Panel Discussion on Ai : Thriving Together in the AI Era

0:00ASTM Here Today, something a little bit different.

0:02I'm posting a recording of a recent panel that I had the privilege to host.

0:06It was a panel titled

0:08thriving together in the AI era.

0:10This panel was run off the back of a really great conference talk that was uh redone here as part of this user group.

0:17And then on top of that, we had a little bit of a demo of what's coming up in the Tableau ecosystem this year with some headlines and some timings that I think are going to be super important.

0:27You should be seeing them on screen

0:29Um the really key thing is that we talk a lot about AI and sometimes we don't get a chance to do it as a community.

0:35And so this panel was exactly that.

0:37We had five incredibly talented individuals from the community coming together from different walks of life to discuss what we'd just seen and also talk about some of our concerns, challenges and opportunities

0:49that AI has to offer.

0:51And so I posted it here on my channel to give it a little bit more exposure.

0:54And another thing to call out is that this panel is pretty much the same feel

0:58and uh kind of vibe you get from a tableau user group.

1:01And so that's actually how uh many people attended this session.

1:04We had roughly five hundred people attend

1:06And many user groups around the world have that kind of audience when they're online.

1:10And many others are in person and I highly encourage you to go to those because you get to build your network.

1:15So uh another call out to just really engage with Tableau user groups, they're super valuable

1:20I've learnt so much from the user groups that I've both presented at and been part of.

1:24So highly encourage that you do that.

1:26You should be seeing the link on screen

1:28Enjoy the session.

1:29I think there's so much everyone can learn.

1:31I've I've done a a decent job putting the timestamps so you can pull out the themes that you care about and see how the panel answers the question

1:37Everyone had so much to share.

1:38It was my turn to sit and listen.

1:40Um I hope you enjoyed this video.

1:41I'll catch you in the next one, but for now, enjoy the panel.

1:44Amazing.

1:45Thank you.

1:45Thanks for thanks for having me.

1:46And let's um sort of do a round warm welcome to the rest of the panel.

1:50I want to just briefly open it.

1:52first by I think highlighting something really important.

1:55My job here is to moderate this this panel, but I think the thing we sometimes forget is that when we have an opportunity to learn something

2:02I think learning is best done in a community.

2:04And so we've just seen a sort of a a series of amazing talks from really incredible people.

2:09We've seen the product, we've seen a little bit about AI.

2:12Um what this panel is really about is talking about sort of what we're all thinking and sort of processing some of that and I think this is a incredible opportunity to do that.

2:21And so I'm joined uh by some great individuals.

2:24Um I'm joined by Joy, um who's a data analyst at Data EdX Group.

2:29She's also Tableau Ambassador for the uh Tableau AI Tug.

2:32Uh Joy, um, are you with us?

2:37Hi everyone, I'm so excited to be here and thank you for the introduction team.

2:42You're welcome, welcome.

2:43Next up we have Will, who's an executive director

2:45for Finance, Data and Insights at JP Morgan Chase and also Career Pathmakers User Group Leader.

2:51Will, how are you doing?

2:53Doing great.

2:55Good.

2:55Good.

2:57Next up we have Maya from the Data Leadership Collaborative, your advisory board member and CEO of Savvy AI.

3:05Hi everyone, so happy to be here.

3:08And last but not least, we've got Rain who runs the user group program here at Tableau.

3:14Rain, how are you?

3:16Hi, I'm wonderful.

3:17There are 518 people on this call right now.

3:21That is incredible.

3:22Represent DataFam.

3:24Absolutely.

3:25It's uh it's a great turnout.

3:26I really love it.

3:27Um I think we've got a a great lineup today on the agenda, so everyone's turned up in full force.

3:31Um

3:32The the first thing I want to start with is like what's running through everyone's mind um having seen the talk from Natalie but also the preview from Tableau and the product.

3:40Um maybe Maya if I start with you first

3:44I loved the talk from Natalie, especially around empathy, because we are in a time of tremendous change.

3:52And when we talk about technology, it's almost seeming like everyday technology is becoming easier and easier in certain ways.

3:59But getting folks to embrace the technology, understanding our role in a world with this new technology, that requires a lot of empathy and understanding.

4:08and just a lot of humanity in very many ways in the face of AI.

4:13Absolutely.

4:14Will, you're up next.

4:16Um so I I really focused on trust.

4:20If if I was playing like bingo, um the trust would be like the number one key that I that was quick and easy.

4:27Um and being data people

4:30Seeing how do we understand and trust what's behind the scenes is is really where my brain continues to just like focus on.

4:39Um so that that's really one of the things that I just took out of that.

4:43Thank you.

4:44And Joy, how about you?

4:47My favorite thing was looking at the different perspectives people had on a particular topic.

4:54So

4:54Someone came from a more humanities standpoint where she talked about um the effects of these things, trust.

5:02Um

5:03the um our world, how it's affecting um climate and everything.

5:08And that was very interesting.

5:09And we also got to see from like the product aspect.

5:11So the fact that we could see

5:13different melting points of the same topic was very interesting for me.

5:19Amazing.

5:20And Rain?

5:22I think I got a lot of reassurance from that talk.

5:25So I'm gonna acknowledge right now that technology is moving at the speed of light right now.

5:31And I think it's giving us a lot of anxiety as data analysts and just data people about what our role is going to be in the future and if we have a role in that future, right?

5:42But from everything we saw

5:45From both of these speakers, there is absolutely a human element to be had.

5:50We are essential to this equation

5:52And so we're never going to not be part of it.

5:55It's just figuring out what that niche is and allowing these tools to enhance our everyday.

6:03Yeah, it's incredible.

6:04And I think um you've all you've all done me a favor because you've made it really easy for me to pull out um I think some recurring themes throughout today.

6:11Um the first one is obviously embracing change

6:14Um this technology is moving incredibly fast.

6:16The headlines yesterday are different to today, right?

6:18Um the next one is um on that point of trust from will.

6:22You know, that what we're really talking about there is how do we implement these things in a way that we feel that we can control.

6:28So innovation and implementation I think is

6:30is sort of a good headline um for that.

6:32And then the very last thing is um obviously we are people, we are human.

6:36Um I think that came out through Natalie's talk.

6:39throughout today.

6:40How do we do this in a collaborative and a communal way?

6:43So those to me feel like sort of the three important themes.

6:45And I think I'll I'll use that as a sort of a headline of how we can approach discussion.

6:49And I know you've all got

6:52sort of different perspectives and uh deep dives that you've all um sort of talked to me ahead of time.

6:56So let's let's let's get into that a little bit.

6:59If we open up with the first point around embracing change.

7:03Maya, if I could sort of start with you again.

7:06Have you seen examples where organizations have successfully navigated this transition, sort of how they've started to think about AI?

7:14I would say the most successful organizations I've seen are those that include stakeholders from various departments, not just

7:23technology stakeholders because ultimately you're building this technology for a set of folks individually and so it's important to include them in the conversation or for maybe a set of external users

7:35it's important to collaborate with them, to ask them, are we really solving your problem or are we just slapping some technology at it and hoping that you know you will be distracted by the shiny object?

7:47Also, the organizations that I see most successful with AI are the ones that are educating their teams, getting their teams to have their hands on various AI tools.

7:57and empowering them to upskill with those tools versus saying and and this is a phrase I really get so sick of hearing is like you won't be replaced with AI somebody using AI will replace you

8:09And I'm so tired of that because that's such a trite way of explaining a really complicated concept.

8:14All of us have encountered AI every single day, every time we use Google Maps or

8:20Every time, you know, we have autocomplete, we've encountered some sort of AI or received an offer in the email that was related to a product we just bought.

8:28We are already much more AI fluent than we are giving ourselves credit for.

8:34And you don't have to be a data scientist or you don't have to have these advanced data skills.

8:38You can upskill uh what where you're already starting from.

8:43Absolutely.

8:44And I I think um just to sort of follow on from that, I think AI has become this tagline for a bunch of disciplines that were previously sort of given their own space.

8:52Like

8:53machine learning stats, all of this stuff has just been bundled into AI.

8:56And it's uh it's a really good point you bring up because actually we've we've all been doing this for quite some time.

9:00It's not not that radically new.

9:03And and just to sort of build on that, um

9:06Uh how how Joy, uh if I come to you as as a data analyst who's sort of working with this technology, how how do you see sort of the challenges, someone who's um you know navigating this field as someone who's learning and building and obviously you've got the um AI user group

9:20What have you seen in that user group that sort of um speak to to what Maya's talked about there, how we how we're using AI?

9:27Yeah, I think um what Maya said about upskilling with the tools, that's what um

9:33our Tableau User Group is about.

9:34So the AI and Tableau User Group basically brings everyone to a community.

9:40It's a safe space for us to talk about these tools and how AI is either changing it or affecting it.

9:47So now we have leaders from the Tableau core team itself.

9:51So we have product managers, researchers.

9:53We see people who are actually building these tools come to talk to us.

9:56every month or quarterly and um at our events you also get to see people who use these tools.

10:03So for example you see individuals, data analysts, data scientists.

10:06And they all come and show show people, okay, this is how I use this tool in my day-to-day activities, this is how it has made me more productive, and this is how you can do it too.

10:15And we also have conversations from pessimists and optimists, people who are scared about the tool, and we have people who are optimistic about the tool.

10:24And my favorite thing is bringing the core team on the hot seats.

10:29That is absolutely my favorite.

10:30So we get to ask very, very difficult questions.

10:33So we're asking them, if we keep using these tools, who is going to use um, where is our data going?

10:39And we keep asking, okay, the new research that's coming, what is this tool going to do?

10:43What is that tool going to do?

10:45And how can people use it

10:46Effectively, what are the changes about this?

10:48And um it's always very nice to put the people building this technology on the hot seat and asking them like the really tough questions and getting the answers you want

10:56So the community helps people grow, helps people ask questions.

11:01And again, like I said, it's a very safe space for us to grow and thrive along with this technology.

11:08Yeah, I I think as um as analysts and in this industry, we you know, this in analytics has been moving

11:15quite fast for some time, I think.

11:17And there's always a like a new trend every every couple of years.

11:19So I think we're naturally sort of well built for this kind of change.

11:22We we're kind of used to adopting new things very quickly, new ideas, new philosophies.

11:27This is just another sort of technology

11:29Um but I think it it sort of nicely leads into the point around implementation.

11:33And Will, I know you've you raised some sort of um points about trust here.

11:37If I can come to you as as someone who

11:40um leads a team who's uh probably thinking at a much higher level about how to do this.

11:45So what are the big challenges around trust and transparency that you'd like to sort of uh deep dive into

11:51Well so it's it's one of those things that I look at uh with like a skeptic mindset, not that I don't think they can do this, but the

12:00Again, being data people, we're so used to data where I write a query and anybody can write the exact same query and get the exact same results.

12:08Whereas if I have five analysts on my team

12:12that ask the same thing, ask the same prompt, they're not going to get the exact same results.

12:18And so there it could be based on how they have been training their model, how the model has been, you know, evolved.

12:25There's there's many aspects that I can't just look at it from a show your work perspective.

12:31And so that's where the trust really

12:33We have to work with the the creators.

12:36We have to give the feedback.

12:38We have to understand ourselves that that human element still needs to exist.

12:43I can't just copy and paste this analysis

12:46into into a a report or whatever.

12:50I mean, so it's that's that's how from a trust perspective

12:54I look at it.

12:55But on the flip side, if I'm not talking about like a data analyst perspective, there's so many strengths that we can use.

13:01You know, um I'll t I'll tell you my ADHD brain, I always say the first thing that comes

13:06to my mind.

13:07Uh people can interpret those things as like emotional in one way or the other.

13:13Um so I've really embraced it from a

13:16You know, I'll type an email, put it in the LLM and say, take out the emotion, make this more fact-based, or give me give me a better way to phrase these things, or I'll give the bullets and say, can you make a sentence out of this?

13:29Because

13:29I think in bullet.

13:30So like there's it's not I know we're in a data talk right now, but it's not just AI is only for this piece.

13:38It is so broad and so

13:40I don't know if I've answered your question or gone down another rabbit hole.

13:43No, absolutely.

13:44The more rabbit holes you open, the more we can talk, and that's a good thing.

13:47So not a not a bad problem at all.

13:49If I sort of open it up more to the panel as well, like around this

13:54sort of a point around innovation and implementation.

13:56What have what have you all sort of seen or even thought about?

13:59Because I think this is something that's still a work in progress.

14:01So are there any sort of perspectives that maybe people want to add to what will Will has said there

14:09Go ahead, Mike.

14:10Oh, sorry.

14:11Yeah.

14:11Right.

14:11Sure.

14:12I also want to acknowledge that as we have all of these tools coming out, especially around agents and agent force.

14:21Data analysts are not only going to be training these, interacting with these, but they're also going to be managing the data that comes out of these.

14:30So we have whole new data streams that we're going to be encountering.

14:35associated with the telemetry with these agents and their efficacy.

14:40So it is a brand new world both from the front end and from the back end and I think it's a really exciting time as well.

14:47Yeah, it's good.

14:49And Maya, I think you had something brief to to add.

14:52I just wanted to briefly say I love that point on trust, Will.

14:55It's so important.

14:56And I think that, you know, you touched on it needs explainability

15:00The way to build trust is transparency, explainability, repeatability, but most importantly also guardrails.

15:07I don't think we talk about this enough in AI because we're just plowing forward always.

15:12But there need to be guardrails put in place that are often hard and fast business rules that we already live by every day and we don't need the AI to relearn.

15:20And that's all about building the trust so that folks can kind of take their hands off the wheel a little bit more every single day.

15:29Yeah.

15:29And that kind of feeds into this point around implementation, right?

15:34Because um I think on one hand, understanding how something works is is is is is vital.

15:40But no two organizations are the same.

15:41And so I think that naturally leads into this sort of follow-on question, which is, you know, let's say you find an AI use that's actually good and it works and you understand its sort of purpose as Natalie was talking about

15:52Um how do we implement these technologies?

15:55Sort of where where would sort of people start to think about implementing these technologies in their organizations?

16:01Um maybe Will, if I go back to you again, you know, just to touch on your perspective as well.

16:08I mean so working at a large global financial institution, um we have so many controls, governance, regulations, there are so many things that it's very hard for us to say like

16:21Hey, I've externally found these models and I want to bring these into the firm.

16:26So there's a lot of there's a lot of guardrails that companies should, at like Maya said, be putting up

16:34understanding how do we have controls, how do we make sure that we're not bypassing, you know, the security, the g it's it's it's very

16:43Fortunately, I don't have to make those decisions and I'm not in those meetings to to define the complexity, but I am the beneficiary of once those things do go live, going through all those um models that

16:55We work with our teams to say what what makes sense to apply, when do we apply it?

17:01How do we apply it?

17:02Um and and like

17:05adapt and evolve.

17:06I know you touched earlier, it's like every couple years there's new things.

17:09So a few years back it was big data, then it was cloud, and now everything is AI.

17:14And it's just we're gonna keep

17:17All those things still exist and the people who embrace those early like got to embrace the new things.

17:22And so I'm again going back to this, like I'm embracing this is this is great.

17:26I'm excited about it.

17:27Yeah, and uh it's it's an interesting one because um I I I think uh Rain, I was gonna come to you because

17:36obviously leading the Tableau user groups, you also work at Salesforce and Salesforce is one of the companies that's sort of pushing this change, right?

17:42And so from a from a

17:45um internal perspective, how are you finding that implementation?

17:49Because it's essentially you'll have to you have to dog food the implementation before you you tell customers sort of how to do it.

17:54So yeah maybe share a bit of a perspective

17:57sort of trials and tribulations if you can share some.

18:00I think that would be really useful for the for the group.

18:02Yeah, absolutely.

18:03And you know, I just want to acknowledge that we're we're all learning together as as all of this rolls out, you know, and I find a lot of solace

18:11in in that.

18:14Internally we are doing a ton of internal enablement on what all these tools actually do, and we're trying to make it less of a black

18:24box, right?

18:25AI really seems like this this magical thing that you put in a query to and you get something out, but we're not really sure what happens in the middle.

18:36And so we're really trying to pull back the curtain on that.

18:40And to Maya and Will's point with guardrails, I really do believe that these AI models are reflections of ourselves, right?

18:49So whatever we put into them is what we're going to get out.

18:53So data governance and data quality are something that internally we're always

18:59very concerned about and that's why I'm really excited um that these models that especially like Agent Force is based in um is

19:09guardrailed so heavily and has checks for biases and toxicity and things like that.

19:14So that makes me feel a lot better about it.

19:19Yeah.

19:20Thank you.

19:21Um and and Joy, I think from a from a user group perspective, uh are are people coming together as community to sort of

19:28think about implementation in general?

19:30Like w what are you seeing from some of the early user groups you've done?

19:34Because your user group is still quite young, right?

19:36So you know, you're probably getting quite um sort of a young crowd together.

19:39So yeah, w what what are you what are you sort of seeing happen on the ground?

19:44So implementation has actually ranged from doing like very basic stuff to doing more complex things.

19:51So one of our leads actually created a mockstar GPT.

19:56So what that did was

19:57You could create a mock data set for whatever thing you're trying to build.

20:02So excuse me, if you're just like starting out in data analysis and you need um a data set for marketing, for example

20:09Um not everybody, not every company has their marketing data readily available on the internet.

20:14But you need something to practice, something to work with.

20:17So so far we have like over a thousand people use that particular um GPT.

20:22And um

20:24So from that week we've able we've been able to see people in our community and people in DataFarm actually use these things, implement them in starting out their own journey and in using it to learn

20:35We've also seen people doing um a lot more things with agents, making their work a lot faster and um just improving their efficiency.

20:43So there's different levels of implementation from people who are learning to people who are experts and it's just very interesting to see that um in all works of lives and

20:54um based on like your different expertise you can actually still implement this and having a community where you can see how people use this and also replicate it is very good

21:05Yeah, thank you.

21:06Thanks for that perspective.

21:07Um it's interesting because I I think the other way that we all learn is by making mistakes, right?

21:12So I guess I I'd ask the the the group in general, what mistakes have you seen people make or what mistakes have you made that

21:18you're willing to share that maybe we can we can start to sort of discuss here.

21:22I know we haven't got a lot of time, but maybe the standout things that you've seen.

21:26Anyone anyone willing to go first?

21:30Oh I'll just jump in.

21:32One of the biggest mistakes I see is people try to swing for the fences their first time at bat, which is a sports analogy to say

21:41They try to take on a project that's way too big because everyone wants to transform everything rather than taking smaller steps and learning.

21:49The problem with these big projects is they're often very high risk, very long timelines, very high expense, a lot of folks looking at it saying, where is this going to go wrong?

21:58Where is this going to go wrong?

22:00There are so many smaller things that you can do around your organization.

22:04Pick the right problem to start.

22:06Pick something that's mid to low risk.

22:09that does have impact to the business, somewhere where you can really cut your teeth on AI and then build on that success to build more AI flows into your processes.

22:21It's uh really, really valuable.

22:23You've you've definitely seen a few mistakes, I'm sure.

22:27Um Will, anything that you you you maybe heard of other companies in your sector doing that you think, oh, maybe let's not try that

22:34Yeah, yeah.

22:35No, so I'll I'll take more on the personal route, right?

22:38Like when I first took on Chat GPT and I first learned I was

22:42I was like, give me a lesson plan on how to train Tableau in three days.

22:46Give me this.

22:47Give me like I I was just I was I was trying to swing for the fences personally.

22:51And I was getting stuff and I was just trusting it.

22:53I was like, man, this is so damn good.

22:55And it and it was it was good, but could it have been better?

22:58So I started learning about the evolution of like how do I prompt, how do I give feedback?

23:02Because a lot of it I'd take the initial output.

23:05And then go do something over here from the human aspect because that's my muscle memory.

23:09To the point where now I've gotten so comfortable with it, where New Year's Eve, I did my whole dinner plan.

23:16using Gemini because we canceled our stuff last minute.

23:20I said I told Gemini the basic prompt was I need I've got two hours to make a nice dinner.

23:26It needs to be child friendly and I want it to be kind of special and foodie.

23:30And then it gave me some ideas and I said, Okay, good, but I need recipes.

23:33Gave me recipes of those things.

23:34And I said, Well that's good, but I need a shopping list.

23:36It gave me a shopping list

23:38I went to the store, came back, and said, I really suck at time management.

23:41How can I be efficient with this time to make it in a small time?

23:45And it gave me all the steps.

23:46It combined and it was just like I built it as a conversation, as a companion, like a sous chef.

23:52And it was it was so much so it the adoption of like I want everything, give me everything to the point of where now it's like, okay, now I see how this fits and how it molds.

24:03And I'm applying the same thing at the office as as I start saying, okay, well I'm gonna take that step back and realize what where are we at?

24:11Where can we go, but how do we baby steps to get there

24:15Yeah, and and you touch on something crucial there because I think through prompting you are defining scope in in a way that kind of

24:22really narrows the AI into something specific to you, but also the scenario.

24:26You know, you could have you could have started out with maybe two sentences that asked the same thing, but because you kind of stepped through the problem, I think you got a better quality output.

24:34And as a result

24:35you probably had a much more rewarding experience.

24:37The the agency and your own ability to direct it was actually quite good to borrow a phrase from Sir Nath Natalie.

24:43So that's um

24:44That's super important.

24:45Um and joy and rain, I don't know if you've you've seen anything in the world even, it doesn't have to be from from a work context that you you'd sort of highlight to people to try and avoid.

24:56I think one thing I've noticed is people not asking um a lot of questions.

25:01So this is still new.

25:03And um there's a lot of things we still don't know and there's a lot of things that we do know.

25:09But asking questions is where like

25:12You get the freedom to know exactly what is going on and the issues that bother you.

25:16There are difficult conversations about this technology, and

25:19Feel free to ask.

25:21Ask the tough questions, ask the dumb questions, ask the hard questions, and just ask questions, basically.

25:29Yeah

25:30It's a s simple but great advice.

25:32Just ask Egg yeah, the honestly I that that's just such a it's so easy to forget that point.

25:37Um

25:38Uh we're all not experts.

25:39Uh even the experts clearly aren't experts at the moment.

25:42So that's um that's something that we can all learn from.

25:46And Rain, maybe from your perspective as well, is there anything that sort of stands out as something you think um peop people should learn from?

25:52Yeah, absolutely.

25:53I want to just echo what Joy just said.

25:55Ask questions.

25:57Really interrogate things.

25:58So I'm going to use just a very simple example.

26:01Every single time that you search anything on Google at this point, you are getting an AI summary.

26:07first and foremost.

26:09Um I want us to be really careful about treating AI like it is a silver bullet for all problems and forgetting our critical thinking skills, right?

26:18It is so important now to interrogate where

26:22these responses are coming from and the validity of these responses and the sources that these AIs are that these AI models excuse me are are pulling from

26:34Um when we see something like Google saying, ah yes, a human should eat eight cigarettes a day, we know that that is false.

26:42We just know that, right?

26:43But when

26:44the responses seem more reasonable, we need to come at it with that same level of skepticism and curiosity as well, and we're going to learn a lot more along along the way as well.

26:56And I think that's um links to um a point Natasi made around ownership, right?

27:02I think sometimes when we when we're doing things we're we we have to do work.

27:07And we think of it as a task, but actually we own the process, right?

27:09We come up with a plan, we deliver the work, and then we actually present the work.

27:13So we have a sense of ownership and pride as well that's sort of part of that.

27:17And something I've talked about is this sort of idea of when you sort of hand off too much of that to AI, you kind of lose a connection with the with the with the final output.

27:26And I think it's

27:27that connection is actually what makes it human and that's what sort of makes us build um sort of rapport with each other in the workplace, with our products, with with the with the work we do.

27:36So I think it's very, very sort of important point.

27:40On the point around um AI, I think we talk about it like it's one thing, right?

27:44We talk about it like it's one discipline.

27:46We've said earlier on that it's actually

27:48a bunch of disciplines.

27:49I'd love to sort of dig into that a little bit more.

27:51Um are there any sort of pockets within

27:55I I'd say AI has sort of become the headline, but maybe we we talk about machine learning, maybe we talk about sort of statistics, maybe we talk about predictive analytics.

28:03These are all, I guess.

28:04uh now unfortunately subheadings of AI.

28:08So are there any elements of of of those things that actually that you are sort of excited to see improve or develop or get more airtime because now they're all part of the discussion, right?

28:18And uh maybe Maya, I saw you nodding your head, sorry, I should have directed that to you.

28:22I mean, I love me some old-fashioned mom jeans machine learning.

28:30I think we are not using it enough and we've jumped a lot of a lot of enterprises are jumping from zero to I want to fine-tune an LLM

28:40And we forget that AI is essentially a toolkit.

28:45And just like in a toolkit, every tool is good for different types of use cases, for different types of data.

28:51LLMs might be great for certain types of unstructured data.

28:55There's vision models that take care of looking at objects and making classifications.

29:00There's recommendations.

29:02There's now decisioning that we're talking about more and more.

29:04Multiple models are blended together to make better decisions

29:08We can't think of it as a one-size-fits-all solution, even though it's easy to just bucket it all under the AI umbrella.

29:14I would encourage folks, especially the data people listening

29:18on on this broadcast to if you work in tabular data, do not forget machine learning.

29:25It is excellent in these scenarios.

29:28It is really and and it has explainability and it has transparency and you don't have to worry, you know, it it it's been around for decades.

29:38People have tested it out quite thoroughly at this point

29:41But do find the best tool for the job, whether that be an LLM or not.

29:47Are there are there any other perspectives and linked to that?

29:50Sort of

29:51Uh you know, any of these subsets that you're you've all used maybe in the past that you're excited about?

29:58So um

29:59I know Joy talked about the early on adoptions of some of the GPTs within the the community.

30:05I I was I I loved

30:08one of the things that Adam Miko put together from a dataviz, I forget the exact name of the GPT, but basically you can do a a summary of like

30:17Is this Viz good?

30:19Like what's the quality of the Viz?

30:21Is it readable?

30:22And he put in a lot of criteria behind it.

30:25So to me, the visual aspect, being able to look at a picture, my brain immediately went to

30:31Companies could start implementing, you know, brand guidelines, put in accessibility, put in all these different things and say, you know, looking at the at the vis at the viz.

30:42Are the fonts readable?

30:43Are the colors good?

30:45Are we you you start defining what your guidelines are and start feeding some of that stuff through there?

30:51Um so like my brain goes towards

30:54It it just boggles my mind how it how we can do all of that.

30:57So that's why I'm excited about.

31:00Good, good.

31:01Uh Joy, Rain, anything that stands out for you?

31:06Something stands out to me, but it's not exactly exciting.

31:10So um not to be the kill joy here.

31:13Um but

31:15But an important thing to me is security, right?

31:18So we're putting all our data in models and um how secure

31:23is this, right?

31:24How many companies are um thinking about their security and in very efficient ways.

31:29Because we've we've seen a lot more

31:32um attacks going on lately, especially with like deep fake um modifications of injections and so many things going on.

31:40And these are important conversations to have

31:42So as much as we're excited about like what AI can do, where we can just put our data in and by some miracle it solves all our problems, we should also be very concerned about the security of our data.

31:56Yeah, it's a great great point.

31:58Um we all use uh if you use like a web portal for any of these services, one of the reasons they keep a certain amount of uh usage free is because

32:08in using it you're you're putting data into it, right?

32:10You're you're putting questions, you're asking perspectives, you're you're you're typing a lot about yourself into that text box.

32:17And that all goes somewhere, it's stored and it's used to improve the model.

32:20But as you say, it's it it contributes to the model in some way or form.

32:24And

32:24I've had the the fortunate chance of uh being referenced in an AR model um in in regards to an explainer about Tableau

32:32pie chart in it.

32:33Literally used my wording in the response and it linked to my video and I was like, okay, that's that's slightly, slightly weird.

32:40But um

32:40Yeah, that that point about security is super important.

32:43And of course AI models intrinsically in the way they're built, they you know they use vector databases.

32:48You lose that lineage actually once they're built.

32:51Once these

32:52AI systems are stood up, that lineage that allows you to ask the question about security is lost.

32:58So I think it's a really sort of valuable um point to to to remember.

33:03If we maybe move on to collaboration and community, that's sort of what we're here for.

33:07We're an amazing community.

33:09We have uh over 500 people online watching this.

33:12Um

33:13How do we how do we do this together?

33:15How do we how do we form um sort of maybe maybe call them teams or groups within our organizations that

33:22uh can can take on these sort of small challenges, these small tasks of understanding how best to do this.

33:28What's sort of the best way to do that?

33:30Um uh Rain, maybe uh you know you're you're in Salesforce, you're doing internal enablement

33:35How are you sort of uh uh like solving that problem?

33:39Yeah, that's a great question.

33:41Um so putting transparency first is

33:46absolutely the first thing that we should be doing.

33:48So what we don't want is to unplay onto our our companies without any explanation of

33:57what it's doing and why it's doing and what sorts of problems it's appropriate for.

34:03So being very transparent about what

34:06this tool is doing and what kinds of problems it can actually solve versus what's out of its scope is extremely important.

34:15As far as bringing people together, that's what things like user groups and internal user groups are really useful for.

34:26When we learn as a community, embedded in a community of our peers who are all working on similar problems, we're always going to learn

34:35faster and especially when there's a little bit of joy and play in it uh like I do see in all of our Tableau user groups

34:44I think somebody dressed up as an agent in one of our recent user groups.

34:49That is really when all of that learning comes together.

34:53in real-life applications.

34:55So it's really important that we get hands-on with real data as soon as possible rather than just training sets or just theoretical learning.

35:08So I I would say my um my my favorite thing about

35:14This industry that we're in, data visualization as a career path, the thing that brought me in was the community because I wasn't alone and people were so giving of their time, effort, and energy.

35:27And keeping that mindset, we will be just as successful in in this space.

35:34And so whether it's

35:36having coffee and talking to somebody that works in the same area as you at a tableau user group, you know, it doesn't necessarily have to be the AI user group.

35:45Some people can present at other user groups and kind of say how does it

35:48come together.

35:49Just basically it is it is the more we share, the more we're open about our experiences that might spark ideas that somebody else had.

35:59Um my idea for creating dinner for New Year's Eve was because I'd seen some reels about people saying ADHD people, there's a way to create a shopping list.

36:07And so in my mind, I said, well, if you can create a shopping list

36:10Can it tell me how to just combine all these recipes?

36:13So it it's just it's it's all these little ideas of just us helping each other and we continue to stand on the shoulders of the people you know before us and we'll be the shoulders of others.

36:21So it's just keep growing.

36:23It's that incremental sort of um journey, right?

36:26Like um the more we share, the more we can all learn from each other.

36:30And I think user groups are a great vessel for that, but there's there's also plenty of um

36:35uh experts outside of our tableau community, right?

36:38There's there's experts in the AI communities, experts in other tools as well.

36:42We can bring that knowledge and those best practices into our into our group, then I think it's a really good way to

36:47help accelerate that but also show that um you know we're we're all coming at this from sort of different angles and we can also bring bring

36:55disparate perspectives.

36:56And and I think it's also important to include um uh cultural perspectives because I know that a lot of AI tools uh uh fundamentally just

37:04grounded in English.

37:06And so they lack a lot of uh sort of a cultural awareness inherently just because of the way they're trained and the way they're set up.

37:12So we as a community have to sort of step to the plate and actually

37:15fill those gaps in the way that we use, train, and make sure that we bring everyone along with us.

37:20It's a super, super important um sort of element there.

37:25The other thing I wanted to ask in general is a lot of these AI tools, there's lots of different AI tools, right?

37:30So um you might have heard ChatGPT, Deep Seekers sort of a conversation today, Alibaba did Quen something.

37:37yesterday.

37:38There's so many different ones.

37:39Um in talking about AI, uh are any of you are any of you sort of building a paradigm on how to approach, which model to use, when, how

37:50Um, you know, that sort of decision, because we we've talked a lot about whether or not and how to use AI.

37:55What about which one to choose?

37:56How do you choose one?

37:57How do you sort of

37:58Experience one.

37:59Uh Will, you talked about Gemini Hat.

38:01Why did you do Gemini and not sort of Chat GPT?

38:04Maybe I'll ask that to you first

38:06Um I just wanted to try it out and see.

38:08I've I've I've had you know ChatGPT app on my phone for a while.

38:12Like I I think it knows me and the aspect of like I've trained it to based on my various prompts and how I do things and so I wanted to see

38:21I've not really used Gemini.

38:22I'm an Android user.

38:24I have the app on my phone.

38:26Let's try it and see.

38:28And like it to me it like

38:31I don't know enough between the different models to be able to say, oh, this is why I would choose this ty this version of ChatGPT, this model.

38:39I'm just like I'm just experimenting until someone can tell me the specific reasons or I find the specific reasons.

38:45It's just trial and error.

38:47You and me both, right?

38:52Any other perspectives on that?

38:56I'm a big fan of staying model agnostic.

38:58The industry is just changing much too fast.

39:02to just, you know, bet the farm on one model and say this is going to be the model that wins.

39:07I think we all learned our lesson with, as I call it, the deep seek freak out

39:12That happened earlier this week when everyone discovered a model that was available since December.

39:17Um and what that means for the industry.

39:20I I think if you back into it with starting with what's the problem of I'm trying to solve

39:26you know which model is best to solve this problem and stay flexible and loose, especially if you're designing for an enterprise framework to these different models.

39:35Because look, these models are going to get faster, better, cheaper

39:39And if you overly build towards one type of paradigm, you're really closing the door to these new innovations.

39:47I prefer Claude, for example, for writing.

39:50I think it's a little bit stronger when it comes as an editor and a writing assistant than some of the other models.

39:56But I might flip back and forth between models, you know, I might use perplexity for search, I might use this for that.

40:02So it it really depends on the use case.

40:04So kind of start with the problem you're trying to solve, then do a little bit of research, but remember to stay flexible about it.

40:11Yeah, it's uh it's a great point actually because um the we we we live in a world where I think analytics tools um

40:20uh ha have become commoditized, but commoditized in a good way.

40:24And for example, the number of integrations we have today with Tableau, whether it's DBT or uh the numerous database connections, you can sort of pick and choose uh and and build your own

40:34sort of analytics journey.

40:36And the challenge with AI is that it's it's sort of um landlocked because the investment required to build this model is so substantial that really unless you're

40:45the big six, you you don't have uh a chance of training anything as powerful as that.

40:50So we're all essentially customers of uh you know fundamentally maybe a handful of really good tools.

40:56But if you if you go

40:57into the community like hugging face.

41:00You'll actually find there's a really diverse community of of models that all do very small but powerful things, whether it's image recognition or

41:08uh maths or you know Python, whatever it is, coding, like all of these disparate things.

41:13And the big tools suck up a lot of the energy, but maybe maybe we could all do a bit more to to to sort of

41:20demand more from smaller, more capable, sort of niche models that do very specific things, but very, very well.

41:26And actually maybe run on our laptops as well.

41:29To save us uh having to share things uh in into the internet um as well.

41:34Um any other any other perspectives?

41:36I'm just gonna open uh sort of encourage everyone who's watching this, because I know there's a bit of lag, to ask questions because we've only got uh uh uh five more minutes.

41:43So

41:44uh please um uh ask questions in the q a and we'll start going through them in about two minutes.

41:49But um just before we do that, um we

41:53We probably should sort of touch on sort of any any sort of bits of wisdom if you were to sort of leave this panel and uh tell some of the people watching this session today.

42:02to do something about AI, what would that what would that thing be?

42:06And so if I just go around the panel, I'll I'll start with Joy, because you're top of my screen, Joy.

42:10If you could tell people to do something um about or with AI today, what what would it be?

42:17I'd say to join a community, right?

42:19I joined the data farm because I was interested in it all, but I stayed for the community.

42:24Join the community today.

42:26There's amazing people everywhere, people that would always look out for the best for you and will teach you and definitely guide your experience.

42:35So join one.

42:37That's also a great t-shirt, you know.

42:39Uh pick tableau, stayed for the community, love it.

42:41Exactly.

42:43Okay.

42:44Uh next up, Will.

42:45Um yeah, any any bit of wisdom for the for the audience

42:48Um I'd say just stay curious.

42:50Uh like Joy talked about, or just keep asking questions, just keep asking questions because the more questions you ask, the more you learn, and then those questions are going to lead into other questions.

43:00And so

43:01It it's the the beginner's mindset is is just such a a critical thing.

43:07We just always have to take that on.

43:09Amazing, amazing.

43:11Rain?

43:13I'm gonna echo everything that was just said, join a community, ask questions.

43:18And also

43:20Learn.

43:21Learn as much as you possibly can, even if you are deeply skeptical about AI, if you think it's the best thing in the world

43:31You need to learn.

43:33If you are pursuing a career in data or really anywhere, learning the language is your very first barrier.

43:41So be sure that you are fluent in in this AI world, just in terms of the basic terminology, and that will get you really far.

43:52So stay curious and stay critical and also.

43:56Join a community.

43:57Find people who are learning together so you can learn from one another.

44:05Wow, plus one to everything Joy Willenrain just said.

44:09I would just also like to say from a much more eloquent Maya, Maya Angelou, she said every journey begins with a single step.

44:16Do not get intimidated.

44:18Just get started.

44:19A lot of these tools are free to use, free to try.

44:23Just start.

44:24Start and you'll learn and start and you'll form community.

44:27Start and you'll find others to talk to.

44:28Just start

44:30Amazing.

44:31Amazing.

44:32Thank you everyone.

44:33That's really um sort of thoughtful.

44:35I think it it's a nice way of encapsulating everything that we've we've sort of seen today.

44:39Um we've got time for maybe um one question.

44:42Um I I can see a a question here from um Scott.

44:45Um I'll just ask it, does anyone see AI replacing a dashboard designer developer?

44:50Um he's referencing a video from Andy Cockgreve where he got the AI to sort of prompt him through building a dashboard.

44:57So I I don't know if anyone's seen that video, but more generally, um that question about you know being replaced.

45:02Um anyone want to take that?

45:05Yeah, I will.

45:06I'll take that that question with a similar meme that I have seen.

45:12I don't see AI replacing us because then that's going to require people to give us the exact requirements of what they want.

45:18And so until they're able to give us the exact requirements, I we are we are good.

45:24It's it's part of that partnership.

45:26So can it help us get speed to market?

45:28Absolutely.

45:29But the not necessarily the human touch of the back and forth.

45:34But just being there and understanding and like Rain talked about translating, just because someone says I want a pie chart, that doesn't mean that they want the pie chart.

45:42They want to dissect

45:43They want a distribution, like those types of things is is really where we come into play from the human aspect.

45:50We can get faster, but we're still definitely going to be here for that.

45:54Yeah, exactly.

45:57Yeah, I'm completely aligned with that as well.

45:59I think um uh you know uh all all of these AI chat engines have one thing and it's a text box and unfortunately it doesn't fill itself, right?

46:08So

46:09Um we all have to be part of that, you know, prompts.

46:12We have to write the prompts.

46:13All these AI tools require input for them to be able to output something.

46:18And

46:18I know we're moving to a world with agents, but at the same time I feel those agents need direction.

46:23They need um sort of peers to help judge their work as well.

46:26So it's super super important point to

46:29to uh end with there.

46:31Okay.

46:31Um I think that's been it.

46:32I I've really enjoyed the panel.

46:34Thank you all of you for your wisdom.

46:36I think it's been a really good way to just uh share our perspectives and and and and just really, you know, understand how we're all thinking about it.