Will Salesforce Be the Most Successful AI Company? And Do You Need That in Real-Time? Featuring the Cynical Data Guy

Eric Dodds 00:06
Welcome to the data stack show.

John Wessel 00:07
The data stack show is a podcast where we talk about the technical, business and human challenges involved in data

Eric Dodds 00:13
work. Join our casual conversations with innovators and data professionals to learn about new data technologies and how data teams are run at top companies. Welcome back to the data stack. Show. It is a new month, and that means a new episode of this cynical data guy, Matt, as always, it’s great to have you back in the studio. It’s good to be here. Who canceled?

John Wessel 00:44
What are you saying?

Eric Dodds 00:45
Why is Brooks crying? Okay, if you are new to the cynical data guy episodes, they’re really fun. We do a lightning round where I go through and hand curate some tasty LinkedIn posts related to data and Matt Keller. Keller and Gibson, who is a cynical data guy who has been jaded by years in the bowels of corporate data America, gives us some commentary. We try to balance it out with John, the agreeable data guy, and sometimes I even weigh in with my opinion. But we try to, we try to, we try to let him do good. I

John Wessel 01:21
have to call up for the listeners that Brooks. Brooks just had to go on mute and turn his video off because he was already laughing. So

Eric Dodds 01:26
This is going to be a good show. It’s going to be a good one. Okay, free Lightning Rounds. We’ll see if we get to a bonus round. But there’s some good stuff today, so I don’t know. And we are going to start off with a real banger here. Okay, ding, ding. Marc Benioff, the CEO at Salesforce, first LinkedIn post ever is about AI and is probably written by AI, which I feel like someone we cannot confirm nor deny, deny it, but I feel like someone would have made this up. I mean, you really it’s that good that let’s just dig in for my first ever post on LinkedIn. I’m excited to announce that as of today, Agent force, our complete AI system for enterprises, built on the Salesforce platform, is available for all customers, easy to set up with a few clicks and a simple description of the job you want done. Agent force is ushering in a new era of AI abundance and limitless workforces that will augment every employee, build deeper customer relationships and drive extraordinary growth and profitability. Should I keep going? But wait, there’s more. But wait, there’s more. Okay, yes, many enterprises are caught in a pointless cycle of AI experimentation with llms and Co pilots that lead to costly failures or proof of concept projects with no path to scale with Agent force, you don’t have to DIY your AI exclamation point as part of Salesforce is trusted and fully customizable platform. Agent force seamlessly integrates enterprise data, metadata, AI models, workflows, security and applications, no costly model training, data management or hyper scalers or AI engineers needed going beyond copilot and chat bots. Agent force, agents don’t just answer questions or surface insights. They autonomously execute actions like resolving customer cases, qualifying sales leads and optimizing marketing campaigns. Companies like Open Table SaaS and Wiley are already using agent force today to extend their employees, expand their workforce and improve customer experiences. This is what I was meant to be. Hmm, deep, deep, deep South.

Matthew Kelliher-Gibson 03:47
It’s yeah, okay, fine. Let’s just word salad our way through all of this. I don’t know. It’s yeah, of course. It’s gonna do everything. It’ll clean your windows, it’ll remove stains from your walls, and it’s a salad dressing. I don’t know it’s I’m just so tired of these over the top things,

Eric Dodds 04:07
okay, but for his first How do you think that it came about that we decided we’re actually going to do his first post about this? Let’s just make up a backstory. Yeah, I need a backstory. I need a cynical backstory.

Matthew Kelliher-Gibson 04:21
Well, let’s see, what would this cynical backstory be here? I don’t know. Perhaps it has something to do with the need to do something to draw attention to this in a different way. And so someone had the idea of having it be posted on his LinkedIn, possibly after some comms or marketing people had been asking for months if he would actually post anything on LinkedIn, and he was not exactly sure what they were talking about. Well, see,

Eric Dodds 04:48
I like to think of a very like years long political battle over who, you know, there’s an internal competition with this guard who has survived many. Years at Salesforce, you know, and is that sort of, like the VP, executive level, and there’s, like, maybe, actually there’s even, God, am I, like, usurping the cynical nature, like a large sum of money on the table for who will win in, like, getting Mark’s first post out the door?

John Wessel 05:17
I mean, that’s a real possibility. I would imagine this is part of some kind of campaign, right? Because everything has to be part of a campaign, of course. So it’s like a PR Blitz, and like, oh, let’s do a LinkedIn post. And then somebody was like, what you’ve never posted on LinkedIn, this would be perfect.

Matthew Kelliher-Gibson 05:32
Well. So my big question with that is, is the story going to be that he’s never done it before, has never wanted to, and like, they finally broke through. Or is it going to be one of those where he’s been talking about, this is going to be a great idea. He’s going to do this. And everyone was like, No, you’re not doing that. And they finally came, that’s even better. And they finally came to an agreement that you can post if, but only if, we can take what you wrote and rewrite it and be the ones in control of it. It’s

Eric Dodds 06:03
actually like a masterful redirect of like, the Yeah, he’s like, I want to be like, Elon

Matthew Kelliher-Gibson 06:09
just on LinkedIn. And

Eric Dodds 06:13
someone’s like, Okay, we actually, we’re gonna get you there.

John Wessel 06:18
Somebody’s like, baby steps. Let’s trust this so well, and we have to read, we have to read the top comment.

Eric Dodds 06:25
Read the top comment, okay, more cynical than even, I

John Wessel 06:29
think cynical, yeah.

Eric Dodds 06:30
I mean, this is savage, okay, ready. Your first post is shameless self promotion. Let’s hope your agent does better on the next post.

John Wessel 06:40
But if you read it, it really is like, I mean, I think we were talking before the show, like you’re on LinkedIn, but like, the post is totally on brand for LinkedIn. Of course, it’s all shameless self promotion, yeah? So yeah, it doesn’t feel like it’s more shameless than any other LinkedIn, yeah. So

Matthew Kelliher-Gibson 06:55
maybe, I mean, at least he has more than six months work experience before he used to do that. LinkedIn,

Eric Dodds 07:01
yeah. I mean, okay, that’s the other thing. Like, we can armchair quarterback Marc Benioff, so hard, but he’s, like, he gets a last laugh, right? I mean, Chris, yeah. Okay, so I’m gonna extend, should we talk about the content of the post a little bit? Should we do just well, okay, so I’m extending round one a little bit here, which is probably why we won’t get to a bonus round. But I knew this is gonna knew this is going to be a needy one. Okay, I want the cynical and the agreeable response to this. Salesforce will be the most successful AI company.

Matthew Kelliher-Gibson 07:30
How are we measuring that? Because they that’s

Eric Dodds 07:35
such as a data leader answer. Well, I

Matthew Kelliher-Gibson 07:38
I mean, it’s interesting, like, are we saying they’ll be really successful because they’ll have this AI thing, a lot of revenue. Are we saying specifically, people, okay, using the AI,

Eric Dodds 07:48
Okay, in Okay, outside of the Great question, that’s why we love you cynical data guy, okay, I will put parameters around it, but I had to ask it in a short bunchy, right? Of course, yeah. Outside of the likes of GPT and the consumer facing that use case, right? We’re talking about AI integrated into a SaaS platform, which, you know, so many companies everyone’s doing, yeah, Salesforce, I think, will be the most successful at that, true or false. I just said, I introduced my opinion that Salesforce will be the most successful at that, true or false?

Matthew Kelliher-Gibson 08:22
I mean, I wanna say no, but then I think about it, and it’s like, whenever you kind of count out these legacy platforms, they seem to find a way to be like everyone uses it and everyone hates it, type of a thing. So maybe

Eric Dodds 08:36
agreeable data guy.

John Wessel 08:38
I think there are very few companies still, and this is funny to say out loud, they can out sell. Salesforce, I think they’re so good at sales, so good at it. And it’s silly, right? Because obviously they sell a tool that helps you, like sell better. But that’s not a given. There are a ton of tools to help you do something better that the company internally is not that good at. But Salesforce, I will give them credit. They’re excellent at sales. Yep. So well, if the key to success in a is like, who can sell it the best? Yes, they’ll win. And that’s probably the key. Okay, I

Eric Dodds 09:10
I think there’s another aspect, though, and I’m gonna take the liberty to introduce my own opinion here, which I try to minimize, but I feel strongly about this. There’s another aspect of this, which is distribution. And so the example I’ll use, and I’ll take Tableau out of the picture here, because Salesforce bought Tableau, which is, you know, one of the, one of the largest, you know, sort of BI, you know, sort of traditional BI solutions. But if we take Tableau out of it, and we just think about Salesforce as its own dashboarding solution with, you know, it’s their reporting products within the Salesforce suite, right? I would guess that it’s probably one of the most widely used dashboard solutions in the world, right, in terms of end users accessing data in some sort of dashboard with charts or whatever, right? The distribution is just so immense. And I think, like their dashboarding solution, even if it’s not the most incredible. AI solutions, like a dedicated AI company, are going to come up with a much more elegant, whatever, you know, solution to the problem, right? It just won’t matter, because if it’s moderately useful, and Salesforce, you know, can sell it, and if it’s moderately useful, or useful enough to people to use it like they just will within the Salesforce platform, right? Well, they

Matthew Kelliher-Gibson 10:20
can get it to update people’s opportunities and sales forecasts, then that’s

Eric Dodds 10:29
the conclusion, yes. Is updating opportunities.

John Wessel 10:33
Is that what I mean by the sale, the sales pitch for it is there, because if you ask any sales person in the world what they hate about Salesforce, it would be that I have to use it, yeah. I mean, like, and then I have to, like, go in and manually update. Like, any sales person is like, Oh, I have to update this manual. I have to update this manually, yeah, if this works, like, as advertised, and they have to do less of that, then, like, I think everybody will want to buy it. Yeah.

Matthew Kelliher-Gibson 10:57
I have a feeling that it will get very big. It’s not going to be because of all the things they talk about. It’s going to be because it does one or two really boring things that are high frequency over,

Eric Dodds 11:08
which is the case with Slack. Ai, yeah, right, sure is it. Did it completely change the slack experience? No, do I want slack without it? No, because it’s really useful for things like searching for something specific across, like a wide set of channels, or, you know, summarizing, or whatever, like, it’s so useful for that. All right, any other hot takes on Benioff? I mean, he’s a master, like, yeah, he can post whatever he wants.

John Wessel 11:33
One other. One other take on this post before we move on, is, I mean, the it is interesting that they show they choose to jump in at this intersection because they could have done it a lot earlier, and they’ve jumped in right after, you know, anthropic announced the the new abilities, at least for developers, I think they’re rolling it out, like, more broadly, soon, to actually perform actions on your Computer. Yep, like, think, like macros, or whatever you want to call them, yep, like actions, so that that, I feel it’s interesting, like, the timing of it, that that comes out in, like, a week or two later, this comes out with, allegedly, like, similar futures, where it can actually perform more actions, and it’s not just summarizing information, or, you know, whatever

Eric Dodds 12:18
Penny up is, the boss is perfect as as soon as the curve of the, you know, curve going into the trough of disillusionment, and, like, the budgets and the lackluster results, yeah, start to become acute, he comes in and he’s like, Hey, only thing I’ll say is Agent force. Every time you said it, all I could think of was the Space Force. Yeah, I thought of the matrix for some reason. Yeah, okay, all right, down to this poster will go unnamed. When I led my webinar on how to get value out of your data, I asked attendees why they were interested in the topic. Here are some of the most common responses. We want to figure out how to monetize our data. We want to better leverage our data. I want to be more data driven. At first glance, these responses might seem vague, but upon further reflection, it’s pretty clear to me where most of these people need help. I will elaborate, but first, let’s talk about composting. Compost. You’re gonna love this one. Imagine someone learns about composting for the first time. They’re curious, so they buy a compost bin to try it out. They start cutting some vegetables, and before they know it, their bin is full. Wow. Okay, now I have all this trash. I heard that I can use it to feed our garden. How do I do that? Exactly? Maybe I just throw these carrot peels into my azaleas. You see where I’m going here. Now imagine the early stage founder. They have a digital product. They’re integrating digital systems into these operations. Maybe they’re scraping open data from public APIs, or maybe they have a bunch of random documents from their customers in an S3 bucket. When founders bring their ideas to reality, it starts to become very clear to them what this whole data thing is. They recognize they have a lot of it, and they’ve heard how valuable it’s supposed to be. They know how important it is to be data driven. They just have no idea how. Meanwhile, many people who can help them get lost on learning new tools that vendors shove down their throat. But I’ll save that rant for another so

Matthew Kelliher-Gibson 14:14
first of all, are we saying data is trash, right the carrot, the old

Eric Dodds 14:22
or this sort of you, you bind this really useful thing, but it’s like a it’s a byproduct of,

Matthew Kelliher-Gibson 14:30
you know, so we’ve gone from data is the new oil to data is a carrot, your leftover. I

John Wessel 14:37
I mean, this opportunity to not just name the post, like what I learned about data from composting, like, we should have just led with that. We should have led with that. We should have led with that.

Matthew Kelliher-Gibson 14:46
Five things I learned about SaaS sales from compost. Yeah, exactly.

Eric Dodds 14:53
So is data a care appeal in your alias?

Matthew Kelliher-Gibson 14:57
I mean, there’s, like, some decent. Points in there that, like when people first start, but I think it’s very specific. Like, if you have a digital product and you’re tracking certain things, and you kind of realize, wait a minute, I can use this for more. That makes some sense. And there is some mismatch to it. But I mean, there’s also kind of, you know, there’s a lot of people who have that idea that, like, I remember I was working like, kind of internal consulting at one company, and we, a guy wanted my group to come in, and it was, he was like, well, we want you to evaluate our database. And, like, what, what do you mean by that? Well, we want you, we feel like we get more of us. We want you to, like, evaluate how it could be used. And I’m like, I don’t know what you’re talking about. So you know, like, there’s still kind of that cold start problem, yeah, out of that, when you say, like, well, I want to be more data driven. Well, what does that mean to you, that you want to be more data driven? Yeah, John, I

John Wessel 15:49
think, I think I understand what he’s saying. As far as he started out with, we want to be data driven. We want to better leverage data, monetize our data. And then he goes straight for like, the data, implying that the data is bad and we have to transform it and do things with it to make it better. Because that’s what composting is. I don’t think starting there typically helps people starting in the end verse of like, what are you trying to accomplish? Like, not like, oh, I want to leverage this day to be day driven. It’s like, no, what are you in the business of? What do you do? Like, define that, and then define, like, what would provide value to customers, and then you build backwards from that with a data product, which

Matthew Kelliher-Gibson 16:27
may be data may not be or may not. Yeah, yeah, yeah. I mean, think it’s kind of inverse, yeah, the composting one is a little weird too, just because, I mean, there’s composting is kind of like, I let it sit around and it breaks down,

John Wessel 16:40
oh yeah, it gets better over time. That would not be true. That is not that’s not

Matthew Kelliher-Gibson 16:43
true. It’s like, I mean that, but that does feel like the rotting aspect is true. Good point, it is true, but it doesn’t turn it into something useful at that point. I mean, it’s a little bit right the way some people perceive data in companies where they’re like, well, we’ve got 10 years worth of this. 10 years it’s been sitting around, it should be valuable. And it’s like, well, no, there isn’t, like, a whole micro, you know, going around and breaking it down and turning it into nutrients like that. It’s just, you know, dirty data that’s been sitting there for 10 years. Yeah.

John Wessel 17:15
I mean, what’s your opinion? I mean, if you have history like maybe that helps if you’re trying to build models because you have more data, yeah, to build a model from,

Matthew Kelliher-Gibson 17:23
that’s good, but it’s not like it’s got, it’s not matter over time or something like that. And then I will agree with him on the part, there’s a lot of people who focus way too much on the tools aspect of it, but that, that more has to do with where people get very lost there. You know, it’s kind of like the person who says, like, well, I just really want to, like, make models, or I just really want to make dashboards or something, and it’s okay, but that’s like a component of this. That’s not, like a job, you know, this isn’t a widget factory where we’re turning stuff out. Yeah,

Eric Dodds 17:54
I love over analyzing, and I love these, like, way too deep, unfair analyzes of people’s analogies, right, right?

Matthew Kelliher-Gibson 18:04
You don’t want me to over. You should have thought it

Eric Dodds 18:09
Okay. One more question on this one thing that’s interesting, and of course, this is an unfair, you know, unfair analysis of the analogy. But there’s this implicit idea that you can’t have any data waste, right? We have to figure out some use for all of this. Yeah,

Matthew Kelliher-Gibson 18:27
That’s wrong. It’s legal. There’s a ton of data that’s just pointless and not going to do anything. Yeah? I mean, that’s because I remember when I started that was, like, the way people thought was, well, we’re going to find, we’re going to refine all this. It’s a real mindset, which is, right? Yeah, it’s a real mindset of like, well, we’re going to collect everything, and then it’s all going to be useful, and it’s like, no, 80% of that’s probably going to be a waste.

John Wessel 18:51
I actually like how the industry has gone with that, where collecting more data, storing more data, is marginally more expensive, but collecting more data is a lot more expensive now, because everything’s usage based, because before, if you just scoped out, I have this fixed amount, then it’s, I’ll just collect whatever I want, and then I’ll, you know, ask for more resources later. But now that it’s like, so scalable, I think there’s a little, there’s a lot more pressure to be collecting useful data versus all the data, right? Well,

Eric Dodds 19:20
The other thing is, it’s getting a lot cheaper too, you know, with things like icebergs, it’s going to be that it’s going to drastically, yeah.

Matthew Kelliher-Gibson 19:29
I mean, the data, like, idea is one that it’s like, you can collect a lot more, and, yeah, the storage is cheap. Yeah, yeah, you can see what might be useful later. You just run into the problem of, Oh, crap. How do I figure out the user well, and

John Wessel 19:43
I think people are more concerned about liability with keeping data forever than they used to be. Yeah, that used to be a thing where it’s like, we’ll just keep it forever. Now people are like, Oh, maybe it’s better that we don’t have that data. Yeah, all right. Round number

Eric Dodds 19:55
three, this has been fun. Oh man, this I forgot what three. Was that this is such a good one, something we’ve talked about on the show for literally years. Okay, as a data practitioner, I can’t count the number of times I’ve been asked the same question, can I have this data in real time? And for years, my answer has been some version of No, you can’t get data more often than once an hour. It’s absurd. We’ve come to accept this as normal, stakeholders often have perfectly valid reasons for wanting up to the minute data, but we’ve been stuck between the limitations of batch processing and the complexity of building real time systems at twirl we’re rethinking orchestration to address this. Traditional orchestrators force entire pipelines to run in one go, meaning the whole pipeline needs to finish before starting the next run. Twirl changes that by allowing each part of the pipeline to run independently on its own schedule. We treat every step in the DAG as a micro service and constantly evaluate if it’s time to run. By default, a job will run if all its inputs have processed new data, but users can override this and specify different triggering logic for each node. This means high priority data can refresh continuously, providing near real time insights. Less critical processes can run hourly or daily. Align with when the data gets used by decoupling the cadence of each step, we blur the boundary between batch and real time data processing. This means teams can deliver low latency data where it makes sense without overhauling their entire infrastructure or rewriting their code, and they tee this up perfectly for me. We think this approach is much better aligned with real world use cases. What do you think?

Matthew Kelliher-Gibson 21:37
I think that highly depends on what business you’re in. I mean, I mean, I’ve worked at a lot of places that had, like, you know, brick and mortar type locations, and the idea of, like, we need to update. It’s like, we don’t even get downloads until the end of the day. So, right, yeah, you’re gonna do it, right? I mean, I think this is, like, probably leaning more towards a digital type product, but you’re gonna be there because, I mean, I don’t know if I’ve had a scenario where someone asked for real time data, where it was actually needed in real

Eric Dodds 22:09
time. That was going to be my question, John, have you run into a true real time use case?

John Wessel 22:16
I don’t think so. I can’t think of any off the top of my head. It depends on what you’re doing, though, because if you’re in reporting and analytics, I’d say almost never. But there’s other things that are not really reporting and analytics that can cross over, like, I work, I like working with a client right now that’s basically written as an analytics type add on, like, they’re using a really old legacy system. I can’t update it. There’s all these like restrictions around it, so they basically wrote like a reporting add on to it, and you can also edit data through that interface. So now it becomes like, does it need to be like, pretty real time? Like, yeah, it does, because they’re editing data and save it because it’s a full featured app now, so to me, that’s like, the differentiation between like, is it read only, which Reporting Analytics that’s like 99.5% that’s how all that, if it’s read only, doesn’t need to be real time. I mean, unless you’re trying to, like, use it to physically monitor something, if you’re actually monitoring something with data through a reporting app, that you just need to know exactly what’s happening. But no, and that’s really rare. I haven’t seen it much at all, right?

Matthew Kelliher-Gibson 23:24
It’s also dependent on a lot of other services in that chain, usually. So Right? A lot of times you’re not going to be fully self contained at every step, and that’s going to put a damper on your ability to be real time anyways, on a lot of things. I mean, like, I’ve worked at places where, you know, we sent data out to get it, you know, appended and matched, and things like this. So it’s like, that would be, you really couldn’t do that in real time. Sure, you can do certain things where you’re like, Okay, if it’s an obvious match with this, then we can move it forward or something, yep, that you’re not going to be doing none of that, right?

John Wessel 24:00
I think there’s, I think the use cases, and they don’t, they just don’t always end up in reporting and analytics. Like, I’m thinking of one right now for sales, where it’s like, hey, I want to notify sales people as soon as, like, this prospect returns to the website. Like, timeliness on that, like, near real time is really helpful and important in that, which is more alerting and monitoring, to your point, yeah, it’s more an alerting and monitoring, but in a sense, like you’re capturing, like, data from a website and then streaming it, right? Part of an analytics pipe. Yeah? It’s part of yeah, you would Yeah, the data would run through a pipeline that also serves analytics use cases, right? Which is, I mean, I think part of the point here where we’re talking about, you can decouple these things, right? And I can have different components of the thing. Make this like, near real time. Make this not like, that’s cool, like, understand that, yeah.

Matthew Kelliher-Gibson 24:52
And I think other people have done some of that stuff where you kind of have, like, there’s the normal pipeline it goes through, and then there’s the Hey, when this. Person leaves the store, I want to be able to ping them right? And that just has its own thing that bypasses everything to get it more real time in there. Yep, right? So, I mean, if it can help with that, might be, you know, it might be useful in that sense. But, I mean, the majority of the cases that you’re gonna run into are one of those where, you know, it’s like the meme, if it’s like, Can I have this real time? Well, what’s it for a monthly review? Yeah,

John Wessel 25:22
QBRs. It’s for QBR Well,

Eric Dodds 25:26
I mean, the like, the, you know, the most the real time reporting. When I think about real time reporting, I think about refreshing a Salesforce dashboard, which there are limitations. It’s funny that we talked about Salesforce dashboards earlier in the show, but yeah, it’s getting towards the end of the month, right? It’s the last day, right? You know, the sales team hasn’t purchased an agent for so they’re having to go in and update opportunities manually themselves. Yep. And so, you know, as a marketing leader, I’m refreshing the dashboard as fast as sales. Let me refresh it. Yeah, to see, like, you know, how close am I going to be to the number, like that, 15 minutes, you know, or whatever it is,

John Wessel 26:07
yeah. And you can imagine a scenario where you’re not using Salesforce, or, for some reason, there’s some other outside data you want with the Salesforce data, where you’d want that, right, like, near real time in that, yeah, yeah. I

Eric Dodds 26:17
I mean, I think that, you know, I don’t, I haven’t. I’ve never, that was the first time I’ve heard of twirl and sort of the orchestration. It is actually interesting to think about the orchestration layer as a way to set up, you know, different set up pipelines that run at different cadences for different uses. Like that I’m sure there’s a lot of utility there, but we’ve had multiple real time vendors on the show. We’ve had companies, you know, data practitioners come and talk about real time, and I think to the point that both of you are making the actual use cases more rare. I can’t remember the name of the company, but there’s a company that does financial information. It’s like, stock ticker stuff like, you know what? That’s real time, you know. And they, like, have a super, a deep Haven. Is the company super cool product, right? But, you know, they come from the world of finance, and they have a lot of customers in the financial world, and it’s like, yeah, I mean, they literally are doing, you know, serving all sorts of real time use cases with, you know, giant feeds of stock ticker information that are literally happening. You know, there’s changes happening all the time, you know. So it’s legitimate, but I do have this slight feeling of throwing a really cool solution to a problem that isn’t practical for a lot of companies.

Matthew Kelliher-Gibson 27:36
Yeah, exactly that, John, you did bring up something beside this point, that is one of my favorites, which is, when you have a legacy system and everyone gets caught up in we got to replace it. We got to throw it out. And it’s like, Nah, let’s just hollow it out. Yeah, you just suck around something, if

John Wessel 27:52
somebody does a good job of that, like, I think it’s one of the best solutions often you can do because, especially like in a large company, where it’s going to be millions and millions of dollars to rep out a system and replace it, to hollow it out, like it really can be good well,

Matthew Kelliher-Gibson 28:06
and depending on what it is. I remember one place I worked, we actually ran into this where it was like we got to completely replace this blah, blah, blah. And my team, we started using it. And one of the things we realized was the data model inside this system was actually really strong and straightforward, made sense, like, wasn’t overly complicated, but it was well made. And we were like, well, this works the way it is, so let’s just sit something on top of it. We’ll do all the work it won’t do, and then we’ll just drop it in as basically storage at the last minute when we need it. So it’s just an underrated thing like hollowing out legacy systems under one of those things that people get very caught up in. No, it’s got to be this big fancy now it doesn’t have one.

John Wessel 28:46
One other thing, though, and I think it’s easy as a data practitioner, to kind of downplay real time because it’s hard, and we don’t want to do it. So I want to call that out. But I do think if they were easy, the same amount of easy and the same amount of cost. I do think there’s a human psychology component that probably long term will help adoption of reporting, analytics solutions, if they were all like real time by default, long term, okay, the

Matthew Kelliher-Gibson 29:16
The problem is they don’t scale that way. Sure. The cost is the thing that makes people, yeah, well, no, we can have this once an hour or once. But I do think it was a

John Wessel 29:24
psychology thing with some of that like you’re saying, with the like, the like, nobody’s gonna go to their Tableau or whatever dashboard at the end of the month to refresh, to look at the number, because they have to wait till the next day they’re gonna log into Salesforce and look right,

Eric Dodds 29:35
right. Okay, but that brings up a question. So do you okay? I agree with you, right? If that could increase adoption because of the dopamine hit, yeah, right, because of the dopamine hit, right? Okay. And I actually, I mean to your point, Matt, like there are cost implications of that, but those are increasingly going away, right? Like Tech. Technology is getting to the point and at a huge scale, there’s obviously so many issues, right? But there is a path towards that, you know, towards it becoming, like, more reasonable from a cost standpoint, yep. I mean, even some of the stuff we’re working on internally, you know, this is pretty cool. But is it healthy, right? Do you want to? Of course, it’s going to increase, like, adoption of analytics. But Matt cynical data guy, is that help like, do you want to Oh,

Matthew Kelliher-Gibson 30:26
I mean, I’ve built stuff for sales teams before. So my big thing with them, a lot of times, is like you’re chasing noise. Like, just let it you know, it’s like you will make better decisions if you’re looking at it once a week than once an hour, because you’re just gonna get very hung up on some of that stuff. I mean, though, if you really wanted to increase engagement, which you should do, like Rory Sutherland, who does marketing, talked about this. The problem isn’t that people need it real time. It’s they’re not sure when it’s gonna come, so have it go every 10 minutes with a countdown clock. Yeah, that would get people doing it just because they do it a lot of times also, like, when we think it’s, well, I need to have it now. It’s not that I need to have it now. It’s, I don’t like the uncertainty of when I’m gonna get based this

John Wessel 31:10
the freshness thing, right? Like, if I knew exactly how fresh this was, when it would update again, when it last dated again, because BI tools are pretty bad at communicating that. Yeah, most. So that solves a lot, and a lot

Matthew Kelliher-Gibson 31:21
teams try to hide that fact. Do you like the Oh, yeah, of course. They

Eric Dodds 31:24
I don’t want you to know that. This is two weeks old. Yeah. As far, I’m surprised you didn’t have more of a cynical like, we can do a Pavlov’s dogs thing where, like, you have to update your opportunity information, and it’s a random number of clicks that actually triggers the report, refresh? No, I

Matthew Kelliher-Gibson 31:43
i think you get enough people addicted just by doing the like, oh, look, it’s gonna update in three minutes. I could walk away, but now I’m gonna sit here and wait for three minutes. Oh, no change, but it’s gonna be another 10 minutes. Now I gotta sit and wait for that. Yeah,

Eric Dodds 31:57
the countdown. Oh, I

John Wessel 31:59
get behind that. I mean to answer your question, Is that healthy? Like, I mean, yes and no, yes, and that you want people to care about the numbers. And like, that’s still hard to do in a lot of companies, yeah, to really get them to care about the numbers, no, and that. Like, there’s another extreme of like, I want you to care about the numbers, but I mainly just want you to be selling or marketing or whatever. Can

Eric Dodds 32:21
I point out that we are just both of you? I mean, I made several jokes about sales, but both of you went straight for sales, like the joke. We launched this

John Wessel 32:31
off with sales for that is true. We

Matthew Kelliher-Gibson 32:36
I can go after marketing too. That

John Wessel 32:38
is very we normally go after marketing, I feel that’s true. Yeah. And sales, yeah, you

Eric Dodds 32:43
know what? It’s the end of the month. It’s actually the end of the month for a lot of SaaS companies. It’s the end of the quarter, true. And we started off with Boss, Beni off at the beginning. So I think that was appropriate. I stand corrected. Well done, gentlemen. Well done. All right. Well, that’s all the time we have for today. Tune in next month. We’ll pick on marketing and pick another couple great LinkedIn posts for ya, and we’ll catch you on the flip side. Stay cynical. Yeah, the data stack show is brought to you by RudderStack, the warehouse native customer data platform. RudderStack is purpose built to help data teams turn customer data into competitive advantage. Learn more at rudderstack.com.