ο»ΏSarah (00:00)
Welcome to Undubbed, the podcast that's unscripted, uncensored, and undeniably data. I'm Sarah.
Fiona (00:06)
And I'm Fi Today we're talking about something that's genuinely changing the way that Tableau works and the way that people interact with data.
Sarah (00:17)
Our guest today is one of the most well known faces in the New Zealand Tableau community. If you work with Tableau and are in Auckland or New Zealand and you don't know Steve Holly, I honestly don't know what to tell you.
Steve is a Tableau Solutions engineer at Salesforce and he's spent over 12 years across enterprise sales in the Tableau SE role. But before Salesforce, he built up a serious career across Oracle and IBM. And if you go all the way back to the beginning, Steve started out as a business analyst at Air New Zealand. We've done our research, Steve.
He has seen how organisations adopt and use data tools, not just in theory, but in the real messy, complicated world.
Fiona (01:00)
And have to mention this because it still makes me laugh. Steve has actually been a dub dub data client. Sarah gave his Salesforce push-up challenge a viz glow-up, so he knows firsthand just what we're about. That's linked in the show notes if you want to check it out.
Sarah (01:18)
Most recently, Steve MC'd at the Auckland Tableau Conference Insights event and presented on bringing the Tableau Magic to Salesforce. And we're so glad he said yes when we asked him to come on our podcast.
Now before we dive in, remember to subscribe to UnDUBBED and most importantly, please share this episode with a data leader or a Tableau user who is trying to figure out what all of this AI change actually means for them.
Fiona (01:48)
Welcome, Steve.
Steve Holley (01:50)
Yeah, thanks very much. Nice to be here.
Fiona (01:52)
Steve, we always love starting with the person behind the professional. Can you tell us what you're really excited about working with data right now?
Steve Holley (02:02)
I think it's the the change that's happening. I've been around data for a long time. prior to my stint at Oracle, I was implementing data warehouses and BI products, and the promise was to put the access to data in the fingertips of business users. And that was last century. And we've been working really hard to do that ever since. Promising, coming up with great UIs and
we just never got there. But now I think with AI, we are the closest we've ever been to allowing business users to really interact with data, to be self-sufficient and to to be data driven. And that's been the the dream that we've been working for since I've been around BI.
Sarah (02:43)
Yeah, really exciting times, right? And I think it's been such a hugely talked about topic. When Faye and I have been around probably not as long as you have, but almost. And it's right. It's always been, slowly getting there, but it does feel like it's it's almost cracked now.
So Steve, there's a term that's been flying around the tableau and broader data network lately. And we want to make sure our listeners actually understand what it means before we dive deeper. You've been right at the cold face of the shift, headless analytics and headless AI.
There's a phrase we keep hearing for someone who hasn't come across it before. What actually is happening here?
Steve Holley (03:25)
Yeah, it's a it's a really good question. And and look, I'm I'm expecting most people listening to this are probably in the IT world. And so the term headless is familiar to them. But I I'm married to a lawyer. And if I went to her and said that we're I'm working on something headless, y you can imagine the connotations in her mind, right? Headless means nothing outside of the world IT, right? In fact she talks about being on the farm and headless chickens.
Fiona (03:51)
Yes.
Steve Holley (03:52)
Yeah, and running around. And so the terms term only means stuff to us, but understanding what it means, right? So we've been doing headless from a tableau perspective for a very long time. So that is taking the the head of Tableau when you look at it is the UI that we present to the end users when they interact with Tableau.
But if you take that off Tableau, can they still access Tableau? And the answer is yes. And we've been we've had customers embedding Tableau into their applications so that their end users can do analytics on the data that's relevant to their application without even knowing that they're going into Tableau. They're just exploring data. And that that embedded experience, you know, this there's companies like in New Zealand, So Gentrack provide
billing software for utilities. And they have an application that includes charts and visualisations embedded into the UI so that their end users can explore the the data that's relevant to them. And that's th that's a part of being headless. I guess what's been really interesting is that the new head that is being put on Tableau is the well are the new
Agentic tools is is probably the right term. So things like Claude and and cursor and chat GPT, whatever you want to plug into the top is the new head. So
Headless BI it is is now something that's available through with a new head. And it's a new thinking head. So that that's a really cool thing, right? If you now go and replace the head you've taken off the chicken and put on a much more intelligent head, maybe the chicken will run differently. I don't know. But underneath you still need the chicken, right? The body to do what it's supposed to do. And that that's what Tableau can do.
Some of the interesting feedback we're getting around there is like, well, with Headless BI and with the GDIC tools, do we really need dashboards anymore?
And that that's been a really interesting conversation. And I had this conversation with our own country manager here at Salesforce. And I said to him, What would your world be like without dashboards? He said, No, you can't do that. I don't have a consistent base of analysis across all my reps, all my managers. If they're all gonna go to their own agentic tool and ask a question to give me the data to tell me where their business is at, they're all gonna get different answers. And all of a sudden, we're
Back into the world of Excel where everybody's downloaded the data to their own environment to get the answer to provide the manager. And so,
When I think about headless, I actually like to think that it's two heads are better than one. So yes, use the new agentic head on top of your VI, but also use what's been proven to work for decades now, which is having a consistent way of seeing and analysing data through highly visual dashboards that give the answers, the first up answers. When they have a question about what's behind the data, yeah, flip to your agentic side of things and drill down. So yeah, I guess that's the the view that I see.
Fiona (06:28)
Mm-hmm.
Steve Holley (06:44)
happening in the market with the headless statement that's out there.
Fiona (06:47)
Hmm. And it's quite fascinating the way that you're describing it there, having a a two-headed operation, that's giving the the I'm imagining there the chicken body sort of flapping around, and then maybe you've got the the human head and then you've got the chicken head sitting on the shoulder as well. And to me, what that means is I'm gonna have the flexibility of being able to dig deeper when I r need to dig deeper and perhaps
If I'm not a data professional being able to use natural language to do those queries, but then at the same time I'm going to have that governed, repeatable, safe and trusted information in the dashboard, which is saying, here's our performance, here's how we're performing against the targets that we have in place
or the the categories that we have in place and where do I need to be looking? And I think you're spot on and your your country manager is spot on as well. Without having that governance in place and those guidelines, people can go off and willy-nilly start to take a look at, well, how am I performing? But performing versus what? are they,
amending or massaging the way that the query is actually going to make it look more favorable.
Steve Holley (08:11)
Yeah, 100%. And it comes down to consistency of reporting across a business, right? So yeah, we we at Salesforce it's a a subscription business. So our business model is that we get people excited about the value that our technology can bring. So they start paying a monthly subscription.
And that's what's reported to the the markets about our current strength and our future strength because if they keep subscribing, we continue to have a strong business. But if they stop stop subscribing, then that starts to break our business model. So we have a measure around what we call customers a treading from our platform. And then so one of the questions that our sales teams get asked are what what are the what are your accounts that are at risk?
So that term at risk, right? It it it just y it means something to us. It means something to the individual account managers, it means something to their manager, but it it could mean completely different things. They measure what accounts are at risk by different criteria. They're at risk because the key contact has left. They're at risk because they've spoken to another customer who's using some other software. You know, the whole what what is that risk? And I think it that when you get a dashboard.
That says at risk is defined this way. This is how we consistently measure at risk. Then we're all on the same platform. And then you can go, for my at-risk customers, tell me some more information. What what's going wrong with them? Why is why the why is the level of adoption going down? Or something like that. You know, you get to drill down another level. So yeah, 100%.
Sarah (09:46)
Yeah.
Yeah. And I feel the focus is really on having that context as well. So it's one thing to have all the data, but giving it the contextual background is what's going to make Headless AI super successful and having people get the same results. And you actually mentioned something I find quite interesting and I haven't really thought about this before, around the dashboard.
being the starting point because if I'm a country manager or so forth, I want that set across all my directs or my teams. And I want to be able to see it all standardised. Because when I think of headless AI, I think of it more as the end users coming in and going, you know, what what should I be worried about today? And they're kind of doing it, maybe in the flow. So something like Slack and they're typing that out. And then I think of them going,
Okay, I've got it I've got that information, but now I want to go and see it in a dashboard. So I think it's interesting that it can kind of come from both angles.
Steve Holley (10:46)
Yeah, I like it it's a it should be dynamic, right? Users shouldn't be forced to go to a dashboard first, right? It some of the capabilities that we're delivering is the ability for the system to alert you. So we get back to management by exception. So you you can talk to the sales managers here and you'll go, How's business? Hey, it's trucking along as expected. That there's nothing to worry about. They they keep doing their job, but
when when something goes wrong, they need to be alerted to it and and go and figure it out. And so having the system push out an alert into Slack or Teams or whatever environment is that you do collaboration, that something's untoward, not is unusual, something needs attention, can then drive that that conversation around, okay, I need to know more using an agentic interface to dig into the data. But you still need a consistent
set of metrics, definitions, and level playing field, if you like, that everybody can go to to to understand how the business is going. I guess that's the real nice thing. There is no rules, right? We're not saying you have to go to a particular environment, you have to start on a dashboard. It's it's the ability to access data the way you want to when you want to.
Fiona (11:58)
Mm. And you touched a little bit on the trust side of things. So that, trust needs to be baked in, both in the dashboard and into the AI side. So if insights are being pushed to people automatically or those alerts are being pushed to people automatically, how does Tableau make sure that it's coming through as accurate and not just some sort of hallucination that may sound plausible?
Steve Holley (12:26)
Yeah, it's a really good question. And it's I think something that a lot of people have seen with the agentic interfaces, right? Is the hallucinations. we were talking the other day about yeah, we use it in our own personal lives, right? And just asked it a question and it came back with an answer that that was wrong.
I knew it was wrong. So I thought I'd can you instead please reference this website and give me the answer again? And it came back. Said, you were right. I hadn't thought of that. Here's my answer now. I know more information. So yeah, a hundred percent. So there is in the flow of using the large language models, Salesforce have implemented a trust layer.
So that the questions go through in a complete process, and there's some key elements to that. So removing bias and hallucinations are two key parts to it. we've also built in the ability to have business rules, we call them business preferences, so that you don't have to rewrite a massive prompt every time you want the agent to behave in a particular way, because that's a key part, right?
if you don't tell it something it's gonna start using its default behaviors. So there's that there to guide it along its way and also to make sure that it's using understanding the right terminology and things. So there's a number of layers we've got that are used before it even goes and gets the answer. And then
One of the key aspects is that and this is a key part to trust. So trust is Salesforce's number one value. So we don't want customers worrying about the fact that their data might be shared outside of the realm that we already hold it in or protect it in. And so no customer data leaves the platform. It's all encrypted, it's all hidden away from anybody, so that if transmissions were somehow broken into but they're all encrypted, that they can trust.
That their data is never going to be discovered. So there's a whole lot of technology that is more than just hopping on and sending a screenshot to Claude to get an interpretation about what's going on. It's there's a yeah quite a detailed lot of thought that's gone into that transmission and receiving from LLMs.
Fiona (14:36)
That that's absolutely right. Like and it's is so critical, especially, you know, the larger the organisation, or, our financials, if we think about our own bank, there's so many banks that out there that are using technology. How do we ensure that our own individual information is never going to be shared? Super, super important.
Sarah (14:57)
Okay, Steve, we've been talking about all the all the exciting stuff that's out there. Would you like to give us a bit of a demo?
Steve Holley (15:05)
Yeah, sure. Let let's do that. So I'm going to show how an end user can go into Slack to be able to ask questions of the data in a headless manner, like we've talked about, and then
to be able to see a little bit behind the covers about some of the controls that you can put in place to try and help the large language models understand the questions properly. So let me share my screen and we'll fire into it.
All right, so this is the Slack UI but through a browser. Now you there is a Slack app that you can use to
to be able to interact with and in fact I usually use the app, but for today's demo it's easier for me to use the browser. So if I go to my Slack home after seeing what's happening with my day, down the left hand side here you can see I've got an analytics and visualisation agent that I'm able to access.
Now it's got here, it's sent me an alert, and β so this is part of the proactivity we were talking about. So I've set it up so that to let me know when certain events happen. So in this particular case, the total margins dropped versus the prior period for a particular part of the business year to date. And then it gives me a lot more detail about it and then gives me some links to follow up with and what have you. But
β and if I wanted to, I can go back, I can go through see all the history of words there. β I can collaborate on this, I can have conversations with my colleagues about this. So it's bringing data into the conversation with Slack, which is part of really what the big question is. So
Because as a company we use Slack, I'm in Slack all the time. In fact, Slack is the first place I go to and the last place I go to at throughout the day. It's where all my colleagues know to contact me during the day. So it's about being able to access data where I work, right? In the flow of work, rather than having to go, β I need to now go and find out information about this. I have to go to some other reporting environment to go and do it. I don't want to have to do that. So I might then come in and go.
Ask it a question, say which region has the lowest profit, for example. Like that's triggered a thought for me, the margin dropping, β what what's happening with the profit, and get it to go away and do its work. So now the agent is now.
Doing what agents do, right? Translating that English request, natural language request, into a data question, and then it's going away to go and find the answers through the definitions that we've already provided from a tableau perspective about where to get the data for this information, what what does what does profit mean? How do we measure measure profit?
And how do we move forward with it? And so it's come back with an answer, which is really cool, and a visualisation. that we can go and have a look at that to understand that, okay, South is at the bottom of the rank and that may be something that we need to look into. So
Having done that, I having got that information back, I want my sales team to be aware of that. So I can then just go and say, hey, sales team, you just be aware that that's what's going on, and we need to do some work with it, right? So I'm still in that environment. I'm in that Slack environment, I'm sharing that information with others that that's in there. If I go to that channel,
There can now be a complete conversation in that channel about what's going on, what should we do about it, what are some of the tactics, why is it happening? You still got the human element, right? So two heads are better than one. people picking up what the agents found and bringing the the human mind in to interact with it.
So while we're working in Slack, we might also then want to dive into Salesforce, right? Salesforce is our CRM system. It's where we keep our opportunities, it's where we track our customers. And so when I go into Salesforce, I have β a dashboard built entirely in this environment. You're all familiar with dashboards, but we've done some really cool things with some of the capabilities there. So, as well as the bands that sit across the top, we've included straight out of the box.
prior period comparisons, and these are configured with a five-step wizard, right? You it's really easy to put these in place. And when you put one in place, it has all of these features that it shows around each of the figures that gives more context to the actual figure itself. Because quite often you'll come into dashboards, you'll see the big number, but you won't actually see any of the detail below. So it's got a a trend line that shows you what's happening β
over the period of time that you've filtered on. But also is that light blue band in the background which is showing you what the expected range is. And you can see it's harder to tell from the daily variations what the trend is. But when you look at the bright blue band, you can see that total sales are trending up a little bit, but margin is definitely trending down. And it's a dashboard like any all other dashboards. Like if I wanted to go, you know south was the area that we're interested in, I can click on south and everything gets filtered on it.
And now we get to see a much clearer picture that the margin has definitely been trending down over time. The other cool thing is that it gives you a further insight to what's behind that data that's sitting there. But at the end of the day, dashboards are designed to answer as many questions as they possibly can without making them cumbersome, right? And then you get to a point where I need to ask a question that the dashboard's not answering. So still in the flow work, I don't have to leave Salesforce, I can go and access.
The agent on that other side over here to be able to get some further insights. So β I might go let's go and compare Q2 margins now. We know lumber and building materials were an issue. How does that compare with tools and hardware? is it something that's happening across all my departments or something that's specific to the lumber and building materials department? So
That wasn't something I could answer in the dashboard. It wasn't immediately obvious. So it's going through and providing that information. The interesting thing here you'll see is that it's picked up that we are filtered in the dashboard on the south region. So it can pick up the context from the dashboard as we do it. So that that's that's really interesting. So now getting back to that risk conversation, I want to know what are the top five at risk accounts?
And base it on sales. So there's a concern for me here now. if our margin is going down, what else do I need to think about? So I've gone and asked the agent β to give me what those at-risk accounts are. And this is something that I think is really strong about what we do. And it's gone, I don't know what at risk means. And that's really important, right? It hasn't gone and made it up, it hasn't gone and tried to figure it out what at risk means. It's gone and
Captured the fact that doesn't know what at risk is. Now, as an end user, I can go, I don't like that answer. β it's inaccurate, and β you should know you're an agent. So I'm gonna provide that feedback to the agent and hope it can learn. So, in doing that, β what happens with that is that that information gets sent through to the back end.
For the people who are managing the environment to answer. Now, this is a semantic model. I'll come back and explain that. But if we go and have a look at these list of questions, these things that β initially when you go and set it up, you might want to go and β know the questions that you're going to ask the agent from your end business users. But here's one we didn't predict, right? We didn't predict the use of the word at risk. So that question's come through, and now I want to go and
Do some work and do some analysis on it. So it's going to look at that question. And in this case, because it's working with the data train team, it's actually going to try and solve it. From an end user perspective, it said, I don't understand at risk. I don't know what that means, so I'm not going to even try. But now it's working with a data professional. It's going to say, hey, at risk, and try and figure out what at risk might mean for the for the data person. And what it's doing, if we go back.
To the back to here is that you can see that it was working with β when we look at the sources, it's working with a particular semantic model. So this is what we saw before as we came in. This is a semantic model. What does that mean? these objects are just pointers to where data are. So these could be pointing at data stored within Salesforce, they could be pointing at data stored in an external cloud database, they could be pointing to β
Another BI database, whatever it is, they just pointers to data. And then that we've told it how to join things. We've given them the fields descriptions. We've given the tables descriptions. So this is the context layer that β the agent is using to get through to the data. And this these questions are are part of it. So now we come back and go, okay, so I think top top β at risk is probably by lower sales.
Well for us it's not. That's not how we define at risk. So I'm gonna say that's inaccurate and I'm gonna say that actually I want you to define at risk as something else. So any account with a return rate above thirty percent. So that's where people are returning product, right? So more than thirty percent. I save that, get it to calibrate it, and it's
Now coming up with a term, it can go and put into my business rules. I edit there and β save it. And now I can go and retest it. Now it takes this takes a little while, and I was thinking about this before I I knew we're going to be on a podcast and thinking, okay, this is gonna seem like a long time on the podcast. But if you think about the environment that people have had to work in in the past and you go,
Someone comes up with a fact, hey, we don't have at risk defined. Well, what process would they have gone through in the past? They would have probably raised a ticket internally for the data team to look at it, who would have then prioritized it, found time to do it, scheduled it, maybe it would have done been done next quarter, it would have gone through the process of being worked on, and then finally it would have come back into the business maybe two, three months down the track. And that's that's a long time. I'm sitting here talking.
and trying to stretch it out like a good news presenter to try and get it to happen while I'm talking. But it it does feel like a long time now, but it's actually really fast. And this is one of the cool things to remember that you this is stuff that has really accelerated the speed at which we can do stuff. This the Yeah, you go.
Fiona (25:59)
One one of the things,
β Steve, that I was really interested in, if you don't mind just while that's loading, β it's come back, but if we can just switch for a moment it back into Slack, I think it was. β and one of the things that I really loved in these queries was the source and the explanation in plain language behind where the data is being sourced from.
Steve Holley (26:12)
Yeah.
Fiona (26:27)
and how the data is being analysed. So for us when we look at a bar chart like that, we will automatically know that there's a bar chart and it's been sorted in descending order by β total margin. But what it's saying here for a sales team is in really plain language, how that is actually occurring. So people don't have to be trained in data. They just need to be able to interpret
a sentence to say, yes, that's right. I agree with the way that that's been done. Or perhaps they might come back to you, Steve, and say, actually that's not how I would β look at that particular query and it's nothing to be worried about. So I I really love the additional context that it's giving people on the queries rather than just looking at a a specific chart itself.
Steve Holley (27:21)
Yeah, I I totally agree. And one of the things I do really love about it is that I think in pictures. So I I don't even read this. I go straight to this. Yeah, because that's how I think and visualize information and I quickly understand it. But I some people think in words. And for them it's easier to read this than it is to try and understand a picture. But because we deliver both, β all all users are have their meats net.
Fiona (27:30)
Mm. Mm.
Sarah (27:50)
Yeah. And I
Fiona (27:50)
Abs absolutely.
I've seen even at the really upper level in the executive teams, sometimes they struggle with charts and it's quite embarrass it can be embarrassing for them. They can feel a bit of shame behind, β I can't understand or read or interpret these things. So I just really love the opportunity to have both.
Sarah (28:10)
I was just gonna say I I really like the sources part of it as well. It's really calling out, you know, what the filtering was on there and exactly where it came from. And then it's in the semantic model. So you're kind of serving everybody, you're serving the execs, you're serving those that prefer words, the ones that like visualisations, and then the techie people that just wanna know where that number came from. It's giving it everything.
Steve Holley (28:32)
Yeah.
Yeah.
And and of course when you because it's in Slack, if you wanted to go and share this with your β data team, you could, right? You just forward this to the Slack channel that is the data team and they would get all this context as well, including the link to the semantic model and the actual text of the question that was that was asked. So yeah, it's it's a great environment and ability to share things in a collaborative space as compared to, can you can you send me the screenshot? Which is
Sort of days
Fiona (29:05)
Yes.
Steve Holley (29:05)
of old. All right, let me jump back. That's quickly okay. So what it's come back with now that we've told it what at risk means is the right list of β users or customers that are at risk, right? And it's different from the one that was ripped back before. And we just have to say, yep, that's verified, you're good to go. What that means now is that if we go and let's say
go into go and ask it in some other part of the platform and let's go and use the later version of the agent and say, hey, let's go and ask that question again.
then it's going to go through its process and now it's got that terminology to be able to refer to and it can pick up on it and actually give you the answer that β that you're looking for. this will run.
Fiona (30:00)
So
one of the things that I while while it's running, I do β sort of have a few observations as it's coming through. the workflow that you've been taking us through just then, there are changes to every person's role within an organisation. There's changes to the sales team role and how they're working with data and understanding more about data and even sharing when data isn't
returning or they're unhappy with things. Then you've got the β data professional, whether it's the engineer or the administrator behind the agent, coming in and helping to redefine what that looks like for those queries with the at risk. that's a new role. Previously it may have been the dashboard developer might add a new calculation or put something in there that's
specifically described as an at risk metric. β But now we've got, this semantic or context updates that that are required and being able to review the the sequel that's in there and taking a look at that and what's being returned, being able to analyse is this right or is that wrong, pushing it back through. So there's quite a shift in
people's activities that they may be doing in their roles.
Steve Holley (31:28)
Yeah, absolutely. β there's also where I've noticed there's got to be a shift in thinking. And it's funny that β I hear a lot that β my customers telling me that they're shifting more things left. So they're doing more stuff in the data than they did in the past. And when β the AI interfaces first came out, there was a little bit of resistance saying, Well, now I have to go and change all the data. I have to change it at the data level. And I think the realisation
Is that that is actually the place that people need to be spending a lot more of their time is in that data level, getting the data right, because that's where the agents are going. Because in the past, I've seen some tableau data sources that have a huge number of calculated fields just to make the dashboard look good. And all of a sudden, if you release that to those calculated fields, the agent it gets completely confused. So yeah, there's a there's a change in thinking about.
how you prepare data for analysis now than was compared in the past as well.
Sarah (32:30)
Yeah, exactly right. And I feel like you're talking about this example and how it's it's been changed in the data, which is great because historically someone will just kind of whip that calculation up and fire it out and go off to a presentation. And then the person that's sitting next to them is like, Well, I've come up with my own calculation and it's it's not, the 30% stuff, it's something completely different.
And β and you've got those issues. What I'm interested in is there's a there's quite a lot of what Fi was saying around change management here. Is there a way like so someone in the back end has then come in and changed this? Is there anywhere that the front end person has been flagged that that's now available for them? Or is it just a case of them going back and querying it and seeing that it's available?
Steve Holley (33:21)
That's a really good question. And β I don't know if there I there's not an automated way, right? There's no process of publishing the business rules to the end users. I think it would be β client adopted best practice would be the the best way to to move forward with it. And that's you know, that's been the way of Tableau for a long time, right? We we've had the blueprint process to help people.
work with bus β BI environments to drive adoption, to get people familiar and happy with them. And I think that's that's always going to be the case. You're always going to have to have internal processes processes that support the technology. And I think that's a good example of one where you probably need to have that sort of thinking in place as you are working in this new world.
Sarah (34:10)
Mm. And maybe it's something like a Slack channel, right? It's the business rules Slack channel and and someone's updated it. There's been a question, it's been answered, and here it is.
Steve Holley (34:13)
Yeah.
Yeah, and that that initial person that provided the feedback, maybe they get a message back from the developers to say, yeah, we've implemented that now. You can ask a question. Or a a sales manager at the top level is told, by the way, we've implemented at risk as this rule. Please let everybody know. Yeah. I exactly how you roll it out. will be up to each individual client, I'm sure. But and and maybe in the future there is more
capability that's built into these tools 'cause remember we're still very young in the days of developing these capabilities. So maybe in the future there will be a way of communicating these out. But yeah, right now it'll be up to the individual companies, I think, yeah.
Well look, that the last thing that we always talk about is taking action, right? So that is to say, hey, now in this case we've identified a number of at-risk clients. Let's get them into a campaign and get them so that we're sending them the right messaging and keeping them on board so they don't donate trip. Thank you for allowing me to share that on your podcast.
Fiona (35:20)
That's awesome. And I love the the way that the flow comes through and it's so easy to close that loop in that last one. I think we we brushed over it really quickly, Steve. But being able to then have something that's regular and repetitive and is already in structure means that the teams can spend less time on cobbling stuff together.
and actually more time going out and really understanding how they can help people.
Steve Holley (35:49)
Yeah, yeah. I it it it's letting people live and work where they live and work. So not making them go other places to do things, right? So yeah, a hundred percent.
Sarah (35:59)
Yeah. And just on that, so for years the data model has been where the data lives. Somewhere you go and find it, and it's usually in a dashboard. What do you think's fundamentally changing about that model?
Steve Holley (36:14)
Yeah, interesting question. I'm not sure. Can you give me an example, envisioning?
Sarah (36:18)
Yeah, so I I guess
it's all that kind of disparate piece, right? And it's it even comes a little bit when we're talking about Tableau and and how it's evolving with composable data sources. So there's been almost like quite rigid spaces where you've got your whatever your data storage is, whether it's, you know, traditional data warehouse or snowflake, and then you've got your published data source, and then you've got your Tableau workbook.
And then you've probably got a PowerPoint presentation. And then then you've got a presentation itself and a whole lot of emails and screenshots and everything going around. And what you've showed us is really a way that it's all coming together. and some of the parts of the, with the whole semantic layer, we're pulling it all into one. We're giving it all the context with the business rules. And then we've got
We've got these dashboards and we can come into it with natural language from Slack or we can even be in the dashboard and look at natural language with Slack. It's quite a it's quite an evolution.
Steve Holley (37:19)
Yeah, I I get it now. And a hundred percent. I it's it's a massive change. I mean, I think back to the days where I was building data warehouses and with BI tools, right? And so yeah, we had to choose a database platform, all the data had to be brought into it. There we had operational data stores, we did our transformations, built our star schemas, our snowflake schemas, put indexes in place. We had to know whereabouts on the disk to put tables and stuff, like the depth of knowledge to to build stuff to make.
it happen and you're right now, yeah, the data can be left where it is. And I think one of the biggest changes for me is how unstructured data is now really important in the analysis of structured data. And we're talking before about the the term at risk and how that's shared and that might be shared in a a Slack channel, a confluence page, a a a Google doc, wherever. Well
That can actually those terms can now live in those spaces and be brought together through a single thing, a single layer. And yet one it was one of the things we announced at Tableau Conference was this we acquired a company called WAII W A I I, and it and it gives the ability to create a a living knowledge map that's updated in real time of where all this information sits and how it relates to each other, so that it all of that can be fed through to the large language model. So the different data sources, whether they're in different clouds.
databases, whether they're sitting in business intelligent tool data stores or ingested into things like Salesforce Data Cloud, doesn't matter. Yeah, it's all it all can be brought together and the result is much richer information going to the individual who's able to make real strong business decisions based on the information that's provided. So yeah, it's call out, 100%.
Sarah (39:05)
And and from my memory from from when they spoke about that product at WAII there was also ways to like certify and categorise the importance of some of those data sources as well. So you could almost give more weight to like this is a really formal piece of data that takes priority almost over others, was my understanding as well, which I think is hugely beneficial when you're looking at LLMs and you've given so much different context.
Steve Holley (39:32)
Yeah, it's pretty, it was the first I heard of it with Tableau Conference. it's been pretty amazing to start to understand it. So I guess some of the the key interesting things were it's so a lot of people are managing this thing through things like YAML and and SQL authoring, right? Whereas this is this basically auto builds and updates itself, right? It's really cool technology. it's not it's not
tied to a particular platform, right? So some of the other stuff is is stuck within the the data base database platforms. and yeah just the ability to we st we still have to give guidance I think to the agents and that's it's a really strong part of the learnings that have come through from as we've watched the agents be delivered how some of the
technologies around using the agents have focused a lot on actually the guidance we have to give agents to to make them behave in a much more consistent manner.
Fiona (40:25)
think you know some of the conversation that we're having today is really touching on and you mentioned it right at the top of the episode as well Steve the rate of change that we're going through and the great leaps forward that we're making. Are there any examples that you can share of how organisations or people are adapting really well or or tricks that they're using to help them
keep up with the rate of change or get the best out of their platforms as as we are seeing these new features being delivered.
Steve Holley (41:01)
Yeah, I think one example which came from Tableau Conference, and in fact w we they shared their experience at our recent day in Auckland, is where an organisation has gone and listened to what others are doing and and learnt from that and picked up on it and and moved it forward themselves. And I think that's a really important aspect of a lot of this. There's a lot of competition to be
first to market with using agents, to be the biggest company using agents, just to to win the agentic use cases. And in fact, I think what I'm seeing working really well is where customers are sharing what they're doing with agents so that everybody's learning. So in this situation it was David from Halter, he'd been to Tableau Conference. He'd he'd as well as meeting with product managers and attending things like
Devs on stage and what have you, he actually went to some of the customer presentations and he saw a agentic use of LLMs that he saw that could fit his business back here in New Zealand. So he was able to make contact with the person, adopt that capability, and re-implement it here in New Zealand. So that that process fast-tracked him.
to getting to where it would have taken a lot longer if he tried to solve it himself. So yeah, I I think I can see a a lot more of that happening as people start to find really great ways to use this technology that others can lean on and learn from and it and just continue to do that. And the one of the biggest things
Fiona (42:16)
Mm-hmm.
Steve Holley (42:30)
We certainly do is we really want all our customers sharing their their use cases and learnings with others, good and bad, right? Because there are not just good stories with using any technology. So let's make sure that other customers can learn from your mistakes as well as learn from your wins.
Sarah (42:47)
Mm.
Fiona (42:48)
It it's touching on such an amazing thing. I mean, both Sarah and I have had the benefits of learning from others throughout our engagements in the Tableau community and within client organisations and organisations that we actually work for in the past. And many people find it difficult to say, I need to get to that US conference to their leaders, but the
Payback and the ROI can be so powerful because, like you say, it really accelerates the speed at which you can change and adapt by learning from others. And there are some amazing sessions to attend. So, if you're listening to how do I get myself across the line and get myself across to the US,
It it can happen. It can happen on the cheap as well. So you can share rooms and do different things. You know, there's definitely creative ways to get there. But the payback that people get from being able to implement change quickly by learning from others is is huge.
Sarah (43:48)
Yeah, it's so invaluable being in a, client session. And a lot of those aren't recorded, but particularly if you can find one similar to your industry and be in there and take those notes and, be that person at the end that hangs around and asks those questions that maybe they didn't want to present, things that didn't go so well maybe, or some of the big learnings they learned on the way and establish those relationships is really powerful.
Steve Holley (44:15)
Yeah, yeah, it's there's a little a thing you can say, which is what what's the cost of not going? Yeah, because that can be more than the cost of going. But like you're both amazing ambassadors and the whole data fam community is just makes the whole event really special too, right? So
If if you're a New Zealand customer and you travel, you can be assured that myself and my colleagues from ANZ will look after you 100%. you are precious to us.
Fiona (44:44)
Absolutely. Steve, we could genuinely go on for hours about this. but before we wrap up, what's the one thing you really want our audience to walk away knowing from our conversation today?
Steve Holley (44:58)
I think the one thing is yes, have a desire to move forward and move forward quickly, but also be a little bit patient with what's happening because I was talking about this with my colleagues. Nearly every day I walk in this office
Something new has come out. it's a new technology, there's a new statement. Someone in the world has done this, that, the other. This the pace of change innovation is unbelievable, faster than I've ever seen before. And you can think, I've got to do that now, or I've got to do this now. And I think my
Advice, request, suggestion, whatever you want to call it, is to just listen to what's going on, but be patient with what you're doing and think about ultimately not the technology, but the end user experience you want to give. Because I'm seeing a lot of people get really excited by the technology and completely forgetting about the end user experience. So enjoy the fast ride, but try and slow down and think about your end user.
Sarah (46:00)
It always comes back to what is that business problem you're trying to solve anyway.
Steve, where can people find you, connect with you, and follow what you're up to?
Steve Holley (46:11)
look I'm I'm on LinkedIn. do I post there a lot? No, not really. but look, anybody's welcome to connect with me, ask questions. but my emails
I still I am old school. I still check my email every day. So [email protected] if you wanted to use that. Reach out to Fiona and Sarah. They know how to get hold of me. I'm happy to respond to any context that come through. I'm the I'm very, very passionate about analytics and about what we're doing from a Salesforce perspective in the market and very comfortable having
strong conversations about what is the right direction that companies should take. So even if it's over a beer, happy to do that because that can sometimes be really interesting conversations.
Fiona (46:57)
Love that. we will make sure that we put your LinkedIn in our show notes today. Steve, thank you so much for accepting the voluntell and coming on our podcast today.
Steve Holley (47:10)
No, look, it's been fun. Thank you for inviting me. And I really have enjoyed it. And I again I really do appreciate what you're both doing for the community and for the Data Fam and for for Tableau as a whole. So I'm very grateful for having been invited on. I'm gonna stick with invited and yeah, I look I look forward to having more conversations whether it's on screen or off.
Fiona (47:31)
Awesome.
Sarah (47:31)
Steve, it's
been a pleasure having you here today, too. If you loved this conversation, don't forget to hit follow and leave us a review and share this episode with a leader or data enthusiast you think would get just as much out of this as we did. Till next time, stay unscripted, uncensored, and undeniably data. Bye.
