Black in Data Tableau User Group 18 Nov 2025
BLK in Data & Analytics TUG - End of Year Wrap-Up | November 2025
55 min 8,487 words transcribed
Summary
The Black in Data & Analytics TUG closes out 2025 with a community round-up and a working tour of Tableau Next. Alice McKnight covers what's new — the Iron Viz 2026 qualifiers, the Tableau Next hackathon, the DataFam Slack, and Tableau Neighborhood replacing the old forums — plus two resources of the quarter, Tableau Tim and Kirk Munroe's Data Model Masterclass and the revived live viz-feedback sessions. Chris Williams, Director of Analytic Consulting at Concentrix, then walks through a customer-intelligence solution his team built on Tableau Next during the beta: the semantic model behind it, the thirteen metrics, the analytics-agent readiness checks, and Agentforce answering questions against the data in plain English. The Q&A is the heart of it — how mature an organisation needs to be before moving, why narrow contextual AI differs from a general model, how governance works inside Data Cloud, and why training users to prompt is really training them to read your data model.
Listen to this TUG
Prefer to listen? This meeting is also an episode of the Tableau TUG recordings podcast.
Key points
- Tableau Next runs on Salesforce Data Cloud, so the semantic model — not the dashboard — is the thing you build.
- Agentforce for Analytics lets any persona, CFO to analyst, ask why a metric moved in plain English.
- It isn't Tableau Classic or Tableau Next: they target different audiences, and the same semantic model can serve both.
- Narrow AI is contextual and only sees what's in your model; a general model has no parameters and will look anywhere.
- Governance is enforceable — you can put guardrails around exactly what the AI bot is allowed to see.
- The cleaner and more precise your data model, the more you get out of it from an AI perspective.
- Train users with five or six starter questions built on your own KPIs, then let them work up to their own.
- Follow a metric and Tableau Next emails you when the values move; metrics can also be shared and explored inside Slack.
- Likes and dislikes actually train the agent — ask for a chart rather than text and it learns your preference.
- Optimising the data model took client dashboards from minutes to seconds — the step people skip when learning Tableau.
Jump to a chapter (24)
- 0:00 Welcome and the TUG co-leaders
- 2:17 What's new: Iron Viz 2026 qualifiers
- 3:51 The Tableau Next and Cloud hackathon
- 5:06 The DataFam Slack community
- 6:31 Tableau Neighborhood replaces the forums
- 8:36 Pulse and Tableau Next update monthly
- 9:39 Resource of the quarter: Data Model Masterclass
- 12:24 Live viz feedback returns
- 14:53 Introducing Chris Williams on Tableau Next
- 16:20 What Tableau Next is, and where it sits
- 17:20 Touring the customer intelligence dashboard
- 21:20 The semantic model behind it
- 22:55 Building the thirteen metrics
- 25:12 Analytics agent readiness checks
- 27:05 Talking to your data with Agentforce
- 31:43 Q&A: does it remember your questions?
- 33:54 Filtering and following a single metric
- 35:42 Sharing metrics into Slack
- 38:49 How deep the Agentforce dive goes
- 40:42 Q&A: how mature does an organisation need to be?
- 42:48 Guardrails, and narrow versus general AI
- 46:43 Q&A: connecting to external platforms
- 48:29 Q&A: training users to prompt your model
- 51:50 Closing: hiring, next meeting, staying connected
Transcript
0:01 Alice McKnight: Hello, everyone. Just wanna welcome you to the November 2025, uh, fourth quarter, um, meeting of the Black in Data and Analytics TUG. Thank you, guys, for, um, sharing, uh, the next, uh,
0:21 30 to 60 minutes with us, um, especially with such short notice. Um, we've all been, uh, the team leads have been really busy this last half of the year, but we did want to, um, uh, meet and do some end of the year wrap-up things. Um, we also wanted to do, uh, have at least one meeting on Tableau Next, and also, uh, bring people a little bit up to speed about the ways that Tableau is trying to, um, strengthening the ties of the community.
0:58 They've, um, done several things that, um, we wanted to make sure that we, um, pass that information on to you.
1:07 Um, this meeting's also recorded, so if you know someone who wanted to come but could not come or is unaware of us, um, you can have them join the Bevy, uh, group on Bevy, and they'll have access to this recording.
1:24 Okay. So Chris, let me know, uh, if there's any questions in the chat or if there's any technical difficulties, because I can't see, um, the chat. All I see are the slides, so just let me know.
1:38 Chris Williams: You got it.
1:39 Alice McKnight: All right. As always, I want to introduce our team.
1:43 Um, I'm Alice McKnight. Uh, Chris Williams is also on the phone.
1:49 Um, I believe Sekou's gonna be joining us shortly. Um, this is just our titles and where you can reach us. Uh, most of us can be, uh, reached through Twitter, X, and I can also be reached, um, on Bluesky. But if you have any questions about this group or anything Tableau related, feel free to reach out.
2:18 I just want to begin with some really quick, uh, bites about what is new in Tableau that you may have missed.
2:28 I can't let this go, uh, go by without talking about Iron Viz.
2:33 Uh, I do know that the, uh, last day of submission is November 21st.
2:38 Uh, so if you're not in the process of, of building that viz, you only have three days, uh, to get that in. But, um, Tab, uh, Iron Viz is one of the main events of Tableau Conference and, frankly, you know, one of the, one of my most favorite things.
2:57 Uh, the, uh, category for this year is food and drink.
3:02 So if you're interested in submitting something for Iron Viz, uh, please do so. Um, it's definitely a, uh… I believe the, uh, well, tagline is, "Win or learn, um, never lose." So, uh, from all the people who've been involved in it, you definitely learn a lot through the building of it.
3:27 I have, uh, served as part of the, uh, group of judges for the, uh, theater, or maybe this will be my third year, and, uh, just the quality of the work is just, um, great. So if you're interested, um, you have until November 21st.
3:52 Um, also is Tableau Hackathon.
3:55 Um, there was a recently, a recent hackathon for Tableau Next.
4:02 Uh, this next ha- uh, hackathon is both cloud and Tableau Next.
4:07 And the goal here is for, uh, teams, I believe the team limit is up to eight, is to build solutions using cloud or Tableau Next to solve, um, uh, a client problem.
4:26 Um, if you go to this page, it'll give you all of the, of the rules and also if you need a Tableau Next organization or a… I think there's a very specific, um, way that they want you to work with cloud, but all the information is here.
4:46 Um, there are cash prizes, prizes associated with this, so, um, there's definitely a incentive for people to join. Um, the deadline for this is right after the new year, January 12th. So if you're interested, get yourself in a team and, uh, join here.
5:07 Over, I would say, for the last year, uh, Salesforce has been working to kind of unify the communities of each, each of its tools, um, under the Salesforce umbrella.
5:22 So, uh, uh, on Slack, uh, the commu- the Data Fam community is global, and it can be in the, where people communicated was kind of disparate. You could be on Twitter, you could be on Bluesky, you could be on LinkedIn, but this Slack community is a way to, um, pull all those, uh, chains of communication together and get people to, uh, join in one space.
5:56 There's lots of channels here from announcements. You can ask questions, uh, submit vizs for feedback, uh- There's topics under AI, general analytics. You can ask questions about future product releases, and, uh, members of the Tableau team or members of the community will respond to you. So it is definitely taking off. So if you have not joined the Slack community, uh, just let us know, and we can give you a link so you can join.
6:31 Another way that, uh, Salesforce is joining the, uh, Tableau communities together is Tableau Neighborhood.
6:40 If you are familiar with the old Tableau forums, I believe they have just recently, um, closed those down in favor of Tableau Neighborhood. If you go to the main Salesforce page and go to, uh, Trailblazer Community, you'll see each one of these neighborhoods for each of Salesforce's, um, main tools.
7:07 So Tableau Neighborhood is, uh, replacing forums, but also enhancing what you could do in forums.
7:15 So you can ask questions, you can answer questions, you can participate in, uh, discussions, join and create specific groups. A lot of the TUGs have groups on Tableau Neighborhood, which takes you to a particular page where you can just talk to individuals who are part of that group.
7:39 Um, those groups can also, uh, add their own files and resources that are just available to members of the group.
7:48 Um, it makes things easier to search for by using hashtags, which people are generally, uh, familiar with.
7:56 And your page is personalized. Once you, uh, log into Salesforce, um, and you visit the Tableau Neighborhood, it's personalized to you. So the conversations that you have participated in or have starred the tag will appear on your page.
8:20 And all this should be visible to you if you have a Trailblazer account. I think it's gonna ask you to either to make a Trailblazer account or a Salesforce account, one of the two.
8:36 Um, one of the things I wanted to point out that I think I just recently learned, uh, is that since Tableau Pulse and Tableau Next are, you know, generally new, their updates are going to come more often than the general, uh, larger Tableau releases. So Tableau Pulse and Tableau Next updates, they occur monthly. So if you want to save the updates page, uh, for Tableau, you will see this update monthly, um, for Pulse and Tableau Next.
9:15 So if you wanna stay on top of what new things are happening with both of those products, and a lot of things are happening really, really fast, um, go visit this page and w- at least once a month so you can keep up with those updates. You'll also see, um, updates, I think, from the previous month also.
9:39 Resources of the quarter. Um, what I wanted to… One of the best things about these groups and TUGs in general is the b- the ability to share the things that, um, we find interesting that I think, uh, are useful to the group.
9:55 And, uh, for this quarter, I found two things that I personally have used, and I think, uh, have proven to be, um, really helpful to myself. I passed them on to some of my, um, colleagues and for, to some of my clients, and it's been, both of these have been really helpful.
10:16 Um, the first one is the Data Model Masterclass.
10:21 Um, this is posted on Tableau Tim's YouTube page.
10:26 Um, he did this in, uh, collaboration with Kirk Munroe. If you have been in the community for any amount of time, then you know that Kirk Munroe is just the person to go to for data modeling in Tableau.
10:43 Um, he definitely knows the m- the most, has kept up with all of the changes that has been coming down, uh, from multi-factor analysis, uh, the semantic model, uh, just down to the overview of, in general, data modeling in Tableau. It's about an hour or so long. Um, I included the timestamps that have the, um, topics that he covers. It's really well done.
11:16 Um, Tableau Tim also has a few videos about introducing the Tableau, uh, Data Model cla- uh, Data Modeling process, and that's also very good. But this, uh, Masterclass really is focused and narrows this down.
11:34 I've used some things in here to change some data models for clients. I end up doing a lot of, uh, performance tuning for clients, uh, in my work, and just some of this stuff has been so helpful to, like, bring things down that used to take minutes for, to run down to seconds, and it's really, um, been really helpful. I have it starred. I passed along to, um, some of my other colleagues, uh, to, uh, help them too.
12:10 So it's one of the Things that people forget when they learn Tableau is to really think about, uh, performance and understanding how much you can make things easier by, uh, optimizing your data model.
12:24 And my second resource is, uh,
12:30 I guess this would be a revival of something that used to happen, um, with the old Makeover Monday, uh, crew.
12:40 If you were a part of the community maybe, I don't know, two, three years ago, um, Andy Kriebel, uh, who started Makeover Monday, would use…
12:51 Would come, uh, do a YouTube Live for about an hour on Wednesdays around noon, where you could submit the viz you were working on for Makeover Monday, and he would critique it live. You could come in, ask questions, or you could just listen, and it was ver- I learned a lot about, um, best practices, um, in those feedback s- uh, sessions. Um, not only from Andy, but from people who joined.
13:25 And now, uh, a little bit of that is been, um, gonna be revived. Um, the lead for this are Chimdi and Zinab. Uh, they're also, uh…
13:37 Chimdi is a visionary and Zinab is an ambassador, and they're bringing this process back. Um, if you are in the Slack and you want to submit a viz for feedback, there's a channel called community viz feedback, and when you, uh, join that and submit a post, you can opt to join the viz, uh, review live, and that will submit your, your viz. And the second part is, is that you have to post that viz to, um, Twitter with the hashtag.
14:13 Um, on the website, it'll give you… You can sign up for your slots. And what I think is really good about this is that these are gonna be recorded on the site, so you can go back and look at, uh, previous recordings and still gain, um, that knowledge that, uh, was given. I'm really glad that this is coming back. I think feedback is, um, really important.
14:39 Uh, important for growth. I learned so much, uh, in the beginning, uh, about, uh, best practices by watching these, and I'm really glad that, uh, Chimdi and Zinab are bringing this back.
14:53 Okay. So our featured presentation is from one of our own, um, our… One of our colleagues, um, Chris Williams. He is going to be talking about Tableau Next and his experiences of working with it and implementing it with his clients.
15:12 Um, this is gonna be more of a, a discussion, so feel free to ask questions in the chat, and so that we can, um, keep the conversation going.
15:25 Chris Williams: All right.
15:26 So, uh, let me go ahead and first say thank you for all, all of you for taking the time out to come out to see us tonight, um, depending on where you are. Uh, I'm in the West Coast, so it's still kind of light out here. So for those who don't know me, I'm Chris Williams. I'm one of the co-leaders of this group, Black In Tableau. I am happy to show you a little bit more about Tableau Next, uh, tonight.
15:49 Uh, this is a s- this is a presentation that we've been doing with Tableau Next for, uh, a while now. We've been a part of the beta with Concentrix. So I'm a director of customer analytics with Concentrix and, um, we've been doing this for probably about two years now, so… But I'm gonna share my screen.
16:09 And for those who've never seen, that they've never seen, say, Tableau Next before, let me give you a little bit before I actually share the screen with you about what's going on. So, you know, Tableau Next is on a Data Cloud platform, which is part of the Salesforce suite of software. Okay? So we have a back end where we created dummy, dummy data, and we were able to go ahead and create a model, a data model, which I'll show you in detail, and then obviously create a front end, which I'll call a dashboard for the moment, but it's not your traditional Tableau dashboard, meaning that you don't have a lot of the creativity of, say, like a, a CSS or something where you can put your own look and feel as far as all the nice colors and refined, rounded corners and all that stuff that you would have on a Tableau Classic dashboard.
17:05 This aspect that I'm gonna show you now talks a lot about not just developing your dashboard, but also has a lot of AI interaction with an agent called Agentforce for Analytics, which we've created on the, uh, Salesforce side.
17:19 So what you're seeing right now is a Tableau Next dashboard. So what we were able to do is we were able to create… We imported three data items of dummy data, and we were able to create KPIs from it. And from those KPIs, we were able to develop the visualizations here. And, and the funny thing about this is, is that the visualization's all pretty much created by the application itself.
17:45 So what is the reason why we did this? Customer intelligence, what is that?
17:50 We created this dashboard as a, from the stakeholder of, say, a, a chief revenue officer or chief operating officer.
17:59 They wanted to understand how well their business is doing at, you know, at any given point in the day.
18:06 And one of the KPIs, what we were able to do is interview the stakeholders to figure out which KPIs would tell that story for you.
18:14 You know, do you– what kind of opportunities do you care– information do you care about? What kind of customer sentiment do you care about? You know, top accounts and bookings and so forth. You can see all of these, um, subheadings here, and within them you have a variety of different KPIs that you have. So you have, yeah, eight within the opportunity tracking.
18:36 And one of the things that you could see as far as a common thing from, say, a Tableau Classic, which you would see a similar dashboard that I built like this, which has a, a really more refined look and feel. This is something that we've developed also off of dummy data, similar KPIs on Tableau Cloud.
18:55 And then we've taken that and pushed it into Tableau Next, which is again on the, the Data Cloud side and Salesforce side of the world. So you have a lot more AI capabilities within here, although you do have some AI capabilities within Tableau Cloud called, um, we-we'll– they have Pulse metrics, but you also have, uh, Discover, which allows you to interact a little bit with your data.
19:22 So you see these KPIs. You'll see we have total opportunities lead won. All these here will tell you how you're doing right this minute and where you potentially could be going.
19:33 Because this will talk to you about how, how you're doing with your opportunities, as well as your leads and the leads won and leads lost and your renewal rate and so forth. This gives me a– and, and also gives you a trend. It'll trend– see what happened from the prior period. It'll give you a percentage. You can see a little bit about, uh, the red and the green.
19:53 If something has improved, you'll see it in green. You'll see a plus sign here.
19:58 Um, lost bookings has decreased forty-two percent roughly. That's a great thing because now you're able to see what's happening and that number i-going down is a great thing. This may want you– this may be something you kinda wanna pay attention to. Why am I, my bookings down fifteen point three percent?
20:16 So these are certain things that, depending on the persona that you have, you wanna understand what's going on because this dashboard is meant to promote action, so prescriptive analytics. Because all the other four, uh, three types of analytics are covered here because you have your descriptive, which actually tells you what's going on.
20:36 You have your diagnostic, some of it is why it's going on.
20:41 You'll see sometimes here as you're looking at this, you'll see sort of a shaded, a light, lighter shade of blue down here as I'm pointing to it. That's because those– that's data that's been anticipated of what it's going to be, and that gives you a little bit of a predictive idea. And so it's based off of the data that you already have within your, um, your repository, and it gives a predictive idea within a range of, of confidence of what it could be going in the future.
21:10 And then obviously, what this will do is it'll drive, uh, your stakeholder to do something. In this case, our chief revenue officer will look at this and want to do something from this.
21:20 How did you get to this point? So as we developed this, we have– we've talked about models and data models, and models are so important.
21:28 The semantic model is something you probably heard or, or going to hear a lot of, which is no different than anything that you've, you've created within, say, either an Alteryx or a Tableau Prep, where you create a model where you have your multitude of data object working well and connected with one another via joins. So if I go ahead and click on the semantic model, and this is semantic model is actually the, the backbone of how any of this works.
21:55 So we have updated and uploaded three different objects.
22:01 So we have, uh, account data, we have custo– uh, opportunity, uh, line item, and then we have opportunity data. And so all of these here, there are three objects. Within those three objects we have different fields, of course. Some of these are numeric fields, and obviously these are more, um, dimensional fields. And if I were to go expand all of these, you'll have pro-approximately eighty-one, uh, seventy-seven, eighty or so different, um, or seventy or so different, um, fields that we have in here.
22:35 If you highlight over each of these, it'll tell you the joins.
22:40 So if I were to do something like this, you'll have the joins. You could do something similar to this obviously in Tableau Prep, but this is also again on the Data Cloud side. All of this is important because now you have a, a model in which all of your objects are communicating with one another.
22:56 So with these, what we built are thirteen metrics. A lot of those that you saw on a previous screen, but this is how they were built. If I were to come in here, it could go into exactly what each of these are. So this is coming up with, uh, the metric, and when you click on the metric itself, it will actually start from the beginning of time and give you an estimate of where it's gonna go.
23:18 So you figure out, yeah, this is book-bookings. It'll give you the API name, and we put a description in here.
23:25 The value. When you're creating these metrics, it'll tell you– you can point where the data item is that you're gonna be using for this particular measure.
23:34 You have filter if you want that advanced setup. This is where you wanna say bookings, a higher number of bookings is good. This is where you go ahead and figure this out.
23:43 Um, sum is the aggregation for this particular metric. If you had a percentage, you would probably have an average here.
23:50 And then obviously you have aggregation cumulative, and then the dimension in which we're doing everything operates around the close date.
23:56 So additional dimensions, if you wanna have your metrics sliced and diced for certain dimensions, other dimensions, then this is where you would wanna add those here, like account or roadmap status. These are some of the things and items through interviews we were able to do that. But these are also within our dataset, of course. This you can see the, the prefix here Insights.
24:19 So when you have each of these metrics, you have the ability to go ahead and create– uh, allow, uh, Salesforce to create insights for you. This will show you the trending from period to period.
24:32 You, you can turn this on or turn this off. And the reason why you would toggle some of these would because if you didn't– if you want to conserve credits, because everything is a… When you're dealing with AI, everything's a credit-driven world, so you could get charged for some of the credits. Very small amount, but you can get charged anyway.
24:51 If you don't want that, you can obviously go, come in here and toggle this off. But obviously understanding what's going on, what your trending– trend is for this particular metric is really important, so I would, you know, employ you to have those on.
25:05 So those are the four types of, um, uh, tabs, subta-subtabs you would have within the metric.
25:12 And so when you create all of your metrics, in order for your AI, uh, capability to come into effect, you would wanna go in, click on this analytic agent. Now, once again, you're within Data Cloud looking at this, and this is how– what you would do to set up your model for communication.
25:29 So one of the things that we do here is making sure we go through these four checkboxes. Uh, whenever you're… This is called your analytics agent readiness.
25:39 So when you finish your model, you have your joins corrected, correct.
25:45 You have your metrics created, and you have everything in your model that you want.
25:50 Now you're ready to go ahead and, and check your model for agent readiness.
25:56 Use your accurate field definitions, you know, data objects, calculated fields and so forth. Clear definitions and descriptions. That's also here. You wanna make sure that box is checked. This is really important because now for spaces in between your field names, you wanna make sure you cl- include quotes in here, because if you don't, it could…
26:18 If you have a field that's called count or contact on its own, and you just type it, it may be instead of doing account contact billing address as a whole, it could be looking at those individually as fields. So keep in mind what you're trying to get done.
26:35 It'll also test the relationships as well as the validating the business preferences. So if you have specific preferences that you wanna define within your particular business organization, it'll also check for that too. When you have all four of those checked off and, and it's validated appropriately, this will turn into an enable. Since we've already done it, we don't need to do that, so I'm gonna cl- cancel and close out of this window.
27:00 So this is what happens behind the scenes.
27:02 So let me go back to the dashboard.
27:06 So now if you wanna have a conversation with your dashboard, with the data inside your dashboard, you wanna click on this Agentforce icon.
27:14 And what this will do is we created this Agentforce for analytics icon. This allow you to type in questions against the data that you're looking at on the dashboard. And you don't have to get really tricky. All you can do is just say, "Describe
27:33 the trend for," in quotes, "win rate."
27:40 So what it's gonna do is look through the semantic model, and this is why you went through the, uh, agent, agent readiness for your model. It'll go through all the items that could potentially influence the metric win rate, and it'll give you an idea of what's happening.
27:58 So if I s- expand this, we see a positive thing. So one of the things when you're working with a, an, an agent is if you like or dislike, I may wanna have a, uh, an image associated with my request.
28:13 So I'm gonna, you know, take advantage of this. I'm just gonna say, no, I'm not gonna– I don't like this. Let's say, "I would like to see an image
28:28 this request." And it should remember what I've typed.
28:32 And then at that point, click submit, and then you go ahead and do it. So I'm gonna test this, and I'm gonna say, "Hey, I want you to tr-try again.
28:40 Uh, describe the trend for win rate."
28:48 You like the data, but you want a visualization about it.
28:52 So what you– what it'll do is it'll try to generate a visualization. What this does is actually behind the scenes creating a JSON response to the question that you have.
29:04 There we go.
29:05 Oops, I made it shrunk.
29:11 It gives you more detail. It talks about the high win rate, low win rate examples. Talks about all this detailed information. It's pretty neat to have all of this here.
29:21 Um, I like this, so I'm gonna go ahead and say, yes, I want this, and keep this type of, um, output for me. But what's happening behind the scenes is actually creating a JSON response, and JSON is actually creating this visualization that you see here.
29:37 Um, and it's pretty neat to be able to go ahead and do that.
29:40 And so now maybe I wanna go deeper into this. It's like maybe say, "Why was there volatility?"
29:56 And so you, you're just talking to your data. One of the cool things about this is, is that it does not matter what persona you are. You could be a CFO, CEO, or a data analyst all the way down– er, everywhere in between 'cause it allows you to go in there and have a simple conversation with your seme- your semantic model to try to figure out why a metric is the way it is.
30:21 So you can type these questions, you can communicate with these– with, um, your data object quite easily. Let me give you some reasons why here. So you can do this with any metric that you see here. Um, you can type in something totally different, and you could say, "Show me bookings
30:45 by vertical." Or let me do this by serving service line.
30:54 Do something like that. We've been talking about win rate, we've been talking about the volatility of win rate.
31:02 Totally shift gears on something else.
31:04 Maybe you wanna figure out what the heck is going on and how your bookings really are looking, and this will give you something like that.
31:14 And I'll show you a little bit more.
31:16 So think about all of this conversation, because this is not actually here on here.
31:22 So what I'm gonna do now is I'm gonna go into this and I'm gonna say, "Okay, so now I've dealt with win rate. I like win rate, but maybe I wanna follow that more." Because keep in mind, you have so many different metrics here right now, and I just kind of wanna show you some of this stuff here, and I'm gonna close this real quick.
31:44 We have-
31:44 Alice McKnight: Where is… Will you close that real quick?
31:46 Chris Williams: Yeah, sure.
31:48 I can open it back.
31:50 Yeah, go for it.
31:53 Alice McKnight: Does Tableau Next remember, uh, the questions you've asked in each session?
32:02 Chris Williams: For, for a session, yes.
32:05 Um, if you can also… You have the ability, if it's for some reason not…
32:10 you're not in line or something, or you think it forgets, you can click on here and clear the history, and essentially you're clearing the memory for what you just asked.
32:19 Alice McKnight: Okay.
32:21 Chris Williams: And-
32:21 Alice McKnight: 'Cause I was wondering if you have to… I know I would prefer to see the image-
32:26 Chris Williams: Yeah
32:27 Alice McKnight: … in every response. So now that you've asked, as long as you haven't cleared the history, it remembers?
32:34 Chris Williams: It, it tries to. It does its best.
32:35 Alice McKnight: Okay.
32:36 Chris Williams: It's not foolproof, for sure. But it's getting… It's been… I've been dealing with this, um,
32:43 this version and since beta, and it's improved quite a bit from the very beginning. Um, the memory capacity can still be somewhat challenging from time to time, but for the most part, I can get images when I ask the questions the, the way I've asked them. So in, in some capacity, you're really dealing with a lot with, you know, prompt manipulation.
33:07 So you're trying to ask the right questions to make the prompt de- develop the answer that you're looking for in the, in the detail that you're looking for it to be.
33:16 And so that's why I take advantage of a lot of the likes and dislikes, uh, to let the bot know that this is exactly what I want to see from my output when I ask a question. I don't want a whole bunch of text. I want you to show me what's going on.
33:33 And so that's really important, too.
33:36 So I think that's, uh, that part of it is really good, and, uh, I try to train the bot this way. So as you're learning the bot, it's learning you as well.
33:46 So that's also something pretty neat to be able to go ahead and, and, and really use it and have as part of your, your arsenal.
33:55 So I'm gonna minimize this. I won't close it, but what I'm gonna do is maybe you see all these KPIs and you really only care about win rate.
34:03 So what I'm gonna do is I'm gonna show this menu, and I'm gonna open this particular metric.
34:11 I'm gonna hide this real quick.
34:14 So this will allow you to go and filter into your win rate metric.
34:19 If you have… You can see the shaded here represents the confidence of which these values tend to be in. As you can see, the upper and lower are generally like, like so.
34:33 This part here is predictive. This is where it's anticipating based on the data that it has to this point, okay?
34:40 So obviously it's just dummy data. It's not up to date, so we'll probably put more dummy data so that we can continue.
34:47 But since you really pay attention to win rate, what we did is how we've created our, our, um, uh, security, is that we have a single sign on through Salesforce.
34:58 So I can choose to follow this, and what this does when I follow it, it means that I can say, "Please send me any time the values fluctuate within my win rate." And what it'll do, since this is hooked up to my work email, it'll send me an email anytime something jumps.
35:20 And so I can… I'll get this email, and then if I want to get to that dashboard and I have it on my phone or I have it on my laptop and you're not logged in here, you can actually click on it. It'll take you in here to… on your phone if you have the ability to log in through your phone, or on your laptop to be able to come right to this page.
35:39 So that's really neat to be able to do as well.
35:43 This is also inter- uh, this also can interface with Slack. So with the share button here, I can go ahead, and this is hooked up to my Slack channel.
35:52 So what I can do here is I can say, "I'm gonna share it to my colleague, Karen." And what this is doing is it's gonna share this win rate, and I'm gonna share it to Slack.
36:06 I have my Slack open. Let me switch into here.
36:11 Let's do my Concentrix sandbox.
36:15 I sent it to Karen.
36:19 So when Karen looks at it, she'll be able to see this, and then she can also explore it when she gets this in Tableau Next as well.
36:26 So that's one of the things that's really kind of undersold, is that if you, if you're a Slack shop, you'll be able to go ahead and interact this with your metrics in Tableau Next within Slack as well.
36:39 Teams is coming.
36:40 Uh, it's just not quite here yet So keep in mind that, you know, that, that just promotes a lot of this interaction.
36:47 So even from the Slack environment, I'll keep it here for a minute.
36:52 From the Slack environment, you can actually click into here and just say, hey, um, you know, you can type in something here, and this is something that, that, uh, Salesforce is still working on. Uh, I can just say, hey, you know…
37:08 Let me start a new one, and I can say, "Here are some past threads." I can also come in here and I could say, "New thread." I could say– I could do the same thing, uh, show me
37:22 bookings by service line.
37:28 I can type it here.
37:32 See what it does.
37:37 It should be going right to my model that I already have configured. Well, it's probably not hooked up yet. So what I'll have to do is give it the context, and that's one of the things that they're still working on, but at some point that will also be working as well.
37:50 So you can go ahead and type stuff in here. You can do by roadmap status. You could say, I could do forecast and I could do done. I can hit Apply. It'll give me a new view on this.
38:06 I can check this out and I could do stretch in the upside, hit Apply.
38:13 You may not have anything, which is fine. I'm gonna go back and just do uncheck, Done.
38:19 Then maybe I'll just– Maybe I just care about a specific…
38:26 Let me choose something that's a little bit easier.
38:32 Let me do something like travel. We'll do something simple, then I'll hit Apply. Maybe there's something here. Yeah. So you can do some sleuthing in here too. So essentially you could just use this to answer a lot of different questions. But the cool part about this is, is that you can really go into the Agentforce and really dive deep into this. You can dive so deep that if you actually have Salesforce CRM, which we do, but you can use your Agentforce or analytics bot to dive straight into an opportunity.
39:07 And so, you know, Salesforce right now has kind of limited
39:16 visualization skills within CRMA, which is the, you know, the analytics tool for Salesforce right now, and they wanted to incorporate the strength of Tableau into this. So this is the closest thing that they could get to dashboarding right now and in, uh, on the Data Cloud side. Now, one of the cool things also about this is that the interaction that you're having with Agentforce on the Data Cloud side, you're also going to have on the Tableau side.
39:45 It's just not quite there yet. There's a little bit of a delta, but, you know, Pulse Discover is out where it gives you some of that, but it's not quite there yet. So you're gonna– Or in the new releases that are coming up, uh, that delta will be shrunk so much so that you'll be able to do exactly the kind of things that I'm gonna be doing he- uh, done here on the Tableau Classic side of the cloud.
40:10 Additionally, you can, um, use the same semantic model that I use on Data Cloud within Tableau Classic here.
40:21 So instead of having two separate models, you can use one and then be able to do the same things with it.
40:28 So that's a lot, that's a lot to unpack. So I kind of want to like sit back and, and see what's going on. Um, anybody have any questions as I fade into the shadows?
40:42 Alice McKnight: I just have a, a couple of questions just because, uh, you know, I'm in consulting and right now I don't think we've seen any clients actually, um, wanna make the jump to Tableau Next-
41:00 Chris Williams: Mm-hmm
41:01 Alice McKnight: … yet. Um, like
41:06 what would be the…
41:11 I guess how mature was this or- organization before moving from Tableau to Tableau Next-
41:17 Chris Williams: Um, you could-
41:18 Alice McKnight: … data-wise?
41:20 Chris Williams: So data-wise, um, if you have– I would target, if you're a Data Cloud user right now or if you really are interested in keeping data on the data side, um, then not necessarily interact with Tableau Classic, so to speak. It's really targeting two different audiences in, in a way. So if you have Tableau Classic, Tableau Pulse is going to get there and, and, and they'll have the ability to go ahead and talk to your data, kind of like what I just showed you here.
41:50 It's just not quite there yet with the, with the interaction with your, your end usage. But if you're on Tableau, if you're on Data Cloud now, Salesforce Data Cloud, and you've been mired in CRMA where it's not great with your visualizations, uh, this would be the natural progression.
42:10 And with the AI capabilities that Tableau Next offers, it's going to give you a hell of a lot more bandwidth to explore your data rather than just try to create projects to mine your data. It doesn't mean it's gonna be replacing what people do. It's not that at all. What it does is it actually helps you get a jump. It's almost think of it in a way of allowing your agent to communicate with your model to see if you're missing anything or what it can see.
42:41 It may be able to see something you haven't thought of. And those are some of the things that are, are pretty, you know, cool to be able to look at It can start another way to really protect your data too. I mean, you can put guardrails around the data that you want your AI bot to take a look at.
42:59 So that's pretty neat too, because the security within Data Cloud is pretty pristine when it comes to actually limiting what people can and can't see. Because if you…
43:08 You know, when you're dealing with AI, you have two different types. You have narrow and you have general.
43:14 This is an example of narrow because it's contextual. Your AI is only going to look at what's in your model, whereas if you were to go, say, to, like, OpenAI, which was down today, but it's back up, I think. Um, if you were to do OpenAI, if you ask it a question, it's gonna go anywhere it can find your answer for you. There are no parameters.
43:34 So you wanna be careful, clearly, what you put out there, but you also wanna make sure that you get your answer, too. So one of the things that you can do as far as, uh, you know, governance… Governance becomes very important when you get into, you know, what you're doing within, you know, any sort of AI capability. But within Data Cloud, you can narrow that down completely to protect yourself.
43:59 But the use case that we use to go in there is that we had, uh, enough data within, you know, inside and outside of Data Cloud, and we wanted to understand if we could build a solution and then be able to promote why we decided to use one over, over the other, and I was like, "Well, we can use both." If you wanna have a situation where you wanna traditionally build your, your Tableau dashboard with Pulse-like experiences, then I would wanna use Tableau, um, Classic, whether it be Server or Cloud, because that'll give me the ability and creativity to create the look and feel and the branding that we're all used to doing within the Tableau environment, take advantage of using extensions, some of the other techniques that you would normally use within Tableau to be able to get your complete solution with some ability, with AI ability down the road.
44:52 But you also have those people who are, you know, complete Salesforce users, and they never have seen Tableau before, and they may want to have that sort of m- intermediate jump to understand how to create a dashboard. This is a n- natural way to go ahead and do that and get an enhanced version of your AI capabilities within this environment and give you exactly what you need to get a win-win situation now.
45:19 So it's not a one or the other. It could be both. You really could be… It's just a matter of which audience do you really fall into when it comes to trying to figure out what this is.
45:32 Alice McKnight: Okay.
45:34 Okay.
45:37 That actually does clear a lot of, uh, a lot of things up for me. W- I don't think any of my clients have been Salesforce users, so.
45:46 Chris Williams: Mm-hmm.
45:49 Yeah. All this is coming for Tableau Classic. It's d- it's demanding. I mean, now we have a API that you can run on the server to get you some Pulse stuff now.
45:59 Alice McKnight: Mm-hmm.
45:59 Chris Williams: And that's only going to get bigger, and that request is gonna get bigger as people really understand some of the strengths of Pulse within the Tableau, um, the, within the Tableau Cloud environment. So stay tuned. Things are coming, and as we keep talking to them, uh, them meaning Salesforce, the product's gonna get better, as we're in pretty much in the AI arms race right now.
46:25 So…
46:26 And the r- the AI arms race is a great thing because everybody's gotta try and outdo the, do each other, and that means the consumers win.
46:34 We're gonna get the best of everybody.
46:36 So I think we're gonna get the best of Salesforce as well, and, you know, carry on with that.
46:43 So how do you create… How do you connect to all the external platforms?
46:47 Um, so in this case, I created, uh… There's an agent that's built into Data Cloud that I can configure and create to communicate with all of my data source. So any s- you know, Data Cloud as well as Tableau, and this is gonna be a blanket answer to this question, Lucy, um, is it can connect to any external data source that you want, whether it be a, a database, a flat file.
47:16 You could do API downloads. You could do JSON, whatever. It can read… Data Cloud can read it. They have various connectors. Tableau Classic can read it because it has various connectors. So you can connect to any kind of data source that you really want. Um, there are a multitude of different ways, um, whether it be cloud, whether it be hard files, whether it be flat files, whether it be, you know, random database, old ass databases, anything you want.
47:47 It can connect to pretty much anything that you want to. Um, and then ultimately what you're gonna be doing is taking those data sources, or data source or data sources, and create a model.
48:00 And then at that point you can build whatever you need to in order to, to get whatever you need to build the visualization.
48:07 An a- agentic AI can communicate to whatever model that you have.
48:12 It's just an… It's all a matter of the AI bot knowing what your model is and preparing that model for it to be ingested by and used by AI. That's, that's the only, you know, step that you really have to care about.
48:27 I hope I answered your question.
48:29 Alice McKnight: And Chris, I have, like, one final question just about, just on training of the, of your users-
48:37 Chris Williams: Mm-hmm
48:38 Alice McKnight: … uh, for this. Was there any kind of, like, formal training, uh, for users on how to engage with the agent? I know we've had, uh, conversations just internally in my job about teaching people how to prompt cor- correctly, uh- How to ask questions to the agent in a way to get the answer that you want. And, and some of that is an art
49:07 Chris Williams: It, it is an art to a point. Uh, but it's also a situation where a lot of this really revolves around your model, the data model that you use for any dash- for your dashboard, and for anything. And this, this is where I don't want you to have amnesia as far as what we're all used to doing within the Tableau environment to begin with.
49:30 Everything revolves around how clear and how c- how precise your model is be- is when it's built.
49:37 Um, the more robust or more… Not even robust, but the more clean it is, the more you're gonna get out of it from an AI perspective. So it's just a matter of if you have a… I have a simple model. I have three objects that I connected, and they were all inner joins.
49:54 Um, and I was able to get, you know, close to 70 fields, tw- 13 KPIs, an endless amount of questions that go deep into those KPIs.
50:06 What you train your individuals on is what's in the model in an English sense, because they're not gonna know some of the, the business terms, so it's up to us to make sure that that's apparent. And then it's all prompt manipulation, and understanding how to ask that question. So what we try to do is I, I try to get them in a situation where I paint them a little square of what kind of questions, like five or six different questions based off of the KPIs that we have listed, to allow them to go ahead and get started off the ground.
50:41 And then I said, when you get comfortable with these questions and the answers that you're getting from these questions, you're understanding how to manipulate your own questions. And you start asking, instead of those, just those primary questions, you start asking those secondary questions. Then you get to the tertiary questions, where they're more specific, and then you're starting to get more comfortable with your model.
51:05 It's all a matter of training your end users, no matter what level they are at, whether it be technical level or business acumen level, is you're training the user how to read your model and how to ask questions against your model so it can be helpful to him or them.
51:25 So that's what our training was revol- uh, revolved around.
51:29 Alice McKnight: Yeah. Sounds good. Do we have… We have, like, uh, five more minutes left. Do we have any more questions for Chris?
51:38 Chris Williams: Forgive my shadows. I mean, I'm getting the afternoon sun now, so it's, you're starting to see the light come in here, so I fight with this right around 3:00.
51:50 Alice McKnight: Well, thank you, Chris. Let me go through, uh, quickly some closing slides really quick.
51:58 Um, as always, we ask, uh, does anybody know who's hiring? If you are, uh, choosing to join the Slack group or…
52:13 Let me fast-forward through this. To our LinkedIn group, um, let us know if y- if you're hiring, if you're looking. We do post things there. Uh,
52:28 if you're just looking to stay connected, network, keep, uh, keep, uh, keep in the loop of when we will have our next w- our next meeting, we generally have meetings once a quarter, so we'll have another one in Q1 of 2026.
52:46 If I was guessing, I would say, um, sometime in February.
52:51 But if you want to stay in contact with us, um, you should be able to scan either one of those three codes to, um,
53:02 get in contact with us. And I see that Sekou did add the link to the Slack, so join. I, I'll encourage everyone to join that Slack.
53:13 Join, uh, Tableau Neighborhood. Um, Tableau's doing a lot of work of keeping people connected, in the know.
53:23 Um, we are not the only TUG. There are TUGs dedicated to industry, from healthcare to government to education to, uh, city-based TUGs to, um, creativity TUGs, Tableau, uh, public TUGs. There's TUGs for basically everything. So if you have other interests, please take a look at Bevy so to see, um, which other groups may interest you. Um, and you can always see those different, uh, TUGs.
53:57 Most of them are recorded, and you can watch those recordings also.
54:01 Um, I did want to go back.
54:05 If you have any suggestions about what you would like to see us talk about, bring to our next TUG, or if you're interested in speaking, um, please let us know. Um, we're always looking for people to present or ideas to, um, add to our next session, so please let us know.
54:28 Um, I just want to say thank you guys for joining us.
54:31 Um, happy to see you. Gl- uh, thank you all of you joining for the la- you know, at, uh, and this, I think we got this out in the last two weeks or so, so thank you for, um, joining us really quickly and spending the last hour with us.
54:47 Chris Williams: Thanks everyone. I hope we all have a great holiday.
54:51 Alice McKnight: See you again in first quarter.
Shorts for this TUG
- Iron Viz: win or learn, never lose 2:28
- Cut Tableau load times from minutes to seconds 11:39
- Tableau Next: anyone can question the data 29:59
- Train the AI bot and it learns you back 33:16
- Follow a Tableau Next metric, get an email 34:58
- See your Tableau Next metrics inside Slack 36:26
- Narrow AI vs general AI, explained 43:08
- Tableau Classic or Tableau Next? Both 44:52
- The AI arms race means consumers win 46:20
- Your data model decides what AI can do 49:22
- How to train users to prompt your model 50:24