Ridge Ai with Tableau Public’s “Godmother” - Ellie Fields

0:00Ellie, how are you doing?

0:01I'm great.

0:02How are you, Tim?

0:03Good.

0:04It's good to have you back.

0:05Yeah, that's great to be on.

0:07Absolutely.

0:08We spoke

0:09uh just just over three weeks ago, I I I think you reached out because you wanted to talk a bit about the thing you're working on today.

0:16So you've come on to talk about Ridge data, is that right?

0:19That's right, Ridge AI.

0:21Ridge AI.

0:22Fine.

0:22Ridge AI.

0:23Okay.

0:23So before we get into that, obviously for many people watching, uh, they might know who you are, but I also assume a large part of my audience probably started learning Tableau well after

0:33you'd moved on from Tableau.

0:35So why don't you give a bit of a history of who you are, um, what you're doing now and um yeah, a bit of your heritage.

0:41Yeah, sure.

0:42Um I was at Tableau uh

0:44Starting in oh eight when it was under a hundred people and was hired to do a split product management product marketing job and ended up doing that and launching Tableau Public, which is still kind of one of the highlights of my career.

0:55Um one of the things I'm most proud of is uh

0:58Uh I got the nickname Godmother of Tableau Public, which I just think is a and I'll I'll I'll hold on to the bottom.

1:03That might be the thumbnail, be careful.

1:05Yeah.

1:06I uh you know it was back in the days where we couldn't get any press coverage for Tableau.

1:10We had this

1:11core group of really excited customers who could see what we were doing and we believed in it, but press wasn't covering us.

1:17Um, you know, for Tableau Public, the first few

1:20publishes the first six months or so we're calling up friends and family.

1:23Do you have a blog?

1:24Do you want to publish something?

1:25So it it it really took off from there and uh it was a great community and a uh great product and I stayed for over a decade.

1:32uh which is not what I had expected to do, but really loved um what we're doing there and uh believed and still believe that data can can transform conversations, organizations.

1:43one.

1:44One of the things we did there, uh just leading into Ridge a bit is we we would sell embedded analytics, which are really analytics that you you put on the web for usually for other people to consume.

1:54And um

1:56Uh you know, I think we had a really good offering at the time uh for the technology available, but it was still a little bit slow and it took um

2:03Uh the way we had to price it was was uh pretty inflexible and high.

2:07Um anyway, I left Tableau after um after the Salesforce acquisition about a year and a half after, and I went to a company called Salesoft

2:15Uh and I went there because they had this nascent uh sales workflow, so action workflow.

2:20And I'm a huge believer that data and action, you know, peanut butter and jelly, they go together.

2:24Yeah.

2:25I wanted to bring data into that workflow.

2:28Because a lot of people will say, hey, sellers aren't analysts, they're not data people.

2:32And and that's true.

2:33You don't want them kicking around in Tableau or somewhere else.

2:36Um they have other selling things to do.

2:38But if you give them data in the right moment, it can really help them take action.

2:43And and and I believe that about almost anything out there.

2:45If you position the data right, you you make it relevant, you make it strong

2:49uh it can really help us have better conversations, take better action, and so on.

2:53Right.

2:54So I stayed there almost four years and and we built out a platform that uh really expanded on the sales workflow and did put action and data in this in this tight loop.

3:03Okay.

3:04When I when I ended up leaving there, um I was gonna take some time off 'cause it had been about what it between those two stints almost sixteen years and about a week in between, two children along the way

3:18And I figured I could use some time to uh to to just uh rest for a bit.

3:22Um so it was nice to have a holiday season with no meetings and and uh no acquisitions or anything else going on.

3:29Um

3:29But uh right around what the time I left I started talking to my co-founder, Jeff Hare, who is sort of legendary in the visualization space.

3:38He's over at UW Leads um

3:40He leads the visualization center over there, he's been part of Project Soak.

3:44Vegan D3 and most recently Mosaic Open Source

3:48And we were r ruminating just on kind of the state of data on the web.

3:52Uh some of his recent work had made data on the web much faster, which was always a big problem with putting data on the web.

3:58Obviously the web's the universal communication mechanism.

4:01That's where that's where everything is.

4:03But data on the web is still in a really poor state.

4:06And it's always bugged me.

4:07Um when I was at SalesOff, we actually had to put data out for our customers.

4:11And um

4:13We looked at all the tools.

4:14I had someone on my team kind of do an evaluation so that we can just bring in Tableau because I'd been there.

4:20And he came back and

4:21Wow, you know, one of our best engineering managers and he said, you know, for the the price for for what we're gonna get is just not worth it with any of the leading tools out there.

4:29Wow.

4:30And so we did not buy one.

4:31We did end up putting in um data uh data lit uh lake with uh

4:36databricks.

4:37Um and we ended up building in-house and okay.

4:41We got some dashboards out for our customers, but it was slow and painful.

4:46And

4:46It was really existential for us.

4:48Um, sometimes people think of, oh, well, there's some reporting, whatever in that platform, who cares?

4:52But if you're building value for your customers.

4:55You need to show them the personal value you're building for them.

4:58You know, they need to see where the opportunities and gaps are in whatever you're offering for them, whatever job you're trying to do for them as a software or

5:06or um as a product, they need to be able to see that value, see, see what they're getting from it.

5:11And that's the purpose typically of of dashboards and reports in a product.

5:16Yeah.

5:16So it was existential for us at Sales Office.

5:18We were expanding what we were doing to do this well.

5:21And I was really frustrated by how long it took and how much time.

5:24And so when Jeff and I started talking about this

5:27I had a moment where it was almost like, you know, I really was hoping to take six months to a year off and read a ton of books.

5:34And as soon as we started talking about it, I was like, I'm gonna get excited about this.

5:38And

5:39Yeah.

5:39You know, we we just have both felt this need to go out and and make it better, make data on the web.

5:44So that's what we're doing, I think.

5:46So that's what we're doing, I think.

5:48Amazing.

5:49What a s what an arc.

5:50What a story.

5:51And I think, you know, for so many people listening, so much of it is like aligns with their experience, right?

5:57Um, you know, I think embedding in Tableau has always been a little bit of a thing that's possible, but it's

6:03J just like a little bit too painful.

6:05Um, we've gone into this world where people have started building their own solutions more in-house, right?

6:11Like um

6:12was streamlit was a little bit of that sort of energy inside of the um uh the data warehouse space, right?

6:18And then more recently, um I don't know if it's because of AI, we'll get into this later, but I I feel like the

6:26the the general community has discovered what software development is, right?

6:31So everyone's vibe coding their own analytics solutions and trying to stand up these things, you know, put put the vendors aside.

6:37Let's see what we can build ourselves.

6:38And so it's probably had, you know,

6:40a resurgence for the wrong reasons because I should say to people, no, you don't want to build your own solution.

6:44Like it's a lot of work.

6:48And this is why I actually wanted to talk you talk to you because I think there

6:51that balance is required, like making things safe and secure.

6:54We'll get into this, I'm sure, is is is a big challenge.

6:56And so I'm pretty excited about um rich AI because I think

7:01It is a need that is still unmet.

7:03It's still way too hard to go from a data source to being able to share something uh to the world.

7:08But even just the embedding

7:10playbook itself is it just feels over engineered and and still really hard to do in a world where you know you can take social media embeds and just put them anywhere you want.

7:19But yet with data, it's still quite a hard thing.

7:21So I'm I'm really happy to be talking about the product and yeah, I'm keen, keen to see what what you're gonna show us today.

7:27That's great.

7:28Well, should we jump into it?

7:29Absolutely.

7:30Let's do it.

7:30Let's do it.

7:30I'll give you the floor.

7:32Thanks, Tim.

7:33Yeah, I'm excited to show you Ridge.

7:35And uh, you know, there is a there are three big problems we've been trying to solve with Ridge, and one of them was performance.

7:41Um embedded analytics on analytics on the web, I believe, have been artificially

7:46uh just depressed because performance is so poor and and I've seen people trying to uh optimize, you know, down to ten seconds or five seconds or something that

7:56even even then it's just too much for users to deal with.

8:00Right.

8:01And so we've made everything very, very interactive.

8:03We've used um

8:05uh mosaic and duckdb and web assembly uh to get this performance which is amazing.

8:10It's really crisp.

8:12It's all uh the computation is all happening in the browser so you don't even have to worry about

8:17hallucinations on computation because it's happening right here.

8:20And if you're hitting something like a snowflake, you're also not paying for all that.

8:24Exactly.

8:25What's happening right here.

8:28And one of the st one of the other big problems has been that dashboards, while they're wonderful and we love them and they present a story and it's visual, they don't always let you ask questions.

8:37And so you might see

8:39Asked a question right here in the data agent.

8:42And this all of this data that we're looking at is usage data, adoption data of some product.

8:47And so I'm asking about different resource types that were used.

8:50Um that is actually not in the dashboard, but the dashboard and the data agent here are backed by the same set of data.

8:56I see.

8:56Yeah.

8:57Yeah.

8:57And they can have many fields.

8:59So they can have 50 fields or or or um

9:02uh or more and you can put those in the ridge and then that's available to the end user as this kind of long tail yeah

9:11Yeah, and that I think was something that the dashboards just suffered from because if people didn't get get what they wanted, they'd ask for another dashboard, and then you had many, many dashboards.

9:20Um so what I've just done is I've actually linked the um the dashboard and the data agent.

9:25Oh wow.

9:26Yeah, that's cool.

9:27That's so cool.

9:28Wow.

9:29So your conversation and your dashboard are like

9:32connected even though the fields aren't in the dashboard um the sort of the universal context you can kind of bring it in it looks like a dashboard now with a chat agent like now that you've linked them it else they they feel like they're one thing

9:44Right.

9:45And you can go and and ask, you know, ten or twenty questions and then use the filters and other natural affordances in a dashboard because that's one of the advantages of a dashboard is not just the ability to select and and touch the data and excuse me, but also the ability

9:59to um to use those filters.

10:00And you don't always want to type every filter out, you know, tell me action type by this, that, or the other.

10:05And so we we think there's a lot of room for innovation on the form factor.

10:09Dashboards have been the stalwart for decades and that's amazing and they do some things really, really, really well, but they don't answer those long tails.

10:18And so we're excited about

10:20with this performance and with AI, you know, browser-based technology, all the things that are happening right now, the ability to innovate on this form factor.

10:30So this is what you get with a ridge.

10:32And I'll show you in a minute how we created this dashboard on the left.

10:36But just to let you know, these are designed to be embedded.

10:39One of the other things I believe very deeply is that data should be where people are working.

10:43And so we've designed these ridges to to sit in your product, in your web page, whatever you might have.

10:50And this is some electric vehicle data on a on a made-up website about where you might want to put some chargers

10:57But you can see we've still got that interactive performance here.

10:59And of course, you've got other questions that you can ask and get those answers right here in the data agent.

11:07That's super awesome.

11:08And and uh the embedding embedding story has been around for a really long while, but I think that the way you you started this demo by talking about performance, that's actually been the problem.

11:17'Cause you you go into a web app, but it doesn't behave like a web app because the performance feels more like something you y you interact with in Excel, right?

11:25So much much faster.

11:26100%.

11:27We're so used to to subsecond performance and then we go to a a data asset and it's

11:32It's five seconds, ten seconds, and it it's just too long.

11:35Exactly.

11:36Um one thing that we're trying to do, and and uh this is really important to us is

11:41is build in best practice.

11:42So you can see even when I ask questions of the data agent, it's giving me a a good form factor over here for the data.

11:49It's giving me the right view for the data question that I've asked.

11:52We are using SQL here in the questions to generate the questions and generate uh sorry, we're using AI to generate the SQL and generate the questions.

12:01But then we do multiple rounds of validation over here.

12:04We've created a harness, if you want to use that term, to validate the sequel.

12:09to uh to use some uh deterministic methods when we know the best practice to create the chart and to validate everything along the way so that you get

12:18a very low error rate with this uh this AI over here.

12:21And the computation is still of course done locally in the browser.

12:25Amazing.

12:27Right.

12:28So let me show you how we build these things.

12:30Yeah.

12:31How does it work?

12:32Yeah.

12:33So the last big problem I think that

12:35data visualization and and uh data on the web in general has had is that it's just too hard to create.

12:41And if you think about people who use Tableau, they need to know

12:44tableau or whatever other front-end environment they're using, they need to know about their data, they need to know best practice, they need to understand the domain and the questions they're asking.

12:53And we think that

12:54You shouldn't have to take that much training before you go and create something.

12:58Right.

12:58So I'm gonna go ahead and create a ridge and uh

13:02I'll go ahead and select that same usage data we were looking at before.

13:05This is just data about some kind of product usage.

13:09Yep.

13:09And yep.

13:11We know in the cycle of visual analysis that as people are working with data, they have to see that data.

13:16And so we've we've given you this preview just as performant as um as what we saw before.

13:20And I can

13:21inspect it, make sure it's the right data, and then uh go ahead and create that new dashboard.

13:26And what we're doing here is we're we've actually added on the left what I what I call a build agent.

13:32Okay.

13:33It's it on the right we had the data agent.

13:36That was questions about your data.

13:37This this one's gonna help us build.

13:39Um we all know what context does these days.

13:41I'm gonna go ahead and say let's build.

13:44And what I'm interested in is actually adoption by department.

13:47That's what I'm trying to understand.

13:49Maybe we sell this product by department.

13:51I need to make sure all

13:53Departments are adopting, my champions are asking me that, where can I go and talk to people about adoption?

13:59And so I'm giving it my highest level goal, the adoption goal.

14:03And so the the build agents going to come back to me and say, well, what does that mean to you if you're going to look at adoption?

14:08Well, I care about coordinate actions and distinct users.

14:12I actually don't really care about success failure right here

14:15Um, but uh I do care about um action duration and actions per user.

14:21Maybe those are things I care about.

14:22I could actually change any of these because Oh wow

14:26Yeah.

14:27The suggested query.

14:28That's quite intuitive.

14:30Yeah.

14:30We're we're trying to solve the blank screen problem.

14:33And and we've inspected your data.

14:35We know this is about usage.

14:36You're getting different questions than you would get if it was marketing data or manufacturing data.

14:41But we also know the data is distinct enough and people's purposes are distinct enough that they need to be able to edit.

14:47Yeah.

14:48And so I'll do the same here.

14:50I'll say, you know, what's the weekly growth rate of actions.

14:55It's now finally asking me about questions I want to answer in the dashboard.

14:58So it knows what metrics are important.

15:00It knows what I'm trying to do.

15:02And so I'll uh uh excuse me, I'll I'll go and select a few of these questions um and uh

15:09And maybe I'll just look for the most common action types per department.

15:14Maybe let's do a correlation here because those are always interesting.

15:19In between maybe average duration.

15:23And um actions and let's put that by department.

15:28Interesting.

15:29And uh and so we've got we've got a number of questions here.

15:31We'll see maybe pick one or two more.

15:34Um

15:36As you're as you're selecting these and you hit refresh, is that sort of helping the context of the new questions that come back in, if that makes sense?

15:44Yeah.

15:44Yeah, exactly.

15:45We're trying to learn from everything that's happening here and ultimately from the different roles in your organization as well.

15:50Like what are they interested in and um what what are they looking for?

15:54Right.

15:55Um

15:55So we've given all that context now to Ridge, but if you notice, I've never had to say, well, I want a bar chart here or this is a this on this axis or whatnot.

16:04Yeah, true.

16:05Yeah.

16:05You haven't had to define the canvas, as it were.

16:08Yeah.

16:08Yeah.

16:09And and just to point out, we're we're dealing with about 200, well, it it was too fast.

16:13About 250 rows here and about 20 fields.

16:17And so what it's done is it's now given me um what should be a best practice dashboard using all the

16:24suppose that we've learned about how people interact with data and it's it's interactive from the get-go and perform it from the get-go.

16:31Yeah.

16:32Gosh, it's so crispy, man.

16:33I wish

16:34I wish uh, you know, a certain tool had that today.

16:38Well, it was always one of the big bugger booze when I was at Tableau.

16:41It just it it it just

16:42It's hard for people to work with data when it's so slow.

16:45Oh, interesting.

16:45You just swap them there and then they they swap positions.

16:48It infers your kind of intent through just the actions.

16:51It makes a lot of sense.

16:52Yeah.

16:53Right.

16:53And I you know, you can do some other editing.

16:55What we're trying to do

16:57is get you to pretty much a 90% good dashboard um without you having to do much of anything and then you can go and do a few more edits and change things around, etc.

17:09But um

17:10It's very funny.

17:11All the toil.

17:12Like even when I sit down to to create something in Tash in in Tableau, I it might take me a half an hour to an hour, even though I'm pretty good at Tableau and I know exactly where I want to go.

17:21I just gotta

17:22click all the buttons and it takes a while.

17:25Um true, true.

17:27And I guess the other thing is uh you have sort of

17:31Uh several challenges.

17:33This is a pretty important one because this is also going to be the end product when you embed it.

17:38So you kind of need to help the user build something that they feel comfortable sharing with the world.

17:43It's not

17:43It's not as simple as a dashboard that you're going to use yourself or maybe just share within your team.

17:47Your target audience for something like this is much bigger.

17:50So the best practice and making it sort of work well really meets the expectation of

17:56a much larger audience that I think most analysts are used to.

17:59You know, we're not we're not used to shipping stuff that tends to have, you know, audiences that start in the thousands rather than end in the thousands, if that makes sense.

18:07Yeah, that makes sense.

18:09And that's it's a good point.

18:10It's it's where best practice comes in.

18:12I mean if you're if you're creating an ad hoc analysis for yourself and you know exactly what you want to get to, that that's fine.

18:17Or maybe you're creating it for someone else.

18:19But

18:20When you are putting it on the web, it needs to use all those principles about how do humans actually understand what's going on.

18:28Yeah.

18:28Yeah.

18:29Amazing.

18:30I guess, and then so once you save this, like what happens?

18:34Well what where does it where does it go from here?

18:37Well, I'll save it and um and then I'll just show you.

18:40We can click over to the explore tab here just to get a sense for what the end user would see.

18:44Yeah.

18:45Um and then we can go back into the app.

18:49So let me go.

18:51back into ridge here and um I can go find that recent ridge ridges got quite a few in here and uh

19:02And I can just get the embed code or you could do a server embed.

19:05And the server embed would be if you wanted to partition it, say if you wanted to see have different customers see different views in the same dashboard but their own data, you you might do a

19:14a server embed using JAT.

19:16Otherwise you just do a simple iframe embed, generate that embed link, and then you've got That's it, you're good to go.

19:22Yeah.

19:22You're ready to go.

19:23Yeah.

19:24So simple.

19:26It's amazing.

19:27And and I think um is there any sort of thought around um I'm gonna call it like MCP tooling and skills if in in a very sort of general sense.

19:36So is it possible if someone has RIGE to be able to

19:40uh use like an agent to help build the ridge because essentially you already have the the skeleton frameworks in order to be able to um I just just replace that sort of chat interface with a Claude or whatever people want to use um if that makes sense.

19:54Yeah, I I love the way you're thinking about that.

19:57Um there's a few things.

19:59One, that when you when I was building, yeah, all I was doing, all that that Rich had was the context

20:06and the data.

20:07Right.

20:07And so we're making that available via an API and and ultimately an MCP.

20:13So that if you do want to just say, hey, I'm going to give you a blob of context or maybe at the end of a workflow, even an agentic book post.

20:20Exactly.

20:20Yeah.

20:21context right you could just say here's the data set I c I care about here's the context and just go and you could end up with a a a dashboard from that yeah um we're also looking at um using mcps in here so say

20:35You have DBT or you you're you're sitting on top of Snowflake or Databricks.

20:40They have um integrations where you could say, hey, tell me about the provenance of this data.

20:44We really do believe in being a pluggable

20:47presentation layer for data for and being very interoperable.

20:51And that means that we don't want to do everything in here.

20:55We want to reach out.

20:57to your other products.

20:58If that's where you created the data and you understand all builds and joins and the provenance of that data, where it came from, you should be able to ask that question here.

21:08Yeah.

21:09And like yeah, go ahead.

21:11I was saying I said I I get excited about this because I think one of the biggest limitations of BI tools is that

21:16Because they've always needed to be driven, it's not been really that easy or scalable to have what I would say like uh it's gonna sound um not ideal, but procedurally generated um sort of

21:28data stories and so the context I think of is sports where you might have lots of different audiences that want to attract different things but they don't they don't they don't share the same universe and so this idea of

21:38And a sports user comes into like a website, maybe they're tracking the World Cup as an example, right?

21:44Um, they can type their own story.

21:46Uh but the data still comes from your universe and then it goes, you know, the you can procedurally generate something of interest to the individual and it's just a ridge.

21:55You just it's just the same data set, it's just different questions.

21:57Um and it's super exciting.

21:59Yeah.

22:00Yeah, we've had customers um one one serves uh local government departments with analytics and each

22:07local government has different legal and regulatory things that they need to show.

22:11So they all have the same data, but for for this little city or miss you need to create

22:23Yeah, yeah, exactly.

22:24Like ever every fan has a passion, right?

22:26And

22:26I think football's a good one because uh no no no two football fans can agree on which statistics matter.

22:31So you're giving people their chance to tell the story their own way is a very good way to diffuse a lot of uh let's say uh

22:40uh fan to fan uh sort of um what's the word debate yeah yeah yeah amazing it's really cool I I'm I'm I'm really excited by this because I think

22:53I I don't know if it's just AI, but it feels like there's there's been a lot of pent-up innovation in the space and suddenly there is lots of avenues for this kind of technology to be explored and it's really powerful.

23:05Yeah.

23:05I love it.

23:06I it it's it's I don't know if it's surprising to me, but like why it's surprising to me, but

23:13It it feels so obvious, like you know what I mean?

23:16Like it's I I guess I guess the core of your product is actually the um repeatability.

23:24And the uh, let's say the the AI, I'm gonna call it the AI interface.

23:29Shoot me down if I'm wrong, but yeah, the the sort of interface you're giving the user with the data through AI to get to that outcome, but also to

23:37to engage with that data in ways that I don't think have been um broadly accessible today if you embed a tableau public viz as an example, right?

23:46You have to educate the user on how to use that specific viz, but that doesn't carry over to anything else.

23:52Whereas you're just using language, you're just using that prompt window, and that's something that people innately have.

23:57Um it doesn't stop them asking silly questions, but it does generally make sure that even if it answers a question, um, yeah, the answer's gonna be pretty well presented, which is which is which is great.

24:08Yeah.

24:08Well I know you've watched people learn Tableau and other tools over time, as have I.

24:13And you know, I just uh after after doing it so many times, hundreds if not thousands, I just I realize people have to understand the data

24:21Yeah.

24:21They have to understand their domain.

24:23They have to understand a front-end tool, which is something like a tableau.

24:27Um, and they need to understand database best practice, which is something not many people understand, which is why sometimes, you know, you get something that's a bit

24:35unwieldy in whatever tool it's been built in.

24:39Um and when you think about that, having the expectation that to communicate with data, you need to have all of those skill sets is really a high expectation.

24:48So trying to take it down to, hey, you know what questions you want to ask, you know your business, you know your domain.

24:56And we can bring a lot of the rest of that into the experience.

24:59Yeah, yeah.

25:01In terms of um you talked about semantics and you you you you know you you highlighted the product's focus is to sit on top of your data stack.

25:08Um, what are some of the biggest challenges there?

25:10Because I know that the industry has, you know, decided finally to align on this thing they're calling OSI, which is, hey, let's all actually make sure that when you run a query in your stack and my stack, they add up to the same thing.

25:23Yeah.

25:23Because we acknowledge it hasn't always done that.

25:25Um that that that I'm a I'm a huge skeptic of that approach.

25:29I I've wrote a newsletter about it because I think it's actually quite a challenging thing to do.

25:34From your perspective, you're building a product that hopes to sit on top of that.

25:38What are some of the biggest challenges there that you you sort of see on the horizon in terms of

25:43how things like that might sort of affect your product, but also where you might have to, you know, get your hands at and fix problems because those those initiatives can't fix them.

25:53Yeah, that's a great question.

25:55I mean, I have yet to see someone who has perfect data and a perfect staff.

25:59Exactly.

26:00Right.

26:00I mean, I I think if if you have perfect data, it means you're out of business and probably have been for quite a while.

26:07But I think I think all of those things are helpful.

26:10One of our founding principles is to be interoperable and to work with people's data stack.

26:15And so uh, you know, you can have

26:18a really well well refined semantic layer and work with rich and we'll sit on top of it.

26:22Or you can have none and just have data in a Postgres and your semantic layer is is table header.

26:27Right.

26:27And we'll work with that.

26:29You know, you can have really sophisticated uh ETL and you can do a lot of validation in your stack or you you can do none and we'll

26:40help you work through that as well.

26:42So really, you know, nobody can make their data perfect, and we certainly can't make people's data perfect, but what we can do is acknowledge the reality of the situation.

26:51that there are different stacks at different levels using different kinds of technology.

26:55And just try to make the promise that what you've tried to centralize in your stack and what you have ready to show to customers will respect that and will will encourage it.

27:05Yeah, amazing.

27:07Um when we when we last spoke, um we talked about so many customers that I think would have loved a tool like this, would still love a tool like this.

27:15And

27:15Um non-for-profits are obviously a big one that come to mind.

27:18I can't think of how many organizations, even just you know, profit for profit companies, you know, journalism, so many of those sectors need a tool like this.

27:27And um

27:28I think one of the one of the things I've thought about since then is how do you how do you s how do you see this tool going from what I would call

27:38um a customer to the analyst.

27:41Because I think that's one of the biggest, let's say, challenges that I think I realize today a lot of BI companies have had, which is that

27:50They sell to a very different persona than the person who uses it, right?

27:55Yeah.

27:55And the the best product fit for the

27:59like the product is not the buyer and never rarely is the buyer.

28:05But with your product where that those two really do need to be so close together, how how do you how do you close that gap?

28:10Because I

28:11I I see it as a big challenge today.

28:12I realize, you know, Salesforce have failed on that front in so many different uh realms.

28:17Yeah.

28:17How how do you solve for that problem given you're building a company um, I guess from scratch?

28:21It was a great question.

28:22I'd love to hear your take on it.

28:24But um one of the things I think Yeah.

28:28One of the things I think Tableau struggled with in embedded and so did some of the more legacy solutions was

28:33the delivery model.

28:35And it you had to do this massive, massive implementation, hundreds of thousands of dollars.

28:40And obviously then you're getting everybody involved, your CFO, you you're doing this long thing.

28:45I think there's

28:46There's something lost when you have to start there.

28:49So what we're doing is we're we're actually making it possible to uh have a have a trial and then have a free

28:55And forever.

28:57And then if you want to go and embed something and make it possible to an external use case, then we start to charge you.

29:03And we do it use case by use case.

29:05I see.

29:06And the reason why we do that is because a lot of companies already have

29:09embedded analytics or external external analytics or what have you.

29:14And they don't want to rip it all out.

29:16But they have somebody, and it's often somebody in the product team if it's a product company.

29:20who has just built out some amazing functionality or is delivering value and they want to show the value.

29:25They want to communicate that to the customers.

29:28And so they have this need.

29:29They can't get engineering resources.

29:31There's no way they're going to go convince their organization to buy look or a tableau or something at 500,000.

29:36And and so

29:38Part of how we're thinking about that is really thinking through that person who is sitting there frustrated that they can't get their story out.

29:46And if you were that person

29:48What would you want?

29:49You'd want something that you could get going fast with, that you could play with before you have to buy, that you could if you if you could buy it, uh, we have price points mostly under $10,000 a year.

29:59And so you can get something up on your website without going and getting a massive budget or doing a big process rigomroll.

30:07So I don't know that's how that's how we're thinking about it.

30:09I'd love to hear what you think about it.

30:11It's a great it's a great one.

30:12I think one of my biggest frustrations is that I've you know, I've come to realise that, you know, I'm sat here teaching the world um products, I guess, and

30:21I've always had access to every part of the product.

30:24So it's always been very easy for me to rationalize this is the best way to use something.

30:29But I've done a couple of talks before and I've realized, you know, I I did one talk at um

30:34uh Salesforce Tower in London, I asked every hey everyone, like put your hands up if if you've got, you know, this license.

30:40And like no one put their hands up.

30:41And then I said, okay, put your hands up if you buy this product.

30:45Everyone put their hand up, plus I so you all have the license.

30:49You just don't know it.

30:51And so on one hand, you've got this huge disconnect.

30:54I guess discoverability is one of those challenges, right?

30:57And then on the other hand, you've got people who when they look at that price, they kind of um the tr you know, procurement is trying to protect the wallet.

31:08They don't like

31:09The number doesn't care, the feature doesn't care.

31:10They just want to say, what's the lowest number available on this price sheet?

31:13And can can we do we really need anything above that?

31:16And that can rip out the innovation out of a product very quickly, right?

31:20And make it very normal.

31:22So the the the the the best example of a solution I've had, like talk to a few people is like uh I use Tableau in this context, is like Tableau should have always had a license that was basically

31:35gave you every part of the product up to a point.

31:38So basically you'd you you have to play with the boundaries very well.

31:41But if I take um data management which is massively underutilized,

31:47They should give you a certain amount of credits that are always free.

31:50Yeah.

31:50Yeah.

31:50And what you do is you create a problem for the admin where they have to decide who gets those free credits.

31:55The product can always do it.

31:56Everyone can do it.

31:57But if you want to increase that utilization, then you basically already have a good problem, basically, right?

32:03At the moment you have no experience of it, so you have no value to demonstrate to anyone

32:07And you have to take such a big risk because your license is tied in for the whole year, so you have to go to someone and say, Hey, can we just increase our budget by twenty percent?

32:15um for this feature that I think might be useful.

32:17It's like no one's gonna no one's gonna have that conversation.

32:20You never try it in a trial because you're so busy trying to set up the whole environment right in the first place.

32:25Governance is like the big issue.

32:27So I've always said that every product should have just basically let you try every part of the product free up until the point.

32:35This is typically called freemium, but I think it should be a bit more aggressive in that

32:40Companies are also different scales, different sizes, right?

32:43And success looks different at different sizes.

32:46So if you're, let's say, in a big 20,000-person enterprise,

32:50That boundary needs to be a little bit more permissive because you'll hit that limit very quickly.

32:55Whereas if you're in a small like you know 50 to 80 person organization, that boundary can be a lot lower because again, success looks very different.

33:02And

33:02It just feels like in all the pricing strategies that I've seen in the industry, like it does that that's the bit that's never been innovated on.

33:09Like it was just, can we figure out sort of

33:13Dare I be the one to suggest this dynamic pricing that actually helps the license be discovered

33:21I can't believe I just suggested that as a solution.

33:23But you know what I mean?

33:24I mean it in a positive sense rather than this negative sense where we're trying to find out

33:29Like what's the what's the what's the maximum you're prepared to pay?

33:32Instead flip it, let's try and find the maximum value that a client can have.

33:35But anyway, that's that's Yeah, I know I love that.

33:37I I mean I think there's also we we we

33:40just discuss this uh in so many ways over the years.

33:43But um I think there's also the fundamental

33:48question that I have is is does everyone want to be an analyst?

33:52Yeah.

33:52I go back to I think er I think anybody who's really trying to do their job well, they want to use, they want to be informed.

33:59They want to use data.

34:00But that doesn't necessarily mean going into a tool like a tabloid.

34:05You know, they may not leave need the license.

34:07And that's why we keep coming back to this embedded scenario, right?

34:10It's like you the person who wants to share it pays pays the fee.

34:15You know, after the freemium, after the after the free trial.

34:19And then it's it's free.

34:20Now it may be authenticated via JOT or something if it's your in your product because you need to make sure the data is secure.

34:27But once you're you to your to your end customers,

34:30To them, it's just seamless.

34:31It's just part of the data product.

34:32It's completely white labeled.

34:33It's completely seamless.

34:34There's no extra auth.

34:36And I think you can do that internally as well.

34:38Because

34:39getting into another product is a hurdle, right?

34:41It's a hurdle for everyone.

34:43I think as an industry, being more creative about pricing, I agree with you on that, will help help us kind of broaden the conversation.

34:52And I think

34:53recognizing that data just has to sit alongside work.

34:57And there will be people who help the data sit alongside work.

34:59And maybe we call them analysts and data engineers and others.

35:02And then there will be people who do the work.

35:04But those people don't have to be analysts.

35:06Mm-hmm.

35:07Yeah, it's true.

35:08It's true.

35:09Uh you you raise a good point as well, which is the surface of work.

35:13Um I like your product because it's

35:16Maybe if maybe if I say this, you can challenge me on this.

35:18I think it's one of the easiest ones to place exactly where it fits into the flow of work, if that makes sense.

35:24Um back in the day you would have heard of this phrase of being in the flow of analysis, right, inside a tablet, right?

35:30And more recently I've been having debates about with it with AI coming around, like where are people going to be doing the work?

35:36Are they going to be doing it inside of your claws and chat GPTs?

35:39Which will have MCPs and hooks to reach into your you know infrastructure, or actually will it be more um in in things like Microsoft Teams and Slack and then those are actually controlling the agents and going out there?

35:51The the jury's out on this

35:53The thing I like about your product is it just lives on the web and the web is everywhere.

35:57So really, it can be anywhere you want it to be.

35:59As long as it it it has the internet and it has some ability to

36:05uh run some sort of browser technology that's modern is gonna be good and actually even I I don't wanna spe speak on your behalf here, but

36:13Phones today have as much sort of browsing compute as even like you know in a typical laptop.

36:18So actually this can be a really universal um uh product as well, which is pretty exciting.

36:23Yeah, I I the web is a universal communication mechanism.

36:27100%.

36:28Yeah.

36:28Yeah.

36:28Same, yeah.

36:29So I actually won't challenge you on that because I think uh it's I mean what

36:33obviously we need to we need to do the work to to get together people.

36:37But I think you know, i i Tableau tried and and did such a uh did so much work in terms of taking us from the world where everyone had to kind of

36:48code to to uh uh you know a more drag and drop world.

36:53Um but again there's just those fundamental assumptions some people are not gonna go

36:59Some people won't write documents, you know, and that's okay.

37:02That they may want to read them, but they don't want to write them.

37:04And others will edit them.

37:05And so I think we just need to have a a more robust sense of what's happening.

37:09I do know that

37:10Humans are always going to need to work with data.

37:12Because we get people who will say things like, well, in the age of AI, um, all the data analysis will be done kind of in a black box, and you'll just get the answer.

37:20The answer is four and a half.

37:22Yeah.

37:22Like, well that I don't believe that because if you're working in any kind of a complex environment where you need to understand what you're doing

37:29You need some kind of sense making about what's happening that got you that four and a half.

37:34Maybe you don't have to do all the computation, but you need you need a way to process the data.

37:39And we

37:40We know that that is visualizing data.

37:42We know that that we work well with visualizing data.

37:45Yeah, yeah, absolutely.

37:47Yeah, fantastic.

37:48So um yeah, I guess if people want to try Ridge AI, how how how do they do it?

37:53Like what's yeah, what's the wait list look like?

37:55Yeah.

37:55You you you uh platform to tell them how to get involved.

37:59Yeah, thanks.

37:59Uh the wait list is a couple hundred people now.

38:01Um we're working through it.

38:03Every week we're we're opening up more licenses.

38:07Yep, and you can just go to ridgedata.

38:10ai to get on the wait list.

38:12I'll put it on screen so you can uh go and have a look.

38:15And um one of the uh interesting things I noticed that you launched roughly the same day as uh Golden and Francois.

38:23The thing I really loved about that is actually uh this this community of, let's say, people launching products who've all come, let's say, from the same heritage.

38:31I think that's super special in a really fun way because it shows that, you know

38:35You can have many different ideas out of the same great foundation, which I which I think is a really strong message as well.

38:42Positive message to send to people that

38:44Um, although we have tribes, um, you can build new ones and that's that's how innovation works, which is super fantastic

38:52Yeah.

38:52Tim, I I have a question for you.

38:55How have you seen, I mean, you're working with all the kinds of companies and enterprises.

38:59How have you seen

39:01AI adoption, AI value, like what what's going on with data and AI enterprises?

39:07Um

39:09This is a hard one to answer because um I use one of my favorite terms at the moment which is leading and lagging indicators, right?

39:18So there's there's

39:21The indicators that I would say are leading indicators are when you have um teams who've traditionally taken really good care of their warehouse.

39:34um are already at the point of maturity where they can have, I guess, really advanced questions and discussions around

39:43what the next sort of phase of their innovation looks like.

39:46And I say phase because it is a very strategic move.

39:48You can't just like click a finger like the big startups and suddenly be, you know, running compute in AWS, standing up your AI agents.

39:55It's

39:56For a lot of, you know, normal enterprises who don't have venture capital to the hilt, like you have to take a lot more measured approach

40:04What they're starting to do is they're starting to have what I would call like a really high level of semantic hygiene, which is, okay, for a long time we've had

40:12This discussion about semantic analytics.

40:15Like now it really matters.

40:17Like, okay.

40:18For anyone who's on the fence, let's just put that to bed now.

40:21Let's let's get behind this program.

40:23Let's stand up these projects.

40:25Those take time and uh you know the people who are doing that are far and few between, their salaries have gone up, so it's also pretty hard to find those people.

40:34So that's sort of phase one, leading indicator, semantic hygiene, standing up these projects.

40:39lagging indicators, um, you're starting to get people in the line of work sort of going off and doing their own thing.

40:48And I mean this in a positive way.

40:49I don't mean this in like

40:50you know, someone rogue on another laptop with with Chat GPT and Claw taking screenshots to to do their work.

40:55What I mean by this is people are starting to find new ways of doing their work.

41:01in anticipation of this innovation arriving.

41:04So I say this because that still s sort of signals the gap, which is you've got the the infra work sort of starting to happen.

41:12You've got your Databrixes and Snowflake standing up capabilities to put agents on top of your warehouse, which I think will enable a lot more people to do it, especially those who don't have the resources to build these big infrastructures.

41:23You've got the people broadly, I'll say broadly, because not everyone's really sort of bought into this, but broadly speaking, people are starting to try and find, okay, what does actually work today and how can we get that set up for the future?

41:35Where there's a gap is how do we close that gap, right?

41:39So that that that is still a thing I just do not know because

41:45That is a missing piece, right?

41:47You can build a castle, but doesn't mean anyone will walk into it.

41:50Likewise, you can enable your whole entire organization, doesn't mean the tool's gonna work.

41:55Yeah.

41:56And so I say this to say that like

41:58I still cannot, for the life of me, see yet how you know your everyday enterprise closes that gap.

42:07Your big startups, your big um, you know, uh venture-backed things

42:12They're just throwing money at the problem, right?

42:14They'll they'll they'll set up AWS Turk, which is that, you know, service where you can hire like human beings in another part of the world to literally do the work that you think a machine should do.

42:25And they'll just blow through it.

42:27They'll treat it as a brute force uh problem.

42:29Um an everyday enterprise can't do that, uh, not least, you know, mm it's not profitable, it's not a good idea.

42:36So

42:37I'm still waiting for that answer.

42:39Um, I wish I I wish I'd seen more examples.

42:42There are definitely companies who have done this as well.

42:44I'm just not exposed to them.

42:45But I think if I talk about the masses, that's the gap I see at the moment.

42:49Um

42:50I am excited though.

42:51I think in two years' time we'll have this discussion and these these problems will largely be solved.

42:57Um, the models will largely be much, much more capable than they are.

43:02I don't know if maybe that's part of the solution.

43:04The capability of the model closes that gap for people.

43:07And then what was the software problem just goes away.

43:10Um

43:12But uh I think the biggest risks are companies who've not had good semantic hygiene

43:20just get left behind, woefully, because y you can't rush that work.

43:23If you just not if you just not had the heritage or the um discipline to

43:30Organize that work, you just get left behind.

43:32Um I worry that a lot of companies are gonna get found out.

43:37Yeah.

43:38See some solutions that help

43:40gather that context with AI.

43:42I think it can be I think it can be accelerated.

43:45Um but I I agree.

43:47And and I think one of the hardest things about semantics.

43:50is that people don't know they want it.

43:52It's one of those things where it's so valuable if you have it and it it accelerates so many things.

43:59But nobody wakes up in the morning and says, I want, I want some semantics because I love semantics.

44:04They want it for a hundred other reasons.

44:06And it and to your point, by the time they figure out that they need it for those hundred other reasons, they've got to go back and start over.

44:13Exactly.

44:14And I think the the other risk with AI is actually that the um the proliferation of tools that we're seeing now sucks semantics into more places.

44:21I mean that that was something I have

44:23last wave of BI too is that y you you had all this capability, but then your semantics are just spread everywhere.

44:29And and that's almost as bad as not having any semantics at all.

44:32True.

44:32And so I think there's

44:34You know, modern data professionals have to be very thoughtful about their stack and what lives, what layer of the stack, because otherwise they're they're going to get caught with a huge problem.

44:45Yeah, yeah.

44:46I've I've had a few um people at Celsius sort of push back and talk about, oh, why why is it bad that the data gravity isn't in like in in in our platform and not not elsewhere?

44:56And I think it goes back to this exact point you're making, like

44:59When you look at that problem you've just highlighted, it does make sense for the data gravity to be as, you know, as close to your warehouse as possible because, you know

45:08you if you're gonna go to you know twelve other places you want them all to have the same foundation and you don't want to have to multiply out um the effort.

45:16The other thing is um

45:18Yeah, if I look at sort of I'm gonna call Salesforce like a CRM as like an sort of end user capability.

45:24There is data that sits outside of that ecosystem that is also valuable to the enterprise, right?

45:29So you might want to say connect your CRM to

45:32Um, I don't know, your E um your um EPOS system if you're working in retail, right?

45:37So your transactions versus your Salesforce CRM records, whatever.

45:41Those two need to uh be brought together.

45:43There's only one good place to do that, and that's not going to be in either in either place.

45:47It's gonna be somewhere

45:49Central.

45:49Um and you actually earlier said something super important again that's relevant here, which is no organization has perfect data, right?

45:57Yeah.

45:57And so the job of semantics is never done.

46:00That's another thing I forgot to highlight, which is I think people think this is like a one-time strategic endeavor and they don't realize what they've started is a journey they can never stop.

46:09And that's gonna be a that's gonna be a rude awakening, I think.

46:12You know, people who've put investments thinking, oh yeah, this is this is just like to get ourselves on the AI train and then

46:19Your analysts will come if everything works, if everything goes to plan, what you should see is an explosion in the need for your semantic capability, right?

46:26Because people asking new questions, understanding better um ways of working.

46:31And your business should be evolving.

46:33I mean data is just a representation of what's happening in the real world.

46:35Of your business, exactly.

46:36It's just a a representation.

46:38So if your data stops, you have no more new data or no new semantics.

46:43It means

46:44It means something's not, you know, you're not reflecting what's happening in the real world.

46:47Yeah.

46:48Yeah, exactly.

46:49Exactly.

46:49I love it.

46:50I love it.

46:51Uh thank you so much.

46:52Um I don't wanna don't wanna sort of overextend uh the the

46:56the time i i really appreciate you joining the channel and i love i love what rige ai is building i'm gonna obviously have a go at the products we'll we'll make some videos over the next month or so

47:06Trying out the product.

47:07Maybe I'll try and build something uh for the real world.

47:10One of the things I'm trying to do is put together like a little like I'll call it just him stack or whatever.

47:15So you know

47:16I like that.

47:17Tools that aren't your everyday tool that we can we can get to do um very very interesting things.

47:23So uh look out for that in the in the near course.

47:25But again, um

47:26Ellie, thank you so much for joining us.

47:28Um hopefully we'll have you back maybe in three, six months.

47:31Who knows?

47:31You tell me when it's appropriate and uh yeah, we'll we'll see where Ridge AI has gone.

47:36Yeah.

47:36Thanks for having me, Tim.

47:37No worries.

47:38Pleasure.