one of the things I tend to do is I like read it all, I take it in, I'm like, okay, that's really interesting.
31:59The main thing I want to know, like, but is it stable or spiky? Like, is this something that is something I could count on and predict and be like, great, this is their new normal? Or is this a one off event, right? And so it helps kind of analyze that, right. And again, prior to having the MCP and prior to having a dive, I probably would have actually had to go pair with my customer engineer, who is more adept at writing SQL, and we'd have to then go investigate the data directly and try to figure out what's in there.
32:28And then maybe intuit some things. And there's just a lot more value in me being able to go in there, try to hypothesize, ask some questions, and then pair up to be a better account partner. So, you know, takes a look, says it's fairly stable. Here's the reason it looks like it's fairly stable. And so my final question is just like, can we just summarize this in a simple sentence, then, so that I can just like know is a dead giveaway, like, what's happening here and review that with my CE.
32:55And the, you know, the summary is, hey, they're duckling hours, their usage essentially stepped up to x on April 27, has held steady. Seems like this is, you know, something that's they're fairly much their new normal, right? So to me, that's like kind of the beauty of this handshake that's happening between the dive and MCP is I build a dashboard I actually love and can use, it helps me find things in the data that I might not have been able to otherwise.
33:21And then I can use something like Claude and the MCP to ask a couple levels deeper of questions to make sure I know what's really happening here. So I can be a good partner to my customers. You bet. And I know that also the dive, you've not kind of kept the dive static, you've kind of improved that over time. So what does that workflow look like? Do you ask the question three times and then realize, oh, maybe I could build that into the dive? Or what is your iteration look like?
33:46Yeah, I would say a couple of things. So one is I first built it for myself and had a handful of kind of version and updates based on what I wanted to see from it. And then I was like, hey, this seems pretty decent, helpful to me, I'll share it with a couple of other folks that, you know, work here as well and have similar workloads. So other AEs and a couple of them used it. And then they found it helpful, but they provided some feedback on what would make it even better for them. And then finally, I just shared it broadly with all of the sales team. And then I got some more feedback.
34:14And then just over time, there's just new things like again, this idea of I was like, oh, what about starter customers and like a watch list and that and then we introduced kind of the annual capacity plans. And so then I added that. So I'm actually on version 100 of my real dive. And it's been like a really fun, beautiful journey to get there. It's not like, oh, my gosh, like the hundred edits came over the course of the last, I don't know, four or five months. And it's just like every once in a while, I think of a tweak that would make a lot of sense. And I tweak it. And I always do.
34:44Part of just my standard process is at the end of every quarter, I look back at the prior quarter, what I thought went well, and what didn't and what I'd want to do differently in the next. And so I'm doing that right now, because we just had a quarter in. And I'm thinking, what about my dive? Might I want to edit? And also, I'm kind of working on kind of a morning command center dive, because that's one of the things I realized is like, what if I just surface like the five things I need to know each morning. And so I'm actually working on that right now as well. So just being able to build that for myself is really just giving me a lot of power.
35:15That's awesome. Yeah, that's, that's cool that you're finding additional spots to use it. I think maybe one or two other quick questions, then we also get to see another demo on the other side. So can you tell me a little bit about the data sets that you're querying here? Are these, you know, really small data sets really big kind of how, what's the data look like under the under the hood?
35:34Oh, I mean, it's all of our customers in the actual data set, not the synthetic one. And so I mean, I think we're in like the eight to 10 terabyte range of like the amount of data that it's looking at in terms of all customers, obviously, like, what I tend to be interested, just like anyone else is, I'm asking questions of a smaller subset of that. But in terms of the total available data that we're building the dive off of, it's all of our customers, it's all of the usage that's happening, it's constantly being
36:04updated with, you know, new information as it's coming in. And so it's an immense amount of data. And that's part of why the dive is so powerful is like, you can see that the interactions and again, this is even though this is the demo data, like, the demo experience is no different from my own experience, and how quickly the stuff loads and how quickly I'm able to edit and jump between different reps and like, you know, see these experiences changing, like, it feels just this fast when I'm even working with a really large data set.
36:34Just given the speed of MotherDuck and the power of the dive. So yeah, it's not it's not like I'm talking about like, oh, it's four accounts in a handful of rows. Got it. And I think that's part of part of our sweet spot. So it's, it's, we like to call it duck fooding around here, instead of dog fooding, where we where we use our own product. And, and I think that was a great, great case. Well, thank you so much for sharing what you're up to, Logan, any anything else you want to add?
36:58No, just a huge fan. And I think if anyone listening is intimidated anyway, by this, like, the best way is just to start using it. Like, for me, that was it. And that's part of why I try to be an early adopter for some of this stuff is, I just don't want to be behind on things that can make me a better resource and a better partner. Again, I keep thinking of good questions are good ways to be a resource for my accounts. But I didn't always know how to build them. And now I feel like I have some pretty great tools.
37:26That's awesome. Well, very cool. Well, well, thank you for giving us a look into the the modern sales side. Ryan, love to hear a little bit about how how dives mother can CP and AI are helping on the marketing side.
37:38Yeah, absolutely. So I, I generated a similar dashboard as to Logan. So I, you know, I have sort of the marketing official dashboard for our team. And this is a version that is based off of synthetic data. So it's actually only like a handful of prompts where I basically told Claude, okay, now now generate something that I can show publicly in a webinar.
38:07Which was was also very helpful for me as a marketer to take the real data that I'm using. And, you know, tell Claude to clean it up for for public presentation. But this is the exact same dive as we use to manage our team. It just is is with a different set of data.
38:28And so you can see the you can see the different tabs at the top here. These are the things that we concern ourselves with. Our primary goal as a team is around these sales qualified accounts here. And you can see the you know, the trends over time. This is I'm suggesting or I'm choosing just Q1. So I manage all the data by quarter.
38:55And, you know, even have detailed methodology here that explains on how all of this is calculated. And so we have a sales qualified accounts target every quarter. And then, you know, how many are remaining to target. And now some of you might be wondering what the color bars here. These are our XMR charts that are that are showing us.
39:21Basically, you know, when we have outliers in our data, when we have exceptional events in our data and converting these to from regular charts to XMR charts was a simple prompt. And so it's very helpful as we do our business reviews.
39:38But, you know, this is the lead funnel. So we have, you know, we're concerned about sales qualified accounts. We're concerned about, you know, how many marketing qualified leads we get and how those lead into sales qualified accounts. And then we have, you know, we have new marketing programs. that we're starting to run and we want to see how those new marketing programs are doing.
40:05And, you know, you can see our. . . Alex, fill me in with sparklines. I'm sorry, I was trying to remember the name of this. But literally, I said, okay, add the first column as a sparkline, and it shows me the trend there. So I can see the numbers, which the raw numbers is often what I want to look at.
40:25But then I can also see a quick visual view of each of these lines. And so this is new marketing programs, showing me where people are at each of the lifecycle stages, and then looking at all the different programs. And then, you know, a very simple visualization of the lead funnel.
40:46You know, this was the. . . I think this is a one-shot prompt version of a visualization of the lead funnel. I'm sure we could get a lot nicer looking here. But, you know, this is the main dashboard of the overall business and the top goals that we have.
41:04The important thing as, you know, Logan had shown you is just how do you can drill down. So in all of my dashboards, I basically say every data point I want to be able to drill down on. And so I click on a data point, and I can say, okay, what were the sales qualified deals this
41:25week? So, you know, the week of April 27th, in this case, this was the last quarter I'm showing here. And then being able to look at it by source and say, oh, wait, you know, my paid campaigns on YouTube are actually working out. And I can see, you know, each of the specific deals that came
41:46through it. And really just kind of understand the data better by just drilling down into it within the dive. And again, this was just one or two more prompts basically saying, you know, show me a pop-up when I click on any data point in the chart that gives me the details. And that was probably the first prompt. The second prompt is, oh, now I want these extra columns.
42:11So what use cases, who the competitor was, you know, those types of things. So, you know, it really allows that very quick drill down. And we've used traditional BI tools in the past. And yes, they allow drill downs. But I never did them in many cases, because it was too
42:31complicated, took too much time. And this is literally just one, you know, one click. And then oftentimes we'll have questions around, you know, is it, here's the overall lead traffic, is it paid or is it organic traffic that's growing? And, you know, I added this toggle
42:53again in sort of one prompt as to whether it's paid or organic. So that's our main, you know, our main metrics here that we use. But then to talk about, you know, the other types of things that we concern ourselves with is, you know, we've had
43:11sequence emails that we collaborate on with our business development reps, BDRs. And you want to see how those sequence emails are doing. You know, this was, this is probably a half hour, 45 minutes of work to say, okay, I want to look at those sequence emails. I want to look at, you know, build this chart of how clicks are doing or how opens are doing. You know, this is all relatively
43:36steady. But in our real data, as we do different experiments, you can see the changes. And you can see the changes over time. On the webinars, you know, this is one of those examples of a domain expert. So this was originally built by Jerel in a separate dive. And then I said, okay, take this
43:58dive and move it into my dive with, you know, I just copied the URL and said, move it into my dive and use my dive styles. But he was able to build this. He's a marketer with a non-technical background, the non-technical ducks, NTDs, as we call them. And, you know, he was able to build this as a domain expert to really understand, okay, the different webinars that we're doing
44:23and how it breaks down between ICP and non-ICP attendees. And he did a similar. . . It's a little meta. That's a little meta. Through the looking glass. Yeah, exactly. And we have in-person events too. You know, and then I'm like, oh, I want to see my web analytics data. So this is using post-hoc data and pulled into our data warehouse and then
44:48analyzing, you know, all of our top URLs, how referrers are doing, you know, Google search console data, closed one deals, and where those closed one deals are coming from in terms of competing solutions and current solutions. Like this was, you know, all combined probably
45:09two days of work. Like that is what's the most impressive thing, right? Like is how quickly you're able to build some of this. So, you know, this is what excites me is the speed at which someone who has the business context and doesn't need to have translation to be done between their
45:33business requirements and their technical side is able to build these things. And now I'm going to just switch the screen I'm presenting here very quickly. And let's see here, share screen.
45:50And I'm going to share a different window. All right. So this is just vanilla plot, right? So you can see at the top here it says free plan upgrade. The data, the synthetic data that I generated is called task. It's, you know, a B2B task management product. And I want to say, like,
46:15use MotherDuck to show me tasks. Business data schema. So using a slightly technical word there,
46:34schema. Here's what's super interesting is like I'm telling it, okay, it can look and do discovery of different connectors. And I actually don't even have, I never have mother, I not have MotherDuck set up in this account. This was a brand new account. MotherDuck is not connected here.
46:59So it was like starting from scratch and Claude. And it says, oh, I found the MotherDuck connector. Would you like me to connect? I'll say connect. And say accept. And hopefully Alex timed that as per our earlier discussion. Oh, man. And we can see how quick it is to connect your
47:22Claude to your MotherDuck. I would say anecdotally that seems closer to 20 seconds than five minutes. Exactly. And if I clicked a little bit faster, it would have been even better than that 20 seconds. I love being proven wrong in all the right ways. That's my favorite. Exactly. So, and I didn't even look through the directory. I was originally going to show you
47:45looking through the directory, but then I thought about just in this webinar and I gave Claude a chance of discovering it on its own. But here it says, okay, found this task marketing demo share database. Let me pull its tables. We have 17 tables. Let me fetch columns. We'll do that
48:01in parallel. And it will start to describe the type of data that is here. And the, let's see here. You can see all the different things that it's doing. So, it's doing list columns on all of
48:20the different tables. So, the event analytics data, the Google search data, et cetera, et cetera, et cetera. So, it's doing a lot of different requests to kind of discover what this data is about. My HubSpot contacts, the deals tables, you know, how contacts are associated with orgs.
48:41So, it's a fairly large data set in terms of the different tables and the different columns. And that means that maybe I should have tested this exact query before. And because I'm getting running out of time here. But you can see it's really doing a great discovery. Okay. There it
49:04goes. So, found everything. Now, let me render this as a clean. Claude is slowing down on me. Maybe I shouldn't be using the free plan here on Claude. I actually experienced this also in the demo that we kicked off as well. Claude, you know, Anthropic just doubled their usage limits yesterday with their announcement. So, I have a feeling that people are having fun with those new usage limits.
49:27The I think that's what's happening. Okay. Well, I mean, it's generating a schema diagram for me, which is awesome. You know, I also do have MotherDuck itself open here. So, you can see the actual share that it's looking at. And also, with a demo account. But you can see I have event
49:51analytics data. I have Google search data. All my HubSpot data. Our different targets. So, this is where I kind of manually update our targets for different quarters. So, you can see all of this data in here. And then, you know, if you did want to preview any of it, MotherDuck
50:14actually has, where do I go here? Previewing data is pretty easy to see what that data looks like. And it will even show you, allow you to search for columns in particular tables, all of that jazz. But here, we actually got a schema diagram. And again, this was fresh, right out of the box,
50:36never connected to MotherDuck. It generated this schema diagram that, to figure out how I can zoom in here for this purpose, but generated a full schema diagram of all the data, which might be more useful for the geeks in the audience. But, you know, at the same
50:55time, is it also then described all of the data that's there. And so, anyway, that's what I wanted to show you folks today. I want to leave some time for some questions. But thanks all for letting me show this. Well, thank you. Yeah. So, we'll do some Q&A. So, if anyone has a question, chime in. We've got a couple. So, thank you for your questions so far. Put your other questions in
51:20there if you've got questions for Logan and Ryan. So, first one here. I'm interested in how you guys upgrade an exploration dive to an official dashboard. Do you have an internal review process where a data person looks at it? What does that look like here at MotherDuck?
51:40I can talk about that a little bit. I mean, this is early days, right? Like I said, November is really when things started to get really good and when we started exploring these things. We do have a concept called blessed dives internally at MotherDuck, which we maintain
52:03in a Git repo as code. And so, someone can take a dive or have one of their more technical duck friends take a dive and submit it to GitHub for review and approval by our data team. And then
52:22kind of that concept of being blessed so that people know that it's one of our official dives within our organization. For the marketing standpoint, what I've done is I've created this marketing official dive where different people on the team do things. And when I, as the marketing
52:44leader, feel that it's ready to include in the marketing official dive, I just use words to describe that this is more blessed. And then I'll pull it into the marketing official dive so that all the other leaders in the company can see that and know that it is blessed by me.
53:08But this will get better over time, right? It's brand new. We'll see better processes in place and best practices in place for doing this. But I do think it's really important right now to be at the bleeding edge for some of these things because that is your competitive advantage.
53:29Yes. And I'll chime in briefly, then I definitely want to hear what you're thinking now, Logan. So, I think for me, what's also interesting too is having worked on the data side, often what I would do is I would actually go to the business expert to do the validation on their side. But now the person that has the domain expertise has actually built the dive. So, in some cases, it's actually pre-validated in a really powerful way compared to the past.
53:53But this also means that now the two sides of this artificial fence we've created, right, can talk to each other on the same exact data set with the same exact environment. And I think you can get the validation from both sides. Logan, what are your thoughts? Yeah, I mean, I was gonna say something similar that you said earlier at the beginning that we often have like a gut instinct or a gut read on something because we're really close to it.
54:19I wouldn't have built Ryan's dive because it's not something I'm as close to. He would have a better sense of both what he wants and probably the story that the data is telling and how that aligns to his gut. So, before he would even need to consider it being blessed or that I would need to with mine, like I can just look at something and have a pretty good idea if it seems way off
54:38or not, right? And so, I was building an updated, like I mentioned, a golf leaderboard, a master's one. And there was something I was seeing that it was like, that's impossible, you know? And sure enough, my own kind of context helped there. So, I think there's ways you could do it officially, like Ryan mentioned in this idea of like blessed dives and using GitHub. And there's also just
55:00like trusting your gut and making sure that you're close enough to it. And then I also start, once I built something that I feel pretty good about, then I share it with just a couple of people so that they can also then bring their intuition and gut to it. And then eventually, it just goes broader from there. Yeah, there's this fear out there that you need all sorts of
55:20metadata, all sorts of context in all of these things to get the right answers, which is true if you don't understand the data, if you're not working with the data day to day and in that business realm. But every time that I've had a problem with it being off,
55:43it being wrong, it finding the wrong answer, it's always been, it's been off, like tremendously off, right? It's been like, this is not even anywhere close. And me as a domain expert, I immediately sense that and know, okay, I did something wrong, it looked at the wrong column, it looked at the wrong. . . I've not had cases where it's like, oh, it's off by a tiny bit.
56:07It's either way off or where I can detect it, or it's right. And so, we'll think about like the percentage of the queries that it gets right and things like that, which all sorts of interesting research that people on our team are doing. But the reality is just by having the domain expert doing these things, the detection possibility for when it's off is so much better.
56:34Fantastic. Well, thank you. Well, keep your questions coming, folks. We'll do our best to answer them. We'll do maybe one more question here, and then I can show at least the demo of the dive that was created in the background while we were chatting today. So, John says, I've only started playing around with dives with Claude, and he agrees it was super fast to get going. So, thank you for that. Appreciate that. It's good to always hear it from the real world
56:57side. Thank you. And he was able to. . . He loves XMR charts too. Thank you, John. Yes, that was good. I was obviously limiting myself because I didn't even think about drill downs, multi-tab scheming diagrams. I think that's part of what we like doing on these types of sessions. It's just we've had a chance to really dig into this and look at it for a long time and really think about, it would be cool if. . . And then we try it and it works in like
57:22a couple of minutes. And that would be my encouragement to everybody here is just give it a try. The cost to try is now like pennies and a couple of minutes. And by and large, this stuff really works well. And so, I'm also just a huge XMR aficionado. That's sort of like how Amazon does their reporting is using this stuff. And it allows you to focus on what really
57:44matters and tell if something is a real signal or a fake signal. And so, that's my favorite type of chart. To inspire you a little bit, check out the Dive Gallery. You can just search for MotherDuck Dive Gallery. And there are some amazing dives here that have very detailed, very
58:08interesting data sets, whether you're looking at business data or you're a sports ball person, which I am not. We all get to have our own fun things that we like to do. So, I did post that link in the channel. So, have a look there. And then I wanted to briefly show the dive that was created while we were talking here. So, this is looking at air quality. And I
58:30said, I want to feel good about the. . . I think in our news media, we don't often get the good news. So, I want a little bit of good news out of this data. And so, it's given me a lot of different filter options where I can toggle to specific places. And it shows that, hey, there's a lot of cities that improved by more than 20%. We've got these metrics here, which is like a measure
58:55of how many particles are in the air. And when there's particles in the air, they go in our lungs. So, we like less particles. And that's improving a lot. And oh, look at that beautiful downward trend. I love it. There's countries that have improved the most. And you can see even at the city level, it's able to break this down. And it's got specific success stories of exactly where things are going. So, we got something that's highly interactive. We have these kind of
59:19abilities to maximize, minimize that we asked for. And we've got some encouragement for today. And for those of you who are joining late, if you can scroll up here, Alex, it was literally just one paragraph prompt that he pasted in there. No knowledge of the data beforehand, just pointing at a sample data set of who air quality data, like World Health
59:42Organization air quality data. And then he just let it go while we were demoing everything else. And that's what it produced, like super detailed analytics, very interactive analytics. You can see it directly inside Claude. then you can actually, you know, that is a dive that is then in, in your MotherDuck account
60:06and you can share with others in your organization. Correct. So that link will take us straight to MotherDuck and see the same thing, but this was even built into your, your Claude environment, right? In your Claude browser. So there's, you know, not even a separate webpage to get used to. It's, it's right there. And you can also go super fast back and forth to iterate. It's funny, I actually forgot we launched this. It was like a week or two ago and there's so much coming out. It's so cool.
60:30It's great because it made just being in Claude instead of having to do Claude Code with a local iteration. Like you can just see it all in real time, which has been awesome. Yes. So that iteration speed is key to be able to go back and do this. So we have, we have run out of time a little bit today. Thank you all for your questions. We can follow up with you afterwards as well and get your questions answered.
60:51It has been a pleasure having you all here and happy diving. Thanks all. All right. Thank you.
