Tableau Einstein 1 Co-Pilot Detailed Breakdown | Dreamforce 23

0:00Hey, it's Tim here. In this video, I've

0:01taken yet another segment from my Dream

0:03force keynote

0:04breakdown. In this one, we take a look at

0:06Einstein Copilot. This is a tool being used

0:09to help

0:10analysts build and explore data much, much

0:12faster. It's the first use of AI inside of

0:15the

0:15Tableau product itself that's helping

0:17analysts build dashboards, figure out

0:19calculations. I've

0:20broken it down in the past, but in this

0:22video, we take a look at what was demoed at

0:24Dreamforce.

0:24It's changed a little bit, so if you want

0:26to know a lot more detail, this is your

0:28video.

0:28Thanks for watching, and as ever, let's get

0:31stuck in.

0:32I use it now every day, and it's really

0:32made my daily experience easier. I get all

0:34of the

0:37information at a glance, and I can take

0:37action on it really, really quickly. But

0:39you know,

0:42when you want to explore data, this is

0:42really... Again, this is another sort of

0:44framing,

0:48then use the build dashboards. Now you ask

0:48questions. And again, this is really

0:50important.

0:53I cannot... Like, if you build dashboards

0:55today, if you're becoming a data analyst

0:57today,

1:00and you think learning how to build dash

1:02boards will be the way that Tableau is

1:05heading,

1:06I think it's changing. I think you need to

1:07become more of a data engineer, data

1:09modeling expert,

1:10metadata expert in order to enable people

1:13to do what Tableau is talking about now,

1:16how to answer, how to frame, how to

1:18contextualize questions. And ultimately,

1:21the data is going to be

1:22treated. It's going to need to be treated.

1:24It's not just going to work out of the box.

1:26When you get, let's say, purchase data from

1:28a superstore, it doesn't come ready to do

1:31this.

1:32When you get transaction data from your

1:34bank or from an online store, it does not

1:37come ready to

1:38do this. In order to get it to this place

1:40where you can actually use AI on top of it,

1:42you're going

1:43to need to clean it, prep it, put it into a

1:45data model, warehouse it, make it available

1:48, think about

1:48security, think about all these different

1:51things. That is where being a data analyst

1:53is going to be

1:54heading to. And that's where being a data

1:56engineer or being a data modeler will sort

1:58of pay dividends.

1:59I think the people who build dashboards

2:00today are just going to move further back

2:02into the stack

2:03and do more things to enable these kinds of

2:05experiences. So everyday people can just go

2:07and ask questions. >>

2:09See where the Tableau superpowers come in.

2:12Tableau is easy to use. You can easily

2:14slice

2:15and dice your data however way you want.

2:17But normally when you start in Tableau,

2:20you got a blank screen and some data, and

2:23you have to kind of learn the product. You

2:25have to

2:25know how to drag and drop, where the

2:27features are. You need a visionary around

2:30you. You need

2:30the documentation. You need experts to help

2:33you get successful. But what if we brought

2:37AI into

2:38the experience? What if you had Einstein

2:40with you that understood Tableau, could

2:42help you answer

2:43your questions more easily? Well, the

2:46future of Tableau is to have AI embedded in

2:49the exploration

2:50experience, where AI fully understands how

2:54to use the product, where AI understands

2:57the meaning of

2:58your questions and can help essentially do

3:00the drag and drop for you. So it's not

3:03about either

3:04or, it's an and. It's augmenting the

3:06experience, making it 10 to 100 times

3:10easier and making all

3:12of you more productive. That is the goal.

3:16And so today I'm pleased to announce the

3:18Einstein

3:19Co-Pilot for Tableau. - So, Einstein Co-P

3:25ilot, June 24th, that's almost a year away,

3:28just under

3:29a year away. This is going to be

3:32sensational. I think this is why it has

3:35such a long run-up.

3:37So much has to happen. And I think there's

3:38some fundamental questions that Tableau

3:40have to answer

3:41before this kind of tool gets deployed. And

3:44yes, you guessed it, it'll probably be

3:46cloud first.

3:47You won't get this for Tableau Server, no

3:49way. Tableau Server's probably got a 25

3:52release,

3:52if that makes sense. I just, the more I

3:55think about sort of Salesforce being a SaaS

3:57company,

3:58the more I think about where Tableau is

4:00heading, I just cannot see how some of the

4:03ideas they're

4:05thinking about here come to Tableau Server

4:07in an expedient way. Because like I said

4:10before,

4:11and for these things to work on Tableau

4:13Server, the requirements on infrastructure

4:16are just going

4:16to keep going up and up and up until the

4:19cost of doing those things kind of pushes

4:22you to the

4:22cloud, honestly. I've not sort of been in

4:25touch with Tableau Server for the last

4:28couple of years

4:29now, because I just haven't needed to use

4:31it as much. Tableau Cloud has been the

4:32predominant

4:33side where clients are working. So actually

4:36knowing how to manage the backend and

4:38infrastructure

4:39of that is almost sort of non-existent,

4:42because all you have to do is go into

4:44online.tableau.com

4:46and manage the front-end user face there.

4:48So I'd be really intrigued to know what are

4:51the

4:52server requirements, what is the server

4:54usage of a server today, and what are the

4:57features that we

4:58get in cloud that aren't available yet that

5:00would sort of increase that and as Tableau

5:02roll out

5:02these AI features, how is that going to

5:04play out long-term? Anyway, interesting,

5:07interesting,

5:07interesting sort of thing to see tie out.

5:09Oh, yes.

5:11The Einstein Copilot is really going to be

5:16a core part of the Tableau experience that

5:20enables you

5:21to ask questions of your data and it will

5:23basically explore it for you. It'll give

5:26you better results

5:28because it understands a lot of the context

5:31. You'll have better best practices built in

5:33,

5:33and ultimately you'll just be more

5:35successful. You'll be able to ask more

5:38questions and drive

5:38more value to your organization, or you can

5:41just put your feet back up and enjoy the

5:44rest of your

5:45day. So let's see Einstein Copilot in

5:47action. So for that, please welcome Honto

5:50May. Honto.

5:52Thanks Francois. In the next four minutes,

5:57I'm going to show you how Einstein Copilot,

5:59backed by the Einstein trust layer, can

6:01speed up and improve the quality of your

6:03data analysis.

6:04Here in Tableau prep, I have customer

6:07purchase data for nationwide chains.

6:09I want to use this data to create

6:12personalized experiences for my customers

6:16that'll drive

6:16incremental revenue. To do so, I need to

6:19know where my customers are, so I need

6:20their postal

6:21codes. Unfortunately, my postal codes are

6:24trapped in this customer mailing address

6:26column.

6:27Normally, I'd have to figure out how to

6:29write a calculation to extract this,

6:31but with Einstein, all I need to do is ask

6:35in natural language.

6:37And on the fly, Einstein is able to create

6:43this calculation for us. Now, all I need to

6:48do...

6:49Now, I think people have seen this demo

6:51before, which is why the crowd didn't

6:53have a big reaction. A lot of the people at

6:55this conference were at Tableau conference,

6:58and so

6:58they've seen this demo, they've seen this

7:00example before. That said, it doesn't take

7:02away from the,

7:03let's say, awesomeness of this, because

7:06this is why I believe AI is fundamentally

7:09going to change

7:09the way analytics is done. You see

7:11previously, if you just look for this

7:13problem, let's say you're

7:15a data analyst and you're not an

7:17experienced data analyst. You've been

7:19working in the field maybe a

7:20year or two years, okay? And this is the

7:22data that you get, and someone asked you, "

7:24Hey, how do you

7:25extract the postcode from this column?"

7:29Your first instinct might be to pass out

7:33the commas to get

7:35the final field of each column, which will

7:38still leave you with Florida 322-444-USA.

7:42And then the

7:42next thing you might do is to say, "Okay,

7:45if you find any one of these states, go

7:48ahead and remove

7:49that which will leave you with 99210 and

7:52then USA." But you see, that's not always

7:54consistent. You

7:55see North Carolina down here is 28405 and

7:59doesn't have a country on the end. So

8:02sometimes it's USA,

8:03sometimes it's US, and you get into this

8:05really messy world where if you really have

8:07to clean

8:08this, you kind of use brute force method

8:10and you apply like a really convoluted way

8:12of passing this

8:13in steps, maybe 15, 16 steps to get to

8:16where you need to, and then you kind of go

8:18from there. Well,

8:19the experienced analysts will be able to

8:21look at this text and say, "Hmm, there's a

8:23pattern here."

8:24The postcode is essentially a certain

8:27number of digits followed by letters

8:29essentially, right? And

8:30so we can actually go and find that in the

8:33string by just looking for that using that

8:36pattern. And

8:37the technology that helps you do that is

8:38called regex. Now you wouldn't know the

8:40term regex,

8:41you wouldn't even, you might stumble across

8:43if you let's go into, let's say you go into

8:45Reddit

8:45and you ask a question, you wait a few days

8:47, someone replies, or you Google, you come

8:49across

8:49this thing and it's called regex. Then you

8:51go to regex 101 and you start trying to use

8:53it and you're

8:53like, "Okay, this is interesting." You go

8:55down a rabbit hole, 30 minutes later, you

8:57're now figuring

8:58out how to write regex for this. And it

9:00works, but it doesn't work some of the time

9:01. Other times,

9:02you go test it. And so you're not so

9:04confident. So you try this thing and you

9:06kind of move on.

9:07That whole flow probably has taken 40

9:09minutes, maybe 30 minutes. If you're super

9:11fast,

9:12you know what to search and you're kind of

9:13adept and you're kind of really going down

9:14this route

9:15of exploratory sort of data analysis. That

9:18said, the simple fact that you can just go

9:20in and type

9:21the question, say, "I want the postcodes."

9:23You don't have to know the term regex. You

9:25don't have

9:26to know regex pattern matching. You don't

9:28even have to go to regex 101 or even ask

9:30the question.

9:31You can just go and type the ask and the AI

9:34tool helps you figure out what you need to

9:37know.

9:38And here's the added bit. Now that you see

9:41that term regex p extract or whatever,

9:43that should pique your interest. If you're

9:46a good day's analyst, that will pique your

9:47interest and

9:48go, "Huh, what is this?" And so you then go

9:50Google that thing and you understand what

9:52it is. And now

9:53AI hasn't just solved the problem. It's

9:55also given you a shortcut directly to the

9:58thing you need to

9:58learn and the way it's working in order to

10:01enhance that. So now you kind of start to

10:03use AI as a way

10:05of discovering things you need to learn as

10:06well as a way of helping you, which is sort

10:08of a double

10:08edged thing. The next time you come to this

10:11, you'll ask specifically, "Hey, can you use

10:13regex

10:14to solve this kind of problem?" It's quite

10:16complex. And now you're having a much

10:18higher level

10:18discussion with AI. You're still using AI,

10:20but you still understand what's going on

10:22and you're

10:22building your understanding as you go along

10:24. So it's also helping you with data

10:26literacy.

10:26So I think this to me is probably the

10:29biggest opportunity that Tableau has just

10:32to help everyday

10:33data analysts who actually still do build

10:35data sets and/or data models and/or visual

10:38izations.

10:38And more importantly, it's also going to

10:41help bring the skill level up for everyone

10:44who's

10:44already doing this stuff. It's going to

10:46bring them right up so they too have access

10:47and awareness of

10:48things like LODs, all these complex terms

10:50like set actions. It's not going to solve

10:53the problem,

10:54but it might just alert you to the

10:55capabilities behind these things. Anyway,

10:57let's keep seeing the demos and see the

10:59examples.

11:00Fortunately, my post-it codes are trapped

11:02in this customer mailing address column.

11:05Normally, I'd have to figure out how to

11:04write a calculation to extract this,

11:09but with Einstein, all I need to do is ask

11:09in natural language.

11:15And on the fly, Einstein is able to create

11:20this calculation for us.

11:25Now, all I need to do is give it a name,

11:30and voila, a new column in my data with the

11:36customer post.

11:36What I would be really interested to know

11:40is what's in the reference tab here. I was

11:42just

11:43thinking about it. It's like, huh, there's

11:44a reference tab. Is the reference tab

11:46showing you

11:47what it's doing? The nice thing with

11:50websites like Regex 101 is that it shows

11:53you how it's working.

11:54What I would love Einstein Copilot to do is

11:56to almost play through an example of the

11:58calculation

11:59in the context of Tableau to show you what

12:01's happening and even show you what's going

12:03on.

12:03I've put this here. I've done this there.

12:05Almost guide you through the steps.

12:08ChatGPT can do this today. It will tell you

12:11do this, do this, do that. Obviously, it's

12:13not

12:13perfect, but if you're training a model

12:15specifically around Tableau,

12:17then actually it should be possible to be

12:19able to instruct it and give you

12:21instructions on what

12:22exactly is going on. Almost reverse write

12:24the blog post that you would write if you'd

12:26figured out how

12:27to do this. Now, all I need to do is give

12:34it a name, and voila, a new column in my

12:41data with the

12:42customer postal code. I did all of this in

12:45a matter of seconds and without writing one

12:51line

12:52of calculation code. So with Einstein, you

12:55and anybody can use Tableau prep to

12:58transform the data

12:59they need into the format they want. So how

13:02are we going to reach these customers

13:04though? Well,

13:05did you know that you can use Tableau to

13:07visually explore your audience data and to

13:10create audience

13:11segments in data cloud? No. Wrong screen.

13:19We're good? Okay, there we are. A little

13:22excitement there. So I have just connected,

13:26with the help of my friends back there, to

13:28data cloud. Now, when I'm presented with a

13:32blank slate

13:33like this, I ask myself, where the heck do

13:36I begin? But I...

13:37The really interesting thing here is that

13:40Einstein is already available on the right-

13:42hand side,

13:42and the right-hand side has become this

13:44sort of contextual place to find out more

13:46about

13:46the data set, more about what's going on,

13:48more about the metadata inside of Tableau.

13:50And it's sort of grown legs. This is also

13:52where explained data used to be. I kind of

13:55feel like

13:55that's going to get sort of pushed to the

13:57side now because you'll have our state

13:59explained it,

14:00those are all going to go away and Einstein

14:02and Tableau GPT and the Carlson metrics

14:04will sort of

14:05sit in this space more squarely ready to

14:07help you sort of answer questions and pull

14:09out insights

14:10rather than sort of forcing you to come up

14:12with the answer question yourself. The

14:15other nice thing

14:16here is obviously this is a web edit

14:18experience and this is a draft, so that

14:20means he's exclusively

14:21using the authoring experience in web edit.

14:24And last edited, September the 12th, that

14:28would have

14:28been the time of the demo. So he's using a

14:30sort of live take of this, if that makes

14:32sense. All

14:33these details do matter because I think a

14:36lot of people think about Tableau in the

14:38desktop sort of

14:39setup when in reality that's not how most

14:41of it works. That's not how most of Tableau

14:44demos

14:44anything anymore. It's all done in the

14:46browser. And so it's interesting to see

14:48that. I don't think

14:49you'll get the same experience in desktop.

14:51I just don't think that will pull through

14:53unless you're a

14:54Tableau Cloud customer. I kind of think

14:56this is going to be a really a smoother

14:58experience in the

14:59web because that's essentially where this

15:01will be running. Otherwise you can just

15:03imagine sort of

15:04the back and forth between your local

15:06client and your laptop and Tableau servers

15:08when this stuff

15:09is running. But it will still be

15:11interesting to see. Now the customer

15:13purchase history,

15:15what is not clear is if this is a data set

15:17in Salesforce and that is why Einstein Co-p

15:20ilot and

15:21this technology is working really well, or

15:24if this is going to work across

15:25non-Salesforce based data sources as well.

15:27So things like in your Snowflake, in your

15:29Databricks

15:30database, whatever those are, that detail

15:33is still not clear. I assume it will work

15:35everywhere

15:36and Tableau will be running this technology

15:38on the cloud, looking at these data sets

15:40and sort

15:40of processing them. That's how some of the

15:42past features have worked. There's

15:44something called

15:45Data Change Radar, which has essentially

15:47been taking snapshots on your server and

15:48then analyzing

15:49that on your cloud instance and then

15:51pushing you alerts when something changes

15:53that shouldn't have

15:54changed. So super interesting little nugget

15:57. Let's see how it actually works.

15:59Einstein has got you covered. Using gener

16:02ative AI and statistical analysis, Einstein

16:05is able to

16:05understand the context of your data and in

16:08doing so, Einstein is able to suggest

16:11relevant business

16:12questions to kickstart your analysis. That

16:14's really good. Let's take a look at this

16:16one about

16:17patterns of my sales over different product

16:19categories. And look, with one click and

16:24without

16:24having to drag a single pill onto a shelf,

16:27I'm able to see the viz that shows my sales

16:30for all

16:31my different product categories. But what

16:33about this pattern Einstein was looking at?

16:35I can see,

16:37yeah, that's right, outdoor sporting goods

16:39are popular in the summertime. That's not a

16:41surprise,

16:42but that gives me an idea. I know that our

16:45in-store experiences drive bigger purchases

16:49compared to online. What if we invite these

16:52outdoorsy and sporty people back into a

16:56store

16:56with an event like a outdoor pet first aid

16:59class? Well, how am I going to do that?

17:03First of all,

17:03we're going to use those postal codes we

17:05extracted earlier and you guessed it, we're

17:08going to ask

17:08Einstein. Yeah.

17:09Show me the location of customers who

17:15bought sporting goods in the last three

17:18months.

17:18By zip code. All right. I see all my

17:24customers on a map. All right. What about

17:28my stores?

17:28How far are these customers from store

17:32locations?

17:33And look, there we go. Without knowing

17:40anything about map layer.

17:41So I'd seen this before, but I was paying

17:44attention to what was going on and what

17:46changed.

17:47And the great thing about this. So see

17:54customer transactions, store locations,

17:58store locations is

18:00in which there we go. So there's two data

18:03sets in this data model. One is customer

18:08transactions,

18:08one is store locations. And essentially

18:11what they're doing is relating the customer

18:14transaction

18:14to the store level data, which gives us two

18:16spatial fields. The city or yeah, there's a

18:23customer location field. So it could be the

18:25city or the whatever of the customer, their

18:28address.

18:28And then you have the location from the

18:33store. And so to bring these two together,

18:36you are creating map layers. They're

18:38putting the two on top of each other.

18:41Because they're in the same data set, they

18:43have a data model relationship. So they

18:45should,

18:45it should naturally work nicely. You can do

18:47that without the relationship. You can just

18:50bring on

18:50your store locations at separate data

18:53source. And a new feature, nearly a year

18:56ago now,

18:57allowed you to basically overlay two

18:59separate data sets on a map without having

19:01to do any

19:02sort of join or relation to them, which is

19:03kind of powerful actually, because it

19:05allows you to

19:05bring contextual sort of map layers without

19:08having to like do the dirty work of

19:10blending it or doing

19:11whatever you need it to do to make the map

19:12work. So this is quite nice. Now, why I

19:15like this demo

19:16is because it kind of shows that iteration.

19:18It kind of shows a Tableau kind of going

19:20through

19:20the steps. And again, I believe the

19:22language is being used here is generally

19:23authentic. It's kind

19:24of what you'd ask in terms of the analysis.

19:26If you're asking good questions, that is a

19:28skill in

19:29itself, but I think this is a fair

19:31reflection of what people would actually do

19:33with AI. And it's

19:34doing what you'd expect it to do. Now, what

19:37we can't tell by this demo is how often is

19:39it good

19:39at doing this? Because you know, sometimes

19:41with AI, these things just, you know, 90%

19:43of the time,

19:43they're okay. And sorry, 90% they're good.

19:4610% of the time, they're okay. And when

19:48they fail,

19:48they fail epically, right? So this looks

19:51pretty good. It's doing a few complex

19:53things, latitude,

19:55longitude, bring it all in. There's levels

19:57of detail here that are on there on both

19:59the

20:00custom and store locations and the marks

20:02pane. There is coloring going on. You could

20:07argue,

20:07potentially there's some buffering going on

20:09. I don't know if the size of the circle,

20:12yeah, the size of the circle represents the

20:14custom account from that store. So I

20:15actually think it's

20:18sort of interesting because the customer,

20:20you know, the custom account maybe relates

20:22to like,

20:23maybe it's a specific town for these

20:25customers. And it's just showing you where

20:28those people are

20:29coming from in those towns. And that's why

20:31certain towns have bigger or smaller

20:33circles. But you have

20:35customer city, customer, I don't know,

20:37something. I can't see the location

20:40hierarchy. So again,

20:41unnecessary levels of details, unnecessary

20:44level of breakdown of the details here, but

20:47it seems pretty good. And the demo adds up.

20:51That's all I'm trying to make sure. Like,

20:54is this a file fetch demo? No, it's not.

20:56And it's probably something you'd get asked

20:58to do.

20:58And you'd be asked to put this in a

20:59dashboard. Now, the super interesting thing

21:03here is just

21:04imagine that the whole of the left hand

21:06side doesn't exist. And all you have is the

21:09Einstein

21:09column on the right, and the customer

21:11location chart. And that's all you get.

21:13What if that's

21:13the experience of Tableau going forward,

21:15right? For every person, for everyone, that

21:17becomes the

21:18experience. But they're trying it here

21:20first with data analysts, to kind of test

21:21if it's good,

21:22and if it's bad, but slowly over time, this

21:25experience where you type and you see

21:27charts

21:28will be pretty much the core experience of

21:31Tableau. What do you think? Let's carry on.

21:34Or geographic roles in Tableau, I was able

21:37to create this intuitive yet complex map

21:40viz

21:41with Einstein. But don't forget, Tableau at

21:44its heart is an interactive and visual tool

21:46.

21:46I can easily grab these customers from San

21:50Francisco and with our integration with

21:53Salesforce,

21:54send these customers up. Okay.

22:00As a audience segment. Now that is a, that

22:05's a pretty sick feature. If you say a Sales

22:07force

22:07customer, just being able to select the, do

22:09the analysis, select the customers, create

22:13a segment

22:14and push it back into Salesforce. I mean,

22:16that is, if you're a Salesforce customer

22:19and a Tableau

22:19customer, that is, that is perfect, right?

22:23Like that is the dream. Now in reality,

22:26not many people have the ability to do that

22:28or sort of trust to do that. But in reality

22:30,

22:31like if that was your flow and you could

22:33enable that and don't forget, this has not

22:35been published.

22:35This is still exploratory analysis and it's

22:38like a, it's an easy thing to forget. We're

22:39not going

22:40from dashboard to Salesforce here. We're

22:42going from a sort of detailed discovery. So

22:45an analyst

22:45has been asked to go and find this customer

22:49segment and push it to where they need to

22:52get to.

22:53As soon as that question comes back, boom,

22:54they can just go back into the chart they

22:56built without

22:56publishing anything and push it off into,

22:59into Salesforce. So I think that's also a

23:01really nice

23:01celebration of Tableau's heritage. It was

23:04always a data exploration tool and it's

23:06kind of easy to

23:07miss in this demo, right? Because you're so

23:09caught up in the feature, but actually the

23:10fact they

23:11didn't publish it before doing this, I

23:12think that's super powerful. It's super

23:14important,

23:15getting straight to the action rather than

23:16this whole, like, you know, governance and

23:18publishing

23:18thing. Like you're empowered to do this,

23:20just go ahead, push it to the Salesforce

23:22instance. And

23:23move on to the next task. Perfect. And I

23:26did all of this without leaving Tableau.

23:30Now, oop, I can't get to the screen. Oh,

23:36sorry. My bad. And now my, uh, my marketing

23:42team has

23:44all the information they need to activate

23:47the segment and send custom communications.

23:52The person who's experiencing it go wrong

23:54at the moment. It doesn't do what they're

23:56expecting.

23:57It doesn't matter if there's on point. It

23:58doesn't matter if they're about to nail the

24:00demo,

24:01nothing matters. As soon as something goes

24:03wrong, your brain just goes into like,

24:05like freeze mode. Cause you're like, what's

24:08going on here. You can't really think fast

24:10enough to

24:11kind of rescue. So for April to sort of

24:13notice that and catch that moment and get H

24:16onto back onto

24:17on track. It's like that is, that is a,

24:20that is superhero stuff. Uh, I'd like to

24:23say not all heroes

24:25have capes, but April definitely deserved

24:27on that because it's a small thing and I'm

24:29sure Honto

24:30super appreciated it at the time, but it's

24:33so easy, so easy in that instance to just

24:36freeze for

24:36like two, three minutes, fix it and off you

24:46go. This, there we go. Like this to all my

24:51customers.

24:53The thing that's really unfair here is that

24:55Honto is not controlling the transitions

24:57between,

24:58uh, the slices from phone to laptop. Honto

25:01is not controlling that. So on top of like

25:03things go

25:04wrong, uh, like whoever's controlling the

25:07PowerPoint is just not working quickly

25:09enough. So, um,

25:10it's not, it's one of those sort of comp

25:12ounding effects. It's kind of funny, but

25:14yeah, anyway,

25:15it looks pretty, pretty good. I did manage

25:19and maybe this is why we didn't have a good

25:21demo.

25:22I managed to sneak my dog into the keynote

25:25presentation.

25:32All right. Now in summary, Einstein Co-p

25:36ilot backed by the Einstein trust layer will

25:38let

25:39anybody who can ask a question visually

25:41explore their data in Tableau and with our

25:44deep integrations

25:44in the data cloud, you can get the insights

25:47you need to connect with your customers

25:49faster and

25:50back to you, Francois. Awesome. Great job,

25:53Honto.

25:53Thank you.

25:54[silence]

26:04[ Silence ]