The Data Stack Show - 247: Consulting Firms, AI Strategies, and the $100K Question with the Cynical Data Guy

Episode Date: June 4, 2025

Highlights from this week’s conversation include:The Cynical Data Guy Returns (1:11)Klarna's AI Shift (2:08)Sensational Headlines in Tech (4:31)Impact of AI on Jobs (8:19)Cost of AI Strategy (10:21)Consulting Strategies and Company Differentiation (12:26)Brand Recognition in Consulting (14:48)Critique of Outsourcing Thought (17:10)Future of Consulting with AI (19:37)The Fragility of Data Systems (21:32)Primitive Data Stacks (24:05)Cynical Data Scientist Epitaph (00:26:06)Rebranding Data Science (27:40)Parting Thoughts and Takeaways (30:14)The Data Stack Show is a weekly podcast powered by RudderStack, customer data infrastructure that enables you to deliver real-time customer event data everywhere it’s needed to power smarter decisions and better customer experiences. Each week, we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.

Hi, I'm Eric Dotz. And I'm John Wessel. Welcome to the Data Stack Show. The Data Stack Show is a podcast where we talk about the technical, business, and human challenges involved in data 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. How to Create a Data Team with RutterSack Before we dig into today's episode,

we want to give a huge thanks to our presenting sponsor, RutterSack. They give us the equipment and time to do this show week in, week out, and provide you the valuable content. RutterSack provides customer data infrastructure and is used by the world's most innovative companies to collect, transform, and deliver their event data

wherever it's needed, all in real time. You can learn more at ruddersack.com. Welcome back to the Data Sack Show, one of our favorite monthly installments, actually our only monthly installments, is where we go deep into the bowels of corporate data America and talk to the cynical data guy about recent topics in the news. Matt, welcome back as always. Thanks for having me. So Matt, you're a couple years out now. Like

at what time does your like, was there an expiration date on your cynicism with corporate America? Are you just not enough here? You are. Yeah, I guess you have been expelled. You have been expelled. Yeah, but the PTSD is always there. It's forever. Yeah, true. It's in data therapy forever after this. Yes, yes. We got it. And, you know, just sitting down for half an hour at a mic is a lot cheaper than,

you know, actually talking to a licensed therapist. Yeah. Yeah, I'm gonna do that. Synthesis therapy doesn't work. We this can't be for you, obviously. And by works for you, I mean, just reinforces your sense. We're glad we can be part of your process. That's really the most important.

That's the whole reason you do this is to see smiles on. For the people. OK. Speaking of people, there was big news with Klarna. So it must have been a year ago or so. In this company, Klarna, they came out and they said, hey, we are going AI first with our customer support, made a bunch of news.

This is one of the first sort of big stories around a company publicly making a move and announcing plans to stop hiring customer support and using AI. I'm gonna read the headline from just a few days ago. If you don't know what Klarna does, what does it do? Yes, Klarna. Add it's like a pay as you go, primarily. Installment, like afterpay.

Yeah, sorry, Clarno's FinTech. So you can essentially finance purchases typically on a smaller scale. Buy now, pay later type stuff. Yep. Here's the headline. As Clarno flips from AI first to hiring people again, a new landmark survey reveals most projects, most AI projects fail to deliver. So I would describe that for those listeners who are not in the studio, which is everyone, I would describe what Matt just did as a snide chuckle. It just, his face is turned away from the mic in disgust. No, not disgust. It's one of these things. I'm actually glad they did this. They did the whole, we're going to go completely by first thing because it gave us a data point

of like, okay, could you really do it? And what they found out, it appears, is that the quotes by the CEO are like, well, it's really important to have a human touch on these types of things. Which I think goes to show the idea of like, you know, we're not completely ready for AI to do everyone's job yet. Is this though, the... Is this though kind of like the sensational self-driving car story? You know the big one where the guy was on his way to the airport and I think it was a Waymo and it circled for like half an hour in a parking lot

and he had to call support. And of course it makes all of this news and in reality Waymo's using self-driving cars to deliver people all over the place usually works. Yeah, is it that though? cars to deliver people all over the place. You know, usually works. Yeah. Is it that though? John, is that, you know, is this is this just a sensational headline? I think that I'm not aware of any business change, technical or otherwise, that works

when you do it that drastically that quickly. It's just, I don't know how that works for anything. I don't care what you're doing. You can reorganization like drastically and like have it fail and not, you know, you think of a change. I don't know how that works for anything. saying, you know, all of you Luddites are falling behind from all of the AI hypesters out there. So like, yeah, maybe it's a little over the top. But there's also probably a bit of schadenfreude going on here since all those people are not going to be mentioning this. It's just going to pretend like it never happened. Yeah.

Well, and the question I want to know is, what are they going to use AI for as they rehire? Will these people be using AI? Are they going like hard, like not going to use AI for as they rehire? Will these people be using AI? Are they going hard, not going to use it at all? Yeah, I think they said they're still going to be investing in it. So, we'll see. I mean, the human in the loop seems to be the sweet spot at the moment. Who knows how much it'll stay there? But I read one person who actually tried it out when it first started, and their comment was,

it just read back verbatim lines for the manual and quickly shooed you off to a human anyway. So, you know, it was when you jumped that quickly, and as John said, like, you're not really vetting it well. So it's a hard one to come through. And I even suspect that with a company like this, it's like, okay, I wanna implement AI, great. Like, what do you want it to do? I don't know, our customer support. It's like, well, what are customers contacting for? I don't know.

Like, there's a component here where like, let's say you're in an e-commerce store and like half your request are like, where's my package? Like, you could do that without AI, but like, sure, maybe it's a little bit more of a personalized touch. And they say, hey, where's my package? And you could do that without AI, but like sure, maybe it's a little bit more of a personalized touch. And they say, hey, where's my package? And you give this nice response of like,

here's where your package is. And you take away 50% of your support requests. Like I totally believe that's a thing. But the long tail is the problem, right? Of like all the like way more complicated things. I respect that's what they ran into. Yeah, my hot take is that I think, we've had actually people on the show,

I remember we've had actually multiple people who've worked in vision for self-driving cars or other autonomous and multiple people. And we've asked them about, you know, the future is here, it's just not evenly distributed yet. And they talk about the long tail of solving you know, the self-driving issue. And you're seeing it, you know, increasingly materialized, right? And I remember one of the things that, I think it was Peter from Aquarium, I can't remember,

we'll have to look that up and put it in the show notes, but he said the press around it is really bad generally for the progression of the technology big way. the excess workforce. Because if it's even 20% smaller than it was, and then you repeat that across thousands of companies, their size, that's a dramatic economic impact in the total number of support jobs that are being replaced by AI, right? And so these individual instances of this

are I think misleading for, I think for the average person thinking about this stuff, I think, misleading for, I think for the average person thinking about this stuff, it's like, oh, well, that's not going to work, right? And it's like, well, actually, even if it's 5%, in aggregate, that's a really non-trivial like economic impact generally on like, you know, how AI is, you know, is impacting. I've been seeing this is the history of this in data and technology. I mean, I just saw it again the other day with the new Pope.

And there was some company that had done this like, you know, text, textual network analysis, and they predicted who the new Pope was. And everyone's like, it's amazing. It's like, you did it once and okay, it worked out. That doesn't mean anything. Like that doesn't prove anything. It doesn't show that you're Sven Golli or something like that. You've got lucky partially. You picked a, you figured out a unlikely but high probability person and it happened

to work out that time. So we need to have a show on predicting the next pope. Okay, next one is also AI related in my gosh is this a zinger so buckle up gentlemen. And let's see this is from Andrew Amman so I think I'm pronouncing your last last name right but Andrew thank you for this incredible post. Paying Deloitte $100,000 to figure out your AI strategy is like paying someone to teach you to swim by showing you pictures of water. You're never going to get there, but they'll keep promising you the insights are just around the corner by the way.

There's another 50k told to get there. Plus they're stuck in the last generation. And there's this image, there's this image and it has some additional text that is really great. Accenture asking for another 250k after burning through the 500K discovery phase budget because they've identified significant LLM integration opportunities that weren't in the original statement of work. That is searing.

That is searing. All right, how much time do we have to talk about this? I've got thoughts. Six hours. There we go. I'll condense it down to, you know, we'll talk about the big consulting groups, not the little ones here. So Present Company included.

The big ones though. Agreeable data asking for another 250k. I mean, your discovery phase is about a half a million. Yeah, if anybody's out there, yeah, need some discovery, let me know. AI discovery for half a million. Yeah, I just. I will do that for half the price. Matt and I will team up on a special task force. Yes. And just tell you everything that you already knew that you needed to do, but pay you.

But then charge you a ton of money to tell you exactly what you already knew. But the large ones there, they are the cause of and solution to every hype cycle that we go through and will go through. They are the one that I've had to fight basically everywhere where I've been, because everyone just wants to pay them a million dollars to, I don't know, give you a fancy PowerPoint. Or the thing when we're like,

all of these projects have failed and we can't figure out why. Oh, because they all stood up these practices for like AI, like 20 minutes after the first big thing happened. So what were they taking all the money for? To figure out how to do AI that they're still struggling with? Yeah, this is like a stand and applaud one from me. I just Agreeable tape I don't know. I don't know if there is one. I mean, I've actually run into that's right You have corrupted now on this issue. I've run into a number of larger practices in my time and I

think at some point, like I say this, I don't know if this is true or not, but I think at some point that pattern like has to be broken as far as like, maybe it won't be because there's some phenomenal salespeople out there that are just really good that work at some of these, you know, big. Well, I think it's the, part of it is the abuse of the idea of using like industry best standards,

right? Right. The idea of like, well, we can't do anything unless it's just like everybody else. And it's like, do you know how you have an enduring strategy and company by doing stuff different than what your competitors do? But that requires you to not be like, well, but what is, what is McKinley say we should do? McKinley is going to tell you to do what they told the other two dozen companies who asked them this question. Well, I think the problem, I actually, I think the main problem here is the upper mid market companies that want to be in that, you know, fortune 500 list.

And they think getting there is by copying the people than Fortune 500 and the consultants will be like, oh, we can help you copy those guys. And I just don't think that is typically the right move. I don't want to think right now. Maybe ever the right move. Can I pay you a million dollars to pretend to think for me? Well, and like part of the like, part of the strategy here is like,

oh, like, you know, find one of the big, you know, consulting companies like, oh, we've got people that have worked at, take a Fortune 500 company and they did this there, and they're going to guide this team to do this thing, and it'll be exactly like here. And the market timing is different, the environment is different, that's not still necessarily the right thing to do. Yeah. Okay, fine. It's OK. You don't have to name it. Yeah. I'll just say it may have already been named in that post there. And it was like, as he said, they've got that senior person with 20 years experience, you know, for the time for they're there at the opening meeting.

OK, you get done. They walk away the next week. It's five twenty three year olds who have no idea what they're doing. Right. And they're just Googling their way to your solution. That you're paying them a billion dollars for. idea what they're doing. and the person essentially does nothing on the project. And they're just there to push it forward. So I think that, I mean, I think it happens everywhere. I actually, okay, I will own the agreeable take here or slightly agreeable takes.

I just let you guys run with a brutal roasting of the big consultancies. The amount of money they're making, they'll be fine. That's true. Yeah, they are laughing, they are listening to this and laughing as they get their paycheck deposit. Yeah, they're not listening. They're getting the bank.

Okay, so there are two people in particular, I won't name, who I've worked with who have had unbelievably successful careers, even in one of them in particular, like in Silicon Valley, phenomenally successful from McKinsey. Okay, and so they attract incredibly intelligent people. There is a very high concentration of talent. But I think that one of the major drivers

that creates a context for them to come in, and I'm not saying that, you know, they're not overly expensive and all of those things, but especially at large companies, it can be really difficult to get things done internally. Yes. And so it is actually, I think, a strategy for some people who have a personal,

not a personal agenda, I don't mean like a personal political agenda, maybe that's part of it, but like they actually want to get something done in the company or they need to restructure a division or whatever and the political forces are unbelievably strong. And so having a third party come in and to your point, Matt, say what you already know is actually part of the strategy sometimes I think right like having a third party come in and like actually be a voice that is objective third party outside of the organization a political like there's a huge context for getting stuff

done even if you generally know like what you want to get done right yeah and so my my criticism there isn't that they're tricking people to just tell them what they want. It's the fact that you need to do that in the first place. That's what I said. Outsourcing my thinking for me, right? Because you're having certain types of companies where it's like, ooh, I don't know, everything sounds risky. Oh, but you know, Boston Consulting Group told me to do this.

Oh, okay, then I'll do it. It's like, yeah, we've been telling you this for the last two years. I think there are some fixes that the big consulting firms are good at. And it's these like objective, like for whatever reason, there's just an extremely unbalanced company.

Like let's say it's run by like by salespeople and they've just got like billions of dollars of sales and like the whole rest of the infrastructure of the company is crumbling essentially. And like, and they just need basic like, hey, we need some like clear roles about like who does what, you know, fine, like that probably makes sense. And then the other one in these like fortune 500 companies

is we have so much segregation of knowledge and roles. Like we need a consulting company to do this because they're gonna be crossing, I'd have to cross so many different teams and levels of like politics essentially to get this done. Like I have to go with a consultant work group, which is to your point. And I think that's valid.

I mean, it seems crazy, but like that's probably the right move. And I will say, I've known a few people who come out of these groups. And generally speaking, you basically put through the ringer. I mean, I worked with a person who said, if you've never been crying in the bathroom, you haven't been a real consultant. How often do you cry in the bathroom, John? Not very often.

Then you're not a real consultant. I know. Maybe I'll get there one day yeah but anyways but like with that still the other thing that I don't like is a lot of those big ones so they're charging for the name right you're charging they're charging this premium because they can put their name on it yeah get you across yep but they've also in the last however many years or decades worked towards it's not okay we're gonna come in and do this yeoman's work of going across here

and doing that, it's how can we package this to something that we can sell that minimizes our costs that people will pay for. And that's where it really gets into this, we're just telling everyone the same thing. We've created a framework that basically is, everyone is exactly the same. You're like, the idea of consulting

is supposed to be it's bespoke, and they're trying to turn it into a factory method of doing this. Which having a company align on a plan and stick to it and actually just do something is super valuable. And having a brand that you trust to like tell you what to do, could that have come from Chad GBT or anywhere? Like, maybe. But you know, but there's value in that because you can get a lot done if people are all just agreeing on what we should do.

McKinsey bot. Yeah, exactly. I mean, maybe that's the future, right? There's gonna be trained bots from all these kids. Now that's margin, right? So the executive team pays a million dollars for each of the executives to get a, like a tuned

McKinsey bot. Basically they tell them what's going on what they think they should do and then it parrots it back in a more refined language chart. This really hit a nerve with you guys. It's really a nerve. Should we move on? So, okay. I love this one. And I know that both of you are going to have a lot of thoughts. This post is from

Sikhar Neer. Thank you for the post. I apologize if I'm not pronouncing your name correctly, but this is wonderful work. Candidate. This is such a high reputation company. Their data must be so structured and clean. The company. Our list of high value customers is a CSV file maintained by George from customer success.

Oh, that strikes so true. It's just man, what a post. And it's probably cleaner than most people's CRMs. Oh yeah. That's why Joe maintains the list is because it doesn't work in the CRM. He does merges every week, keeps the customers like super clean.

Yeah. Yep. So examples from your experience in similar... Like what type of similar... Oh man. What's striking about this is for all of the discussion that we have and all of the advanced data tools out there, the reason this is funny is because we know that in whatever context they are in, yes, it's a physical file, there are a number of things,

but it probably works extremely well for the company. Just extremely fragile once Joe leaves. Yeah, true, but like- Or when they bring in AI to reduce the head count there and nobody realizes that list was being maintained by Joe the entire time. Yeah, this is a really funny thing I've been thinking about. Like, you know, there's a lot of talk on all this like unstructured data, we're gonna have images and videos and stuff.

I've been thinking about Anna like night, let's say, probably four, 50 years ago type things, and maybe what happened in an office. And like, if we will go back to that for some things, because AI is a thing. So for example, physical sticky notes or something. People still use them, but like people use them more. It's like, I wonder if people will be like, you know what? I love the workflow of physical sticky notes all over my monitor.

And I have like a camera that like captures it and then like processes the data and it puts it in Salesforce for me. Or like a Rolodex, I go back to physical cards, like I like that. And I have this like little machine and then processes the data and it puts it in Salesforce for me. Or like a Rolodex, I go back to physical cards, I like that. And I have this little machine that literally flips through it, takes pictures for me, and puts it in Salesforce. I really think it'll be interesting. I really think it'll be interesting where there's some of these like analog things that will be possible and like not not have the downsides that they would have had where we have some of these

processes like maybe even more analog than a CSV file that like all of a sudden work again because it's like oh like we can do pretty instant here you know here recognition and of course it put where we want to go. I know for a fact that there's some of this stuff with just using AI to do screen reading. Oh, we need to go do an integration with an API. It's a screen. There we go. We just log the AI in and then it reads it and it does whatever it needs to do at that point. But to go back to where you think, oh man, this should really be something here. I remember I worked at one company and it was, they bought another company, right? And the company

they bought was like 20 million revenue or something. Then it'd been around for like seven, 10 years or whatever. And I ended up at one point being part of a discussion because the entire company, $20 million in annual revenue was being run out of one Excel notebook for finance. That it was password protected and that only one person actually could understand how to get the information out of it in order to know like who was up for renewal and stuff like that. No automation whatsoever. It was like, it was just one.

Yes. Or fuck. Yep. I think the, it is interesting though, the sophistication, I think one of the points that I draw from this, because if you have a company that provides a product that solves a really big need or is like a really well-loved product, this isn't 100%, but I can think of multiple cases where companies that all of our listeners would know about,

they have a very primitive data stack. And they're just really good at finding signal in the noise which is interesting. I don't know. I kind of it's an interesting pattern to see. Well, I mean, even if you go like, I mean, I've worked at one company where we got acquired by a larger public company and they had a whole system. They did a lot of manufacturing stuff, too. And it was run off of sticky notes on a whiteboard that you moved through thing. I mean, because it's also like I'm of the opinion. Everyone gets really overcomplicated with a lot of project management.

And it's like you can you don't need special software to do project management. I've run it out of Excel. I've run it out of, you know, like all sorts of different things. It really, you don't need it. And a lot of times when you start buying specialized products for some of this stuff, it makes it more complicated

because you're now trying to take your flow and stick it into this predefined flow and it never really fits exactly right. It's like people get all hung up with it and it's like just pull it out there and just here's a list of things we need to do in Excel. We're good. Yep. Okay, do we have time for a bonus round? Yeah. We're on our producer list right now so that means we can do whatever we want, right?

Producer list. Okay. Anarchy on the podcast. This one is an image that just came across our podcast channel. So I don't even know where this came from. I was just scrolling through messages this week. And it's a picture of a tombstone I'm scrolling through messages this week. And it's a picture of a tombstone with the following epitaph carved into it, okay? Here lies a data scientist, 1995 to 2027, copying Kaggle notebooks,

trained models without understanding the data, optimized for accuracy, ignored business needs, never validated assumptions, never spoke to stakeholders, never moved past Jupyter notebooks. The world needed solutions. He delivered charts. While LLMs reshaped industries, he was busy updating pandas. Man, we got some brutal, had some brutal takes on the show today. Cynical Data Guy, what's your favorite part?

It's probably going to be either optimized for accuracy and not business need, or never got past Jupyter notebooks. I've seen, I mean, like I've literally, I was a data scientist, I've managed them, I've seen many of those things in individuals before. Just even the idea of it's like, I need the model to do this. Okay. Well, if I go look at accuracy, nobody said that. I needed to measure this metric.

Couldn't do it. Could not figure it out. John, your favorite? I think the whole data science. I mean, there's a whole group of people right that are rebranding maybe is the right word right like that are moving away from data science and it's basically because of those reasons because there's a reputation maybe not at every company but

a lot of companies of that of data science being too theoretical, not practical enough, not driving enough business value, etc. And... I'd like to meet these overly theoretical data scientists, by the way, because that has not been my experience. Really? Well, but part of optimizing for accuracy without anything. Yeah, I'm thinking more of the overly accurate. Oh, good. Yeah, no. Yeah. Those are actually fairly rare, too. accuracy without like anything like that.

Yeah, those are actually fairly rare too. But here's the thing, part of that is a problem that we've had for a long time in tech, and that's why product management exists. It does not exist nearly as much in data as it does in traditional development land. So I don't think it's actually fair to data scientists to be so hard on them when the exact same problem exists with engineers and that's why product is a thing. So why can't they just become a product person? Data scientists? Yeah, yeah.

I don't know. I don't know that why. Well, it's so hard. I did. Exactly. But that, I mean, that's reality, right? This is a solved problem in that like, we know how we have solved it here, but like, you know,

people haven't done it as much. And for whatever reason, companies have seemed to like, just like, ah, forget it. Like, we'll just ask sales what the forecast should be. Let's not try to predict anything. Yep. You know, interestingly enough, I know several people. So in Greenville, the forecast to be. Let's not try to And I met a guy a couple months ago, and I can't remember exactly which one he was at,

but we were talking about what he does. He actually is a data product owner at one of these large organizations for exactly the reason that you talked about, right? We're still dealing with companies still don't know what they're hiring for a lot of times. There's still some companies that wouldn't hire product coupled with that, but way less. And they still overly focus on technical skills joining us for another edition of Cynical Data Guy, and we'll catch you on the next one.

Stay cynical. The Data Stack Show is brought to you by Rudder Stack. Learn more at rudderstack.com.