[00:00:00] Whitney Lee: Coté, I'm gonna you to stereotype for me.

[00:00:05] Coté: Oh, okay

[00:00:06] Whitney Lee: Yeah. Well, here's the thing is, uh, again, I haven't been in corporate land so long, um, and I feel like of all like the C-level personalities, that there are stereotypes probably for

[00:00:19] Coté: Mm. Okay

[00:00:21] Whitney Lee: like I'm-- Like for a CEO stereotype, I'm picturing like you, you take a cold shower, 15-minute cold shower right when you wake up.

[00:00:28] Like your, your day is planned to the absolute minute. Everything's hyper, um, hyper-efficient and well thought out. Is that, uh, is that an accurate CEO stereotype or is-

[00:00:40] Coté: Hmm. I'm trying to think of CEOs that I've met. I mean, they're all, they're all, they're all tech CEOs, so they're-- have their own peculiarities.

[00:00:49] Whitney Lee: Yeah

[00:00:51] Coté: you know, I-- let me, let me, let me change it slight- like high, very high-level executives. And, you know, most of the CEOs I've met are very, um... Uh, well, as always, this is gonna sound a little odd.

[00:01:03] I used to have this orthodontist, uh, and, and I, I would, I would of course I would, I, I would go there. And, and the orthodontist, you know, he-- Well, I don't remember his name. He was in Austin, of course. And he had-- He was, he was one of these, these orthodontist types who, it-- well, as I got, as I got older, I kinda was piecing it together.

[00:01:25] And every single, I don't know what you call them, the tech, the people who actually do stuff, was, like, a very young, attractive woman. And he was just-- And, and he was, like, very smiley. He had this big smile. And

[00:01:38] Whitney Lee: Do

[00:01:38] Coté: this is the '80--

[00:01:39] Whitney Lee: Yeah, yeah

[00:01:40] Coté: this is the '80s, so he had that, like, uh, Magnum P.I. mustache. And, and he was always, like, very clean, and he had this very particular, like, smell.

[00:01:51] And it wasn't a bad smell. It was, it was kind of like that, like, like if you could imagine if there's, like, a good smell that is like a, "I'm a medical professional" smell. Like, it smells good, and you know that I'm clean, right? And so it was like... And, and it's not, it's not that, like, that, like, the stereotypical CEO is exact- Dr.

[00:02:13] Barnett, that was his name. It, it's not like this Dr. Barnett guy, but he d- he just exuded this, like, "I've got all my shit together, and now I'm gonna get your teeth together." But, but not in that tone. That's, that's a very, like, Gen X way of thinking about it. But just, like, just, like, a lot of, um, confidence and, and like, you know, "We've got it handled.

[00:02:34] We're gonna do stuff." And maybe, maybe not, like, confidence of like, you know, like a coach, like getting things up, but just sort of like, "Yeah, I'm executizing. Let's put together a plan. Let's have a thing.

[00:02:46] Whitney Lee: Yeah,

[00:02:47] Coté: I've got my team here. Yeah, we're gonna, we're gonna execute and do stuff." And so, like

[00:02:52] Whitney Lee: I've executed and, and done plans on my own life. So

[00:02:55] Coté: Hmm. Oh, I smell, I smell good.

[00:03:00] Whitney Lee: They, they just, you know, wear really clean clothes. Like, I feel like my clothes get worn and I keep wearing them. And

[00:03:06] Coté: Uh-huh

[00:03:07] Whitney Lee: I was thinking more, um, CMOs have the, like, really perfect outfit,

[00:03:12] Coté: Oh, yeah, yeah, yeah. I like that. I like that.

[00:03:16] Whitney Lee: Yeah.

[00:03:17] Coté: Yeah, but then, you know, I, I, I've...

[00:03:19] Whitney Lee: I think are very serious

[00:03:20] Coté: Yeah. Oh, yeah, a little frumpy sometimes, but, you know, they're, they're, they're, they're, they're busy. Uh, you know, I've, I've met pl- I've met plenty of other CEOs who are. There's, there's one I'm thinking of who's a very successful CEO, and they were, they were sort of like a, uh...

[00:03:36] They, they, they were, they were, like, financially savvy with, like, how they're, they're d- doing the finances and everything of their business. But really, they were just like, uh, a, a, a nerd who got lucky and, and kind of like, you know, raised that... You know, and then, and then also had the skills to, like...

[00:03:53] Whitney Lee: are? Yeah.

[00:03:54] Coté: Yeah, yeah, yeah.

[00:03:55] And but...

[00:03:56] Whitney Lee: that

[00:03:56] Coté: then there, there's, you know, there's another one that I knew was, is just like a, a, a, a good old boy from Texas who also kinda got lucky. You know, there's, there's a variety. But they all, they all have that sort of like, that Dr. Barnett thing of like, "We're running stuff here. Ev- your teeth are gonna be fine.

[00:04:13] We got a plan. I got it all set up, and I smell strangely good."

[00:04:19] Whitney Lee: Do you know when you were telling that orthodontist story, you were like, "Here's a guy, he has these, um, he employs these young women. He has this kind of mustache. He's, like, really super clean." I thought the end of that w- story was gonna be like, "Now with my grownup mind, I realize that guy's a total creep." Like, but, but instead, instead the conclusion was, "No, that guy has CEO vibes." So,

[00:04:43] Coté: Yeah. Well, that, that, that was just th- that wasn't part of the, the CEO characterization. That was just a side note to this thing. And, and the point of it being older is when I was a kid, you know, when, when I w- when I was getting braces age and, and, um, appliance, I had an appliance, appliance age, like I didn't really...

[00:05:03] I, I don't know. I, I wasn't as observant and sort of like systems thinking as I was back then. So,

[00:05:08] Whitney Lee: Oh,

[00:05:08] Coté: you know, at some point as I got older, and then to your point, it's sort of like, hold on, what was going on there? And then it's just like, it's a very like very, uh, the smile, the mustache, the smell, the young ladies.

[00:05:22] That's, that's sort of like the, uh, the mind palace of Dr. Barnett. Well, my mind palace where he lives

[00:05:30] Whitney Lee: So I 

[00:05:31] um, so now that we've probably managed to offend a lot of people, let's bring on our guest. Ben, will you introduce yourself please? And why do you think this, uh, topic of conversation might apply to you?

[00:05:44] Ben Savage: Well, I would guess it's because I'm a CEO. So I'm Ben Savage, the CEO and founder of, uh, Veritas Automata. I've done a bunch of other things before. So, um, did a little bit of consulting, did a little bit of time in logistics 'cause my actual degree is in supply chain,

[00:06:01] Whitney Lee: Uh-huh.

[00:06:01] Ben Savage: I'm now a CEO of a tech company with a bunch of developers all over North and South America, and we do AI, and we do blockchain, and we do IoT, and we like to do things the hard way at like the edge where we don't have all the nice stuff that everyone else builds. Um, I think because people say that it's hard and they don't know how to do it, so we go, "Hmm, let's try."

[00:06:24] Whitney Lee: needs to know how to do it. That's, yeah.

[00:06:26] Ben Savage: Well, it's,

[00:06:26] Is, is that one?

[00:06:28] to say no

[00:06:31] Whitney Lee: So the efficiencies stereotype of CEO is not applying to you in this case?

[00:06:36] Ben Savage: Um, no, not in that sense. I would not be a character in, uh, American Psycho with, uh, Christian Bale

[00:06:43] Coté: See, I, I think, I think maybe that's, that's, that's a characteristic of a tech CEO, is that, like, we should try to do the thing. Whereas Dr. Barnett would be like, "Listen, these are people's teeth. We know what's going on. We're not gonna try anything interesting. We're just gonna straighten those teeth."

[00:06:59] Whitney Lee: talk badly about Dr. Barnett, I'm hearing. Coté, Coté has feelings about it. Um

[00:07:07] Ben Savage: it is a, there is a big difference though between like Midwestern tech and like bi- and coastal tech, right? Like here, everyone kind of bootstraps themselves 'cause the

[00:07:16] Coté: Hmm.

[00:07:17] Ben Savage: environment isn't here. So you, you find yourselves doing a lot more because not people knocking on your door to give you bunches of money, so you gotta figure it out

[00:07:27] Whitney Lee: Oh, two questions. Where are you when you say here?

[00:07:30] Ben Savage: Uh, I'm saying Midwest. So I'm in Cincinnati right now.

[00:07:34] Whitney Lee: Okay

[00:07:34] Ben Savage: like, you know, anywhere between, let's call it of like Philly and skip over Denver and Chicago and like west of Vegas probably, Salt Lake City, somewhere in there. You

[00:07:45] Whitney Lee: Oh, wait

[00:07:46] Ben Savage: VCs don't usually go

[00:07:49] Whitney Lee: OH.

[00:07:50] Ben Savage: .io

[00:07:52] Whitney Lee: Um, the second question why? Again, why? Like, um, if, if you can go to a coastal place and people will knock on your door with money, why stay in Cincinnati, Ohio?

[00:08:06] Ben Savage: 'Cause I think there's a lot more here than people realize. There's a lot of different tech. It's different than what people think. It's not, you know, it's not always software. There's usually hardware, there's physical things, there's stuff you're building. There's lots of really smart people here who don't want to go travel to different places. Um, grew up with this kind of, "Hey, if we don't know it, we'll figure it out, or we'll figure out a path to go do it and figure out..." You know, it's that term bootstrap. You know, there's a lot of people out here who bootstrap themselves. So that's just what I've grown up with, and it's like, why not embrace it?

[00:08:40] Whitney Lee: Are your clients other Midwesterners?

[00:08:42] Ben Savage: Um, our customers are all over,

[00:08:45] Whitney Lee: Mm-hmm.

[00:08:45] Ben Savage: that really varies. Um, we do do a lot in this area, but we do a lot in life science in Philly and on the coast where those companies are. Um, you know, we're, we're, we're not, um, we're not biased to where it is. We just like to solve the interesting problems

[00:09:00] Whitney Lee: I understand you're a CTO too, is that correct?

[00:09:03] Ben Savage: I have multiple jobs. So I am CTO at a different business

[00:09:06] Whitney Lee: Uh-huh

[00:09:07] Ben Savage: um, that's a company that we spun, uh, Veritas Automata out of. So that's a company called Apex Order Pickup Solutions. So there, um, do order pickup. So it started in industrial, so there we were doing large distributed IoT systems that were in, I think when we sold the industrial part of the business, it was like 120,000, 110,000 devices out in the field. So if you've ever dealt with IoT, like that's starting to be a useful number. Like, okay, now I have issues because all the devices call home and I've gotta go deal with it. Um, then as they moved into food, we've done, uh, the project for Little Caesars. So if you've seen a Little Caesars pizza portal, one of my teams did that.

[00:09:46] We've got other projects, um, that are doing other restaurant brands and things so that way you don't have to wait in line 'cause who likes waiting in line?

[00:09:55] Whitney Lee: So, so, first of all, d- do you think any of those st- stereotypes apply to you as CEO or CTO?

[00:10:03] Ben Savage: Um

[00:10:06] Whitney Lee: Do you, do you project an aura of having your shit together?

[00:10:09] Ben Savage: Most of the time because that's what you gotta do to get everyone there, is you gotta say, "Hey, there's an answer for this. We may

[00:10:14] Whitney Lee: Uh-huh.

[00:10:14] Ben Savage: it yet, but we're gonna go find it. Don't panic." You know, if the systems are down, you can't go, "Hey," as a CTO, "Hey, go panic. Amazon went down today." Mm-mm. Not a valid answer. So it's like, "Hey, okay, here's what the team is doing. Here's how we're gonna go work the issue." And if you're CEO, someone comes to you with a challenge or there's an opportunity and you've gotta move through, I think both you can't be what a lot of people think of engineers or even some of the super well put together C- CEOs because you don't have enough data most of the time.

[00:10:44] So you've gotta make a decision

[00:10:45] Whitney Lee: Mm-hmm.

[00:10:45] Ben Savage: limited data, and that's part of what you have to go refine as you think about it

[00:10:50] Whitney Lee: Do-- Are you conscious about projecting confidence? Like how much of it is intentional and how much of it just kind of happens?

[00:10:59] Ben Savage: It's a little of both, right? You've gotta read the audience you're working with and figure out what message you need to go convey to them to go achieve the final goal

[00:11:08] Whitney Lee: Cool. So then this pizza portal business or this pickup solution, I'm imagining, imagining cubbies that keep food warm that your phone-- you can unlock with your phone or something along those lines

[00:11:23] Ben Savage: Most of the time you don't need heat because an order only sits, you know, when we look at the data, maybe four to five minutes.

[00:11:29] Whitney Lee: Okay

[00:11:30] Ben Savage: out your thermodynamics hat on and not your emotional hat on,

[00:11:33] Whitney Lee: Uh-huh

[00:11:33] Ben Savage: by the time you've got... If you've got insulated packaging,

[00:11:36] Whitney Lee: Okay

[00:11:36] Ben Savage: doesn't actually get through the insulation that's keeping the heat into the food product itself. Most people feel emotional about it though, so sometimes you do have to put heat to make them feel better.

[00:11:45] Whitney Lee: Ah.

[00:11:45] Ben Savage: Um, you know, it's-- but that's all part of what you're doing, 'cause sometimes what you're doing is you're creating a experience when you're doing food pickup. just a, "Hey, I'm doing an operational thing."

[00:11:55] And that's something we had to learn coming from industrial into food. We think it's, "Hey, how do we make this more efficient? How do we make this better?" Well, there's an experience to it, and there's a service. So you've got to think about how do you build some of that in, um, as

[00:12:08] Whitney Lee: It make,

[00:12:09] Ben Savage: with it.

[00:12:10] Whitney Lee: it makes sense to me, like the way, I don't know, like a Yeti mug keeps your hot hot for a long time and cold cold for a long time without adding anything, that, that, that would be the solution for this. I'm interested in this idea, this idea that emotionally you might need to make it seem warm because people have feelings about it. there's gotta be a story behind that

[00:12:32] Ben Savage: Well, it's, it's psychology, right?

[00:12:34] Whitney Lee: Yeah

[00:12:35] Ben Savage: better when they think something is there, 'cause they're not necessarily thinking about the whole experience.

[00:12:39] Whitney Lee: Uh-huh.

[00:12:40] Ben Savage: if I'm doing food pickup and

[00:12:42] Whitney Lee: Uh-huh.

[00:12:42] Ben Savage: picking it up,

[00:12:43] Whitney Lee: Uh-huh

[00:12:44] Ben Savage: I pick it up out of a warm compartment,

[00:12:46] Whitney Lee: Uh-huh.

[00:12:47] Ben Savage: I feel that it's done something. There

[00:12:49] Whitney Lee: Yeah

[00:12:49] Ben Savage: to prove to me it has done anything, and in

[00:12:52] Whitney Lee: So

[00:12:52] Ben Savage: cases it may have melted my milkshake. You know, if you think about it, right, that's the challenge.

[00:12:57] Whitney Lee: So are,

[00:12:58] Ben Savage: to do ambient

[00:12:59] Whitney Lee: are you preemptively doing that or are you choosing we need to warm up our cubbies based on data?

[00:13:04] Ben Savage: Uh, I am avoiding putting heat in because I think it provides a cost and a maintenance issue, and it doesn't really provide a benefit. But if I have to work with a customer that requires it, we'll figure it out

[00:13:17] Whitney Lee: Oh, so it's more like the restaurant customer with the feelings maybe than the end user customer who's actually picking it up

[00:13:25] Ben Savage: And that industry's changed, right? 'Cause it used to be maybe you going to pick it up. Now a

[00:13:29] Whitney Lee: Uh-huh

[00:13:29] Ben Savage: of times it's Uber or DoorDash or, you know, Just Eat or wherever you are in the world. And at that point, they probably don't care, and

[00:13:39] Whitney Lee: Yeah

[00:13:40] Ben Savage: get wherever they're going 'cause they're gonna stop or do something else. And my job then is just to make sure we've got data to know they picked it up, and so

[00:13:48] Whitney Lee: Interesting. I'm also fascinated to hear it only takes people four to five minutes to pick up their order when it's ready. I, I don't know what to ask or say about that, but I, I don't know. I, like, I don't, I don't order a lot of food, but I, um, uh, that just surprises me that it times out like that. I guess people are hungry.

[00:14:08] Ben Savage: their order, right?

[00:14:08] Whitney Lee: Yeah.

[00:14:09] Ben Savage: and wait most of the time. If you're ordering food, you're not ordering it for, "Hey, I wanna go let it sit for three hours." Now,

[00:14:14] Whitney Lee: Oh, yeah.

[00:14:15] Ben Savage: hit traffic or you get a meeting,

[00:14:17] Whitney Lee: Yeah

[00:14:17] Ben Savage: it pretty well. And again, this is back to everything we do.

[00:14:21] There's a certain element of psychology in it, whether you're leading people or you're building product, right?

[00:14:25] Whitney Lee: Uh-huh

[00:14:26] Ben Savage: so there it's working with our customers to make sure they build their messaging stream to say, "Hey, we've received

[00:14:30] Whitney Lee: Ah.

[00:14:31] Ben Savage: order. We're working on your order." And then I can tell them exactly when it's ready for the handoff so they can

[00:14:37] Whitney Lee: Okay.

[00:14:37] Ben Savage: to build that reinforcement

[00:14:38] Whitney Lee: So it's more that four to five-minute window reflects the good communication in the cycle of the food prep and getting it ready and firing messages at certain time in the chain of pr- yeah, putting in the oven or whatever. Interesting. Okay.

[00:14:54] Coté: So, uh, y- in, in, in your, in your, uh, your, your, I don't know, stuff you do in software world and stuff like that, like, like in the broader software world, like one of the bigger problems is, uh, we've got this great new technology and the, the way that the organization that would be using it, uh, are structured, the process they follow, like just doesn't really take advantage of, of what it is.

[00:15:15] Like, I don't know, you've got silos or they don't change how they work. And so this is kind of like, uh, I, I don't know, I would call this a, uh, one of the big problems in enterprise software, something, something like that. Now, and, and then, um, as, uh, so with that as a template of like, this is one of the, you know, solving the problem to be solved is, is an issue.

[00:15:36] Like how do we, um, what's a really boring enterprise thing? How do we make sure that the-- we do purchase orders well? And so you gotta solve that problem. But then there's also the problem of like, when the customer buys the thing, now they have to use it effectively, otherwise they think the product is bad.

[00:15:53] And that might entail like changing process. So I wonder in like logistics, like what are similar problems in logistics? 'Cause, you know, to be naive with someone who, who has a background of that, like I, I always assume in a good way, logistics is like, I gotta get this thing from one point to another point, and like planning that all out and doing it.

[00:16:13] And I'm sure there's a lot of difficult problems there, but like, what are the, these kind of like secondary things that like kind of ruin it, that you have to pay a lot of attention to? Kind of like heating up the food even though you don't need to.

[00:16:26] Ben Savage: Well, think about it if you're a B2B customer today, right? And you are ordering a shipment on a-- need to book a, a freight load, and maybe it's not a full truckload. So when you go book freight, you can buy a whole trailer or you can buy a portion of space on a trailer. If you're a portion of space, you're going through a broker of some form.

[00:16:44] Might be called a 3PL, a 4PL. So they're gonna go find a driver. Maybe it's one of theirs, maybe it's someone who's kind of in a marketplace, and then they're gonna have to arrange for them to go pick it up. They're gonna have to arrange for it to get there. And then along the way, part of this process is managed, it's not high-tech sensors.

[00:17:00] It's what they call check calls. So there's someone calling in and saying, "Hey, where are you today? Did you pick up the load? How much longer do you have? What's your ETA?" So there's all this friction that doesn't feel good to anyone, and nobody really knows when something's gonna get there. You know, had an experience with one project we were working on where we shipped something to LA and back to, uh, test the durability of the product.

[00:17:20] So you, you ship it on bad roads. thought, "Okay, it'll take 10 days." Took 30, and it got lost in multiple warehouses around the way because there's not good data in it. So in an example like that, it could be as simple as, hey, how do we put some better sensors to give people data on it so they know where things are and you can go communicate?

[00:17:40] Um, there was a really interesting podcast from Planet Money, from NPR, um, in the last week on the Domino's Pizza tracker,

[00:17:48] Whitney Lee: Yeah

[00:17:49] Ben Savage: about how much of that was the psychology of just telling people when something might be ready, which reduced their anxiety, which

[00:17:56] Coté: Hmm.

[00:17:56] Ben Savage: them actually pick back up some of the volume they lost when they could no longer have drivers racing around trying to meet a 30-minute delivery time window

[00:18:05] Coté: So is, is that, is that like the, uh, to use one word to, for all of that stuff, is that like tracking the status? Is that actually like a huge part of logistics? Like when you go get a degree in logistics, is it like this is the third class you're taking on calling people to find out where your shit is?

[00:18:22] Like

[00:18:22] Ben Savage: if, if it-- ish. I mean, that's like, that's somewhat kind of what the marketing piece is sometimes and like the customer service piece. Like, if you're getting a degree in it, there's a couple different variations on more business or more engineering oriented,

[00:18:35] Coté: Mm.

[00:18:35] Ben Savage: it's gonna be in planning routes and doing all that and, you know, making sure that Google Maps or something has planned an optimum route and figure out how you build that in so you're making efficiency from your vehicles. And then a lot of it, there's usually a lot of negotiations and finance courses, 'cause you're having to figure out the business side

[00:18:51] Coté: Right

[00:18:52] Ben Savage: or how you're pricing things and all that stuff

[00:18:55] Coté: And then, and then, and then so, so it seems-- So what you were describing, like you have multiple parties with different incentives involved. Like it's not all like one organization. And so is there like, is that, is that like an advantage that like a FedEx or a UPS or a DHL has, is that it's kind of all in-house instead of being like spread out?

[00:19:17] I mean, you still have to do coordination, but maybe it's a little different since you all-- Maybe there's some logistics term for it, but you all wear the same color or something, right? Like you're all on the same team.

[00:19:26] Ben Savage: Well, you, you pay some-- You pay, like if you're a third par- if you're internal, like let's say, let's say I'm Amazon and I've built my own trucking system, yes, for them there's a ton of advantage for that. If I'm using UPS or FedEx, there's some benefit to them internally, but maybe

[00:19:40] Coté: Oh, right, right, right

[00:19:41] Ben Savage: benefit as me as a customer to go use their service.

[00:19:44] Coté: Right. So there's kind of like gradients of it

[00:19:46] Ben Savage: I'll pay a premium to go do that.

[00:19:48] Coté: Yeah

[00:19:49] Ben Savage: to bring that all in-house and have all of that trucking and now air and water, yes, there's a huge advantage for them 'cause they can be much more efficient in how they do all their route planning

[00:19:59] Coté: And so, so and, uh, while, while we're on this exciting topic of logistics, like, like, what, what, uh, like how did like COVID r- change logistics a lot? Like, it, it feels like maybe it forced, uh, i- i- innovating a little bit more. Like, you know, you had like, uh, I think at one point, like someone like Gartner or Forrester had like five types of getting your stuff from the store.

[00:20:23] Like you could order it, you could have curbside pickup, you could have in-store pickup, and, and like it seems like all of that is available now. Or like, is it because of COVID stuff, or was that just already on that track?

[00:20:37] Ben Savage: so I'm gonna give three different answers to that 'cause

[00:20:41] Whitney Lee: Love it

[00:20:42] Ben Savage: it-- logistics isn't, there's not one thing that's logistics. So it's, it's, you know, so if I am in a carrier perspective, that is a highly cyclical industry to be UPS, FedEx, or like a shipping company like Maersk or, um, some kind of rail company because it takes a long time for them to add capacity.

[00:21:03] So they add capacity when price is high,

[00:21:06] Coté: Mm.

[00:21:07] Ben Savage: bullwhip effect, if you guys have heard of that, where, okay, now it's dropped off, so what do I do with all that excess capacity? Then it rises again. And so that industry tends to be highly cyclical. So sometimes where you see changes are driven by both something like COVID and a change in a cycle.

[00:21:23] Sometimes that change because of something like COVID may not impact it because they were already at a low or a high point on a cycle. So during COVID, one of the big problems was actually containers 'cause there was so much coming into the US and there were no containers going back to China

[00:21:39] Coté: Mm.

[00:21:40] Ben Savage: imbalances of where they were.

[00:21:41] So that impacted pricing then. So that drove it from like an industrial perspective. Now you're seeing a different thing in 2026 because of tariffs and other things, and that's causing trade to move around the world and different things and unpredictability in it. if I'm a consumer, like that kind of logistics, like that drove a massive in-increase in buy online pickup in store, whether that's in a restaurant or in a, um, in a, uh, retailer.

[00:22:09] And if you think about from March of 2020 to call it August of 2020, in Mar-- at least here in the area where I was, March, maybe 25 to maybe 40% of places had some kind of buy online pickup. By August, it was at like 85%.

[00:22:29] Coté: Mm-hmm. Right

[00:22:30] Ben Savage: a massive technology adoption cycle there, and it accelerated things like Uber and DoorDash and some of that from a delivery perspective. And sometimes different segments of the market get impacted by these very differently just because of is gonna be the driver and how you go fulfill that need.

[00:22:49] Whitney Lee: What were the three answers? I got two

[00:22:52] Ben Savage: Um, the third would be kind of some of the pricing and stuff,

[00:22:57] Whitney Lee: Mm-hmm

[00:22:58] Ben Savage: which some of those models in the software would change, um, based on how the demand or the other thing is. Sometimes you get a real physical imbalance, and then you've got the how do I do what Coté was saying, like, how do I use more software to make this smarter? that sometimes gets driven by some of those changes as well. But some of those groups are slow to change, particularly on the big carrier side.

[00:23:23] Whitney Lee: How is it that you know so much about and physical infrastructure? Like, not technology, but actual moving real shit around

[00:23:33] Ben Savage: so I worked at Maersk as my first job. So Maersk at that point was the largest carrier, uh, shipping carrier, sea carrier. And then because I've had engineering and product in my life most of the time, if I'm gonna build a physical good, I gotta figure out how to get the parts everywhere and figure out the supply chain stuff

[00:23:50] Whitney Lee: Do you find that one has informed the other?

[00:23:54] Ben Savage: Um, it definitely impacts like how you think about product development, 'cause depending on what you're gonna do, you're gonna think about where do I wanna base my manufacturing and what's my supply chain? And then if I'm going into certain regions, I have to know, okay, there may be a tariff barrier there, so maybe I wanna do a design. I can go manufacture in multiple places to go work around that or figuring out those pieces.

[00:24:16] Whitney Lee: Mm-hmm.

[00:24:16] Ben Savage: And then the dirty secret from a software perspective, a routing algorithm in logistics is

[00:24:21] Whitney Lee: That's

[00:24:21] Ben Savage: much a search algorithm for computer science

[00:24:24] Whitney Lee: It's kind-- Yeah, this is feeling similar. Like, oh, like, um, what do you call it? A barrier, tariff barrier is like a, some sort of firewall, and like doing multiple at once is like running things in parallel as opposed... Yeah. So did, did the same parts of your brain fire off or do like, do-

[00:24:40] Ben Savage: does because so much of it they talk about like Kanban. So I learned about that from supply chain in the physical world,

[00:24:46] Whitney Lee: Uh-huh

[00:24:46] Ben Savage: when I saw people in the software world doing it, it's like, "Oh, okay, I'm familiar with this."

[00:24:50] Whitney Lee: Yeah. And I, I guess we in software literally call it supply chain based on the physical world stuff of it. But I, I've heard, I've heard it talked about in theory, but I don't think I've known anyone before this moment who actually has experience with the physical world stuff. So it's interesting to hear that it is still a pretty direct comparison.

[00:25:10] Ben Savage: I think it's a very direct one. And, you know, the, the best technique from the Toyota production method we should all remember, the five whys. Ask why five times, and then you'll finally get the answer. So if you're trying to figure out why something went wrong on a plant floor, it works. If you're trying to figure out why something went wrong when a developer promoted something and it didn't work right, first time they're gonna say, "I don't know."

[00:25:29] Second time they're gonna give you a little information. By the time you get to the fifth time, they'll finally tell you exactly what happened. Someone hit the wrong button

[00:25:36] Coté: A- after, 

[00:25:37] after a while when you do that, do you find that people just tell you five whys right away? And they're like, "Let's just get it over with. I'm, I'm gonna, I'm gonna answer your five questions."

[00:25:45] Ben Savage: No, because they're not usually thinking of what actually happened the first time. They're still

[00:25:50] Coté: Hmm.

[00:25:50] Ben Savage: working through it. So some of that's, you know, it's like with troubleshooting, right? If you go through a methodology, you tend to get to the right answer, 'cause a lot of times what you see the first time isn't necessarily what happened

[00:26:01] Whitney Lee: So, so I can really wrap my head around the, the pizza portal side of your job, but the Veritas Automatas, did I say that right? Is, uh, is a little more opaque to me. Is that ba- is that consulting at the end of the day?

[00:26:16] Ben Savage: A- at the end of the day, it's consulting. We do have some product we've been working on that's some IoT-based distributed ledger technology,

[00:26:24] Whitney Lee: Mm-hmm.

[00:26:25] Ben Savage: but it's not in the market. So most of what we do, you know, 90-plus percent is some kind of consulting or services, whether that's custom software development or custom, like machine learning development.

[00:26:36] I'm separating those 'cause sometimes we're just building a model for somebody.

[00:26:39] Whitney Lee: Mm-hmm.

[00:26:39] Ben Savage: Um, sometimes it's building an LLM, uh, you know, generative AI pipeline. Um, sometimes it's really helping a customer build a process. We've got a customer, probably can't say their name right now 'cause the, uh, the press release hasn't gone out, but I'll make sure to share it with you guys when it does. But we've helped them build a whole pipeline that will help them transcribe textbooks into a very specific, uh, format so they can go produce more textbooks, 'cause it's really hard for them to find transcribers in that particular use case.

[00:27:09] Whitney Lee: Hmm.

[00:27:10] Ben Savage: So we're taking technology and saying, "Hey, we're not replacing people.

[00:27:14] What we're doing is creating something that will let them go faster 'cause they can't find enough people to go meet the demand that's out there." that to me is like one of those cool kind of process technology combinations that we get to do. Sometimes it's less exciting and we're building, working with customers to verify robots that check filters

[00:27:32] Whitney Lee: What sets you apart from other consulting companies? How do your customers find you? Yeah

[00:27:38] Ben Savage: so we have, I think, a unique combination of scientists who can do the machine learning. Like we can go all the way to build a custom model that runs at the edge or do something like that. So we're not just taking something repackaged from open source or something like that. All the way to we can do that with LLMs, to we can go build software, whether we put conventional teams to do old school .NET to new school techniques where we're using assistive, uh, AI to go build all the software. then we have the ability to go package all that together. And most people will know just the cloud or just a piece of it or just the machine learning, and we can kind of talk across all of it, and we have experts who are comfortable working in all those spaces

[00:28:22] Coté: So you, you know, I, I think, I think relative to, uh, what a CEO would do, like how do you, uh, you know, you're talking about you have cycles in, uh, uh, in shipping and, and a, and a, in a, in a down cycle you're kinda like, "Well, now what do I do with all this capacity? I can't just have it sitting here." Like how do you, how do you manage that similar thing in, in, in a consulting organization where you're like, uh, "I can't not have 10 people because then when I get a contract to deliver in 90 days, I can't hire 10 people," so like, you know, benches and beaches and all that stuff.

[00:28:51] Like how do you... How, how much time do you spend thinking about that?

[00:28:55] Ben Savage: So I spend a lot of time thinking about it. So one is having a really good talent team that can go help find talent when we need it. sometimes we bring in contractors, sometimes we bring in partners, depending on what it is, if it's not something we have a skill, because we can't be everything to everybody, right?

[00:29:09] We're very focused in the specific kinds of skills that we're good at, and then we'll bring an expert in. Like if I need someone for a life science project that's an FDA, uh, uh, what is it? FDA type Part 11, um, Part 21, Type 11 or whatever it is. It's one of the very specific things for life science that you have to do to make sure that a process is safe and is FDA approved. We'll bring in an expert to go do that to work alongside with our team. a team, what I've focused them in on, on some very specific technologies overall, some very specific industries. So at our core, we're really good at .NET because we have a lot of customers that do that. We're really good at TypeScript and Python. And then from a machine learning perspective, there we're a little broader because some of that is a little more transferable between the different things. So we'll do everything with conventional machine learning to the large language models and the generative AI pipelines for that. then industry-wise, we focus very much in life science. And then because of my industrial background, we pick up quite a bit of stuff in that space as well because people either know us or that we go, we go forward. But it's trying to keep that focus on the team because that, the team bringing in more projects in those is what makes it more repeatable, which makes it so I can redeploy those people on the bench faster so that we can go deal with that

[00:30:25] Coté: And, and then, and then what, what i- what is like the, the, uh, and I don't mean this in, in, uh, to use the wo- the word a third time in the, in the episode in a shitty way, but like what, uh, what's, what's the theory of like the, the branding on y'all's site? Like the feeling of it. It's very like flat and like, um, ki- kinda, kinda like a toned down Jack Kirby style, like very industrial and not...

[00:30:47] I don't know. W- when I see stuff like that, I think there must have been lots of meetings and lots of conversations because it's very of, of its own thing

[00:30:57] Ben Savage: So, so we worked with a friend of mine and he has the same love of punk rock I do. And if you look at our VA logo, it ends up looking a lot like a Black Flag logo. You can see there's some influence between

[00:31:08] Coté: Mm.

[00:31:10] Ben Savage: the, the bars of the Black Flag. And I think that's kind of how we got to it, 'cause we were talking about some of that, you know, want to have a little bit of an attitude but not too much that it's

[00:31:20] Coté: Well, you're like, "We're a business."

[00:31:22] Ben Savage: yeah

[00:31:24] Coté: Yeah, yeah. It, it ha- it, it has a, it has a interesting also kind of like industrial sort of like, I don't know, something to it. It's, it's, uh, quite unique

[00:31:33] Ben Savage: it was, it was to create an identity that when people see it, they know who we are and that we can go use that as part of our story as we go tell it

[00:31:42] Whitney Lee: And which-- So these two companies, you said you're the founder of one of them, right?

[00:31:46] Ben Savage: of Veritas Automata

[00:31:47] Whitney Lee: Okay. What's the story of, of that? How did you... Yeah. What's the story? Founding is fascinating to me because I think there's some amount of, um, audacity or just be like, "This should exist and I'm going to work my ass off to make, make it exist."

[00:32:01] So what convinced you it should exist?

[00:32:03] Ben Savage: So in 2020, I-- as part of Apex, we decided to go sell our industrial business to Fastenal. So we did that deal between March 1st and March 30th of 2020. So just imagine putting a deal of that scale together with a publicly traded company in March of 2020. Um,

[00:32:23] Whitney Lee: Did that happen because of COVID or was that coincidental? Coincidental.

[00:32:27] Ben Savage: Yeah, it, it was, it was already in process and strategically we just completed it then, um, because it made sense.

[00:32:34] Whitney Lee: Okay

[00:32:35] Ben Savage: So as we did that, we were looking at what did the new business or businesses need to be? and I had the opportunity to take my technology and my innovation teams, 'cause at that point I was the chief innovation officer

[00:32:46] Whitney Lee: Wait, I have a question. did the new business or businesses need to be? Makes an assumption that there needs to be a new business or businesses. um, why-- I mean, if you made something and sold something, you have an opportunity, I guess, for one thing, to live on a beach, or two, to plug in, to go be a barista.

[00:33:03] Or like, why, why does there need to be something at all?

[00:33:06] Ben Savage: So I think my dad is an entrepreneur at heart. He's the one who started Apex.

[00:33:10] Whitney Lee: Okay

[00:33:12] Ben Savage: just grown up, I don't know any other way to be. So I don't

[00:33:14] Whitney Lee: Okay

[00:33:14] Ben Savage: you'll ever see me just sitting at a beach. There's always gonna be something I'm experimenting with. as Apex had completed the deal with, with, uh, Fastenal, we then looked at and said, "Okay, what are we gonna do?"

[00:33:27] So Apex ended up becoming Apex Order Pickup Systems, and they focused on food, and I had an opportunity to take my technology team. So I had a technology, um, like software development, all that team, and then I had an innovation team that I'd built at Apex at that point. And I had the opportunity to spin those out and create what's now Veritas Automata to say, "Hey, some of the underlying communication we use in the devices we think has some broader use cases for it." So that became what became Veritas Automata, and along the way, we've picked up all this other kinds of project and other work and this typical startup story. You start thinking you're gonna do A and you end up doing Z

[00:34:04] Whitney Lee: And such a new business wasn't threatening to the people who just bought your old business when you're using some of this, that stuff?

[00:34:11] Ben Savage: they didn't have license to the newest new stuff that spun out,

[00:34:14] Whitney Lee: Okay

[00:34:15] Ben Savage: to go compete with them for a period of time. I don't remember how much longer

[00:34:19] Whitney Lee: Okay. Oh, watch out world

[00:34:23] Ben Savage: That's not a space I wanna go back into. I did that for 15 years. I did it be- when I was in high school at different points. I, I wanna go play with new technology. That's-- My team loves what I love, which is solving really cool, interesting problems, which is why we get to work with exciting customers

[00:34:39] Whitney Lee: How, how often is someone coming w- to you saying, "Hey, here's a really cool, interesting problem and we need someone to solve it and we think that's you"? Or how often is it like, "Hey, we have this real cool, interesting idea. We tried to make it, we failed. We want you to fix what we broke"?

[00:34:56] Ben Savage: It's a little bit of both. Depends on the customer, and we love them both, and they're both, for us, interesting opportunities

[00:35:05] Whitney Lee: How often, like do you, they want you to make something of something? Like if they already made a mess, how obligated are you to clean up the mess instead of just making a new thing from scratch?

[00:35:17] Ben Savage: It depends, and it depends on how well they listen to our advice, right?

[00:35:20] Whitney Lee: Yeah

[00:35:20] Ben Savage: I mean, that's-- and particularly in this world of, and that's a great question in this world of vibe coding, right? People can vibe code anything once. Um, the question is, will it work more than once? Or will it work when they go, have to go update something or there's new CVEs they've gotta go patch? Probably not even thinking of that.

[00:35:38] Whitney Lee: Mm-hmm.

[00:35:39] Ben Savage: we can go do an analysis with a company and say, "Okay, here's what our recommendation is. Here's how to either take what you've got and make it better, or how we can take the core idea you have and make it real for you."

[00:35:50] Whitney Lee: Mm-hmm.

[00:35:50] Ben Savage: both are valid answers. It depends on what our customer's business is and where they are in that state, and how they're raising their business and their, their business model and cost and all that stuff

[00:36:03] Whitney Lee: How much is, like, solving cool, interesting problems, and how much is, um, people that those problems are the ones to solve? Or how much are, like, massaging egos when something that they want, really wanna do does break?

[00:36:22] Ben Savage: So I would say it's 50% taking on boring projects that where people need lots of extra hands to go figure something out and, you know, typical software, enterprise software, right? Where, you know, we've got lots of people on a .NET project or a Java project or something like that. Uh, you know, some kind of technology modernization. And then it's probably 30-ish percent saying, "Hey, this is a cool idea. We both agree. Let's go work on it." And then it's 20% working the hard cases, and those work and sometimes they don't, 'cause, you know, you can lead a horse to water, but you can't force it to drink

[00:36:58] Whitney Lee: You need, um, a Domino's Pizza tracker for your software projects. And there are extra steps involved if you don't take our advice. There's, uh, yeah

[00:37:10] Ben Savage: why you structure things in phases and work through it with the customer

[00:37:14] Coté: Yeah, so- sometimes you have to heat things up even though it makes no sense 'cause, 'cause they like it. Uh,

[00:37:19] Ben Savage: Yeah.

[00:37:20] Coté: you, you know, I, I w-

[00:37:21] Ben Savage: right,

[00:37:22] Coté: I w- I, I was, I was, I was looking over some stuff that you've spoken on and worked on, and I... And it occurred to me, kinda like you were saying, you, uh, it-- Well, relative to AI stuff, you, you, you s- you, uh, you and your, your people spend time on, quote-unquote, "boring stuff," which is to say not AI.

[00:37:38] And I've been thinking recently, like, "Man, I could really go for some news on, like, configuration management about now." Like, like I, I need some variety. Like, it's like all I'm, uh, you know, whatever your favorite pie is, it's like all I've been eating for the past two years is pumpkin pie for every single meal.

[00:37:55] And, like, and sometimes it has whipping cream on it, and sometimes it doesn't, and that's it. But, like, maybe, maybe, like, some miso soup I, I would enjoy. Like, like, what, what, uh... From that perspective, like, what else has been going on in the world? I assume things are evolving. Like, people have new ways of, like, monitoring Kubernetes and deploying it.

[00:38:16] Like, like where, where's, like, uh, where are these-- how are these people sustaining themselves? Oh, you're ruining it. You're bringing back the pumpkin pie.

[00:38:25] Ben Savage: Well, but I think there definitely, there are all those that are going there. I think AI is probably becoming more entwined with some of it than maybe it should in some cases. This is something my team and I talk about because we work in pretty security conscious, safety conscious industries at points, depending on what you're doing with industrial, right?

[00:38:41] If there are robots or things. I can't have a robot-ish when I give it a command. It either does something or it doesn't, otherwise it's not safe. Life science is the same way. Finance, when we work with those customers, it's the same way. So one of our discussions more and more has been when do we use AI to go do development?

[00:38:58] When do we use good review process to catch stuff? And when do we say something is so critical people have to hand code it? I would put some of your kubernetes and some of your DevOps and some of that, you've got to think about it because you can't have your system kind of running or kind of deploying. It's got to actually deploy, otherwise it doesn't work

[00:39:17] Whitney Lee: I've heard this perspective recently that will stop talking about pumpkin pie, Coté, um, uh, pretty soon here, but not because pumpkin pie is going away, not because AI's going away. It's just, like, expected. Like, uh, uh, apparently I wasn't around for this, but people used to say internet-enabled w- all the time about

[00:39:37] Coté: Hmm.

[00:39:37] tech, and then at some point you just assume it, it's gonna have something to do with the internet.

[00:39:41] Whitney Lee: We're just gonna talk about the tech for what it is. So I think we are gonna stop it. Do you agree with that? I see you nodding your head, Ben.

[00:39:48] Ben Savage: One of the first products when we first launched it, we described it as an internet appliance, 'cause Internet of Things didn't exist yet, so you

[00:39:55] Coté: There you go. Yeah, yeah, internet appliance, and yeah, like web TV. Yep, e- going back, yeah, yeah. Yeah, you know, it's making me th- speaking of food, it's like I, I remember at least, uh, back in Austin when tiramisu was a good thing, and you, y- I mean, a new thing, and you would go to a restaurant, and every waiter was very excited to describe tiramisu.

[00:40:15] You got the ladyfingers, the espresso. Like I don't even remember what's in it, but like they would give you the full thing. And nowadays it's just like, yeah, tiramisu. I don't need to explain it to you. It's normal. It's, uh, so maybe that's the AI instead of the pumpkin pie. But yeah. So how, how do, how do you, uh, uh, like, like in that area, you know, when, when, when someone does need, other than it being like, you know, the right thing to do and, and, and challenging and interesting, like, you know, so you've, you've got some team, and you're like, "I know you really wanna work on an AI thing, but we gotta go configure servers for the next three months."

[00:40:46] Like, the, is there, is there sort of like keeping people on track for the, the, sorry to keep using the metaphor, for the pumpkin pie or like, you know, and kind of like steering them away from, from the other thing? Like how do you... Uh, having been a developer myself, I, I know I needed a lot of steering to not like build the fifth f- framework, you know?

[00:41:07] Ben Savage: Well, a-a-and that's a lot of what my delivery team and some of my management team do when they're working with their teams, is making sure that the work is being done as expected. You know, that gets to, okay, are we doing the right code reviews? You know, all that kind of basic blocking and tackling, whether it's humans doing or AI is doing it, you need some way to go follow those kind of good practices. And so that really becomes something they're working with their teams on, both from a coaching and an execution perspective. And yeah, sometimes you've got cases where you see a developer, they, they stray into the pumpkin pie and you gotta go, "Oh, you can't do that. You gotta come back to where you're supposed to be." most of my team responds pretty well to that coaching, and they stay with it and, you know

[00:41:49] Whitney Lee: the-- your customers, I'm sure, have different adoption strategies and patterns, um, as they're adopting AI, and those are kind of evolving industry-wide. Uh, how much-- like, for example, like there used to be people called token maxing, like everyone's keeping track of like, "We want you to use maximum AI all the time," and, and that's kind of fallen away to, to more efficient things. But how much of that, culture shift affects you? Like watching other people's culture shift around AI affects you?

[00:42:18] Ben Savage: So what I've driven our team to focus on is how do we focus on harnesses that are multi-model capable, is the term I'll use. So we're not-- when the next new model comes out, we can switch to it without having to redo all our tooling,

[00:42:33] Whitney Lee: That makes sense

[00:42:35] Ben Savage: plane for it. That's for our internal stuff. But sometimes our customers, because they're in very secure development n- in ecosystems, I guess is the term I'll use, they provide us the tools. In that case, we can be anywhere from, "Hey, we've got cursor and lots of tokens, and go tell us how many you're using," and they're gonna provide them, to, "Hey, you can't use AI at all. You gotta go old school." And so in those cases, we've gotta go work with the, with the customer and work with that process. That's, that's why I say it's so much on the processes that we build internally for our teams and then my management team and the delivery team to make sure that we're following whatever we've agreed to for the customer, 'cause it's different on different projects.

[00:43:13] Whitney Lee: How often is a customer doing something really cool and forward-thinking that you're like, "Oh, I hadn't thought of that, and we're gonna bring that over into our business"?

[00:43:21] Ben Savage: Um, I would say we've seen some interesting stuff On how do you do some of the measurement, sometimes we take little bits of it in. I, I don't think I've seen anyone who's gotten anything that completely blows me away, but that's 'cause we're dealing with a lot of-- Sometimes we're dealing with, you know, 15 multi-billion dollar corporations. Those groups don't move fast in that kind of stuff. When we get to work with the startups and stuff, that's when we see the, you know, the smaller, more agile groups. We see some interesting stuff, and we always try, you know, we try not to suffer from not invented here and incorporate whatever we see that's a good idea, share those ideas when they're worth it and our customers are going to go, uh, listen to us and, um, try and take that approach.

[00:44:02] Whitney Lee: What's, what's an idea you've been excited about lately?

[00:44:08] Ben Savage: So I think this idea of model routing and building, whether it's gateways or, you know, whether you want to call it a gateway or model router or a semantic router, looking at how you do that, because I think from a business perspective, it's very interesting to manage costs, and I can route things to the model that's the right level to go do that. But maybe even more importantly, it gives me some ways to go protect corporate data so I can make sure I'm doing PII scrubbing or other things so that things don't get to models I can't trust. Whether that is, do I trust Anthropic or OpenAI? Hmm, don't know. If you look at some of the news with Kimi and that maybe they were routing stuff to Claude secretly and then training back on the traces, maybe I got to be careful there. And so that gives me different tools, and then back to the ability of maybe I want to host some models internally, whether they're open models or things, and gives me a lot more flexibility. So that idea is really exciting

[00:45:01] Whitney Lee: you-- I think I mostly know what a model router is, but will you say it to, like explain it to me briefly?

[00:45:08] Ben Savage: So if you think about, think about like a typical internet router where I can control, or load balancer maybe is a better example, where I can control what traffic goes to what server or where based on different rules, whether that's an endpoint or other logic. So if I think about a large language model, if I use the term semantic router,

[00:45:27] Whitney Lee: Mm-hmm.

[00:45:27] Ben Savage: I'm being intentional saying I'm looking at what is the prompt coming in and

[00:45:32] Whitney Lee: Mm-hmm.

[00:45:33] Ben Savage: modifying the data that's in it to protect the business so certain things don't go out,

[00:45:37] Whitney Lee: Mm-hmm.

[00:45:38] Ben Savage: based on that kind of request, I'm choosing different models. Um, and you could be doing that with something like a classifier, which is conventional machine learning,

[00:45:46] Whitney Lee: Mm-hmm.

[00:45:47] Ben Savage: you could be doing that with the new, I'm gonna call it fad, the Jev fad and all these

[00:45:52] Whitney Lee: Yeah.

[00:45:52] Ben Savage: one type models. You could do it

[00:45:54] Whitney Lee: Mm-hmm.

[00:45:54] Ben Savage: those as well, and those don't require as much pre-training

[00:45:57] Whitney Lee: Okay, so the logic of where it goes is in the model router itself and may or may not be, um, a model that's doing the routing, the semantic understanding. Interesting. And then I f- to me, like the PII scrubbing, guess I hear that under the words guard, but it makes sense to me that if it's going one, from one place through a thing to another place, you could do routing and guarding and caching

[00:46:19] Ben Savage: we

[00:46:20] all in the same spot.

[00:46:20] to that 'cause we do so much in life science and there's just so much data there you don't want

[00:46:25] Whitney Lee: Hmm. Yeah Um, what other, what other culture changes besides the one I mentioned about, uh, token maxing are you seeing out in the wild as people, uh, as companies' AI practices are maturing?

[00:46:41] Ben Savage: I think the biggest thing, and I, I don't even know if I've seen a good answer to it yet, which is we're generating so much code now, how do we review it to make sure it's reasonable code? then how do we build in all the security models to scan behind it using the conventional static analysis and the other things to make sure there's nothing there?

[00:47:00] 'Cause even in little experiments I've done with AI, it does some very questionable things sometimes. It's gotten

[00:47:08] Whitney Lee: Mm-hmm.

[00:47:08] Ben Savage: but sometimes it still does things you're like, did you do that? Why did

[00:47:12] Whitney Lee: Mm-hmm.

[00:47:12] Ben Savage: do five different mo- frameworks to go call the SQL database instead of just using one?" Some of that you can kind of control through prompting, but some of that you've gotta go manage with your code review and your other processes.

[00:47:23] And that, that's the piece that probably scares me the most with all this, 'cause humans can't keep up with it. But

[00:47:30] Whitney Lee: Yeah

[00:47:32] Ben Savage: yet to really fix some of it either

[00:47:35] Whitney Lee: I see that at Datadog there's like AI-assisted tasks, at a SaaS AI-assisted coding. We're building new Git infrastructure to support all the extra code. Just, um, I don't know this, whether Datadog's doing this, um, this is a gap in my knowledge, nothing about Datadog, but, um, this idea of catching super inefficient code that you just said is an interesting problem one I've not heard anyone talking about before.

[00:48:01] Yeah.

[00:48:03] Ben Savage: And, and sometimes you get unexpected results. So I built a really stupid experiment, and this is when I get bored with AI, I make it do silly things.

[00:48:13] Whitney Lee: Uh-huh.

[00:48:13] Ben Savage: And so I have one of the Spark DGXs, and I was going, "Okay, what can I run just on this box without external things?"

[00:48:20] Whitney Lee: Uh-huh.

[00:48:20] Ben Savage: And one of the-- my colleagues in his, his interview question is: What is your theme song? So with a

[00:48:29] Whitney Lee: Mm-hmm.

[00:48:29] Ben Savage: bit of ADD, you go, okay, well, there are models where I can generate a theme song. I can go recognize an image and generate something from that, can be lyrics that can go feed into a theme song. So I built all that, and it was working in about five seconds from image to song generation, occasionally it wouldn't go do the lyrics.

[00:48:53] So then I said, okay, because I was using AI-based coding to one of the tools to go build it. okay, well, it's not always having the lyrics. So then it puts in a thing. Now it takes like three times longer. It always has lyrics. So it's like, okay, do I really wanna go through the code and figure out exactly why it's taking so much longer?

[00:49:09] Whitney Lee: Yeah

[00:49:11] Ben Savage: I just go, okay, this was a stupid experiment and it works when I hit the red button

[00:49:16] Whitney Lee: And, and honestly, uh, the answer to that's always gonna be different depending on who it is and what they're doing and, like, does it need to support production traffic? Does it represent our business? Is it something I'm, yeah, doing on the side and who really needs a perfect theme song? Or oh no, I have to wait 15 seconds for my custom theme

[00:49:34] Coté: Do,

[00:49:34] Whitney Lee: on my photograph

[00:49:35] Coté: do, do you find that maybe something like that repeats its- repeats itself in business where the... I mean, these, this could go either way, but the, the, the customer is more willing to use a prototype to figure out what it is they want? In, in, in the, you know, the example you had, it's sort of like, ah, at first it seemed like three minutes would be too long, but then I was thinking like, "Well, maybe it's fine."

[00:49:59] And you know, and, I mean, often you, uh, you build some software, and you put it in front of the person who, who said what they wanted, and they have... Maybe they're happy, but they might also have a reaction that's like, "That's not what I wanted." Or it might be like, "I had no idea that I wanted this, and now I want it."

[00:50:15] So I would imagine if you can produce more code, I don't know, do, do people change what they want around more, or is it as it, as it always has been?

[00:50:24] Ben Savage: Oh, I think you definitely see people being much more willing to say, "Well, let's do this experiment, let's do this change, let's do some of that." I think that creates some of the challenges 'cause what makes good communication work in software or good interfaces work a lot of times is that you've taken time to go build those steps, and sometimes you're building spinners or other things to go hide other latency that people may not realize are there as part of the experience. when people are willing to just keep adding, you lose some of that intent, I think does sometimes make it harder to make really good things because you're not spending some of that time, that intent to go say, "Well, I know when they push the button, it's gonna take seconds to get there, so then I'm gonna do a little You know, think about video game load screens, right? Think about how much effort goes in to go load all that data, and they're just, they're making it look like something's happening, keeping you occupied while they do it

[00:51:18] Coté: Mm. Uh, yeah, there's, there's, there's sort of like magic and, and innovation and fun in the waste. The, the, the t- which, which I imagine that doesn't probably apply to logistics, right? Like, you're probably not gonna discover something great because there's, uh, something broke down in the, the, the chain.

[00:51:32] 'cause your package took 30 days instead of 10 days to get back from LA.

[00:51:36] Ben Savage: that drives innovation for other things

[00:51:38] Coté: Exactly.

[00:51:39] Whitney Lee: But I was thinking though, like s- we're talking about like the psychological customer experience, keeping the locker warm or whatever. Maybe sometimes you want it to take a little longer 'cause then it has like, oh wow, they, it worked harder on this, so therefore it's more meaningful to me that like, oh, after a minute they came up with the perfect theme song for me personally.

[00:51:58] And if it comes out in five seconds, it's like, whoa, there's hardly any effort there. I don't care. You know what I mean?

[00:52:04] Ben Savage: I think intent is very important when you're doing things like that, right? When you're building something that's gonna interact with a human, the intent of how you're communicating or what you're doing in that experience, whether this is software or fun software, that is something you need to think about 'cause that's gonna drive how it's adopted, how people use it.

[00:52:22] It's gonna drive, okay, have I made someone so annoyed they're gonna throw something through the screen, or they're gonna feel like they had a good experience? And that I don't think changes with AI. AI gives you different tools to go do it, but also then you've gotta go call out the limits and say, "Okay, do I, how do I get AI to do what I want?"

[00:52:39] Like if you ever tried to create a video with AI, you just type in a prompt, it's a roulette if you're gonna get what you're expecting. If you

[00:52:48] Whitney Lee: Mm-hmm.

[00:52:49] Ben Savage: time to really write a detailed prompt or use some of the new storyboarding things where you give it a lot of guidance, it'll give you something pretty cool

[00:52:56] Whitney Lee: And then when you're working with all these customers, people have different appetite for AI and, and really I guess, um, trust. Like there, there are lots of ways to build AI, like trust with AI, but how do you, how do you get a consistent experience? Or how do you, um, more than that, how do you get the customer to trust that what you've built is gonna give their customers a con- a consistent experience?

[00:53:22] Ben Savage: a lot of that starts in the early discussion with the customer to figure out what is the problem or opportunity or challenge they're trying to work on.

[00:53:28] Whitney Lee: Uh-huh

[00:53:29] Ben Savage: And there it's identifying the right tools, 'cause maybe we use AI to go build the code, and then we've got a whole review process with them to do acceptance on different steps. Or maybe we're building in not generative AI, but a conventional machine learning model to go what they want. They may have said, "Hey, give me AI 'cause I think that's what I want." And then you go, "Well, okay, it's gonna cost this much per call and you're gonna spend this much on your AI budget. But if you do a conventional machine learning model, you don't have to pay any of that." So it's all through that customer discovery and working process, and it's, it's different for every industry. And, you know, I've got some projects where people are saying, "Hey, give me more AI," and I've got

[00:54:08] Whitney Lee: Mm-hmm

[00:54:08] Ben Savage: where they're saying, "We're really concerned about what this is gonna mean to our workforce, so we wanna move very cautiously with it." and both of those are valid approaches

[00:54:18] Whitney Lee: I, I wanna switch topics. I understand you host a podcast too. Is that true?

[00:54:25] Ben Savage: I do with, uh, my colleague Dr. Sarkar. Um, talk a lot about AI news and occasionally they let me show experiments that I've created that traumatize Suren, I believe

[00:54:35] Whitney Lee: What's a piece of AI news that you presented about lately on your podcast?

[00:54:39] Ben Savage: I think last week we were actually talking a lot about the system one models and how those are different from a classifier. Um, I don't think they replace classifiers, but very interesting on how they work. Um, the key is gonna be the speed and the cost and the performance of them, 'cause as you stack, you know, every time you stack another thing in the decision tree, it just takes that much more longer

[00:55:02] Whitney Lee: Well, I always kind of thought of classifier models as not something else in the decision tree, but like we're gonna put this, um, you know, yes or no comparison or, uh, compare one thing to another thing here instead of a full model to, to make things lighter. So the idea that you're saying it might make things heavier 'cause you're adding one more thing, I guess it's just another thing in the tech stack?

[00:55:22] Ben Savage: the system one are true LLMs. They're LLM size, right? All the-- I

[00:55:26] Whitney Lee: Oh, okay

[00:55:26] Ben Savage: AI just came out with one in a 600 million parameter, but most of them are either if they're open source, they're based around like a 27B is where a lot of them 'cause they're just, they're variations of Qwen.

[00:55:37] Whitney Lee: Okay

[00:55:38] Ben Savage: you've got the Jev model, and that's like 250 millisecond response, I think is what it is.

[00:55:43] Whitney Lee: Okay

[00:55:43] Ben Savage: sound bad until you think about, okay, how many other calls are going on on top of that one 250 millisecond response

[00:55:52] Whitney Lee: So, so you're not sold on, uh, on it? Or actually, for those who, who have-- aren't listening, will you bring people up to speed with, with what Jev and models like Jev are?

[00:56:01] Ben Savage: Yeah. So I think, I think they're another tool in the arsenal to solve problems.

[00:56:05] Whitney Lee: Uh-huh.

[00:56:06] Ben Savage: to me, everything is a tool,

[00:56:07] Whitney Lee: Yeah

[00:56:07] Ben Savage: you don't always have to use it if it's in your toolbox, right? I mean, I can hit things with a wrench, and sometimes that's more effective than the hammer because that's what was handy. Um, it's not always the right tool either. So if you look at the Jev models, they are a variation of what they're calling a system one or a decision model where you can go feed it in a question and a list of answers, and it's gonna give you the probability of what is the most right answer. then you can choose something. Um, so the other thing I think they're calling some of these is there's a bunch of names that people are using.

[00:56:37] I think decision model is one of the other names that people are using for it.

[00:56:40] Whitney Lee: And what's the advantage of that?

[00:56:42] Ben Savage: um, it's not generating all the tokens it would take to gen- do a large response. It's not going through some of the thinking type stuff that a conventional model would be. So if you're thinking from a more conventional software decision perspective, will be faster than saying, "Okay, I'm gonna get it in.

[00:56:58] Now I'm gonna get a JSON thing out. Now I gotta go process all this JSON." It's gonna say, "Hey, here was the input. Here were the potential options that you gave me. It's option one or option five," or, "I'm 95% certain here's what it is." So it can speed up where people are using AI that maybe they didn't really need generative models.

[00:57:17] Whitney Lee: Hmm. That's, uh, that's kind of can be often. Uh, so, uh, my-- the last question I'm gonna ask, I can't speak for Coté, is, uh, you say you like to tinker a lot. What's a fun project you've been tinkering on lately?

[00:57:32] Ben Savage: Um, have you ever read Douglas Adams?

[00:57:36] Whitney Lee: Yes

[00:57:37] Ben Savage: Have you read his "Dirk Gently's Holistic Detective Agency" books?

[00:57:40] Whitney Lee: I haven't read those. I act- they're on my list, but there's a, a BBC show. I watched the BBC show, and I loved it

[00:57:47] Ben Savage: Okay. I haven't watched the show, but in the first of those books, there's a key piece of software that was created called Reason.

[00:57:54] Whitney Lee: Okay

[00:57:55] Ben Savage: a tool that would take your corporate filings or your spreadsheet and turn them into a corporate theme song

[00:58:01] Whitney Lee: Okay.

[00:58:02] Ben Savage: With AI,

[00:58:04] Whitney Lee: Uh-huh

[00:58:04] Ben Savage: an SEC filing, I can now feed that in and generate a corporate theme song based on that.

[00:58:11] Whitney Lee: Oh no

[00:58:13] Ben Savage: they're not good ideas, but yes

[00:58:16] Coté: So what, what, uh, what, uh, what, what are, what are these theme songs looking like? What are they, uh, what's like the feel of them?

[00:58:22] Ben Savage: What you get to choose what genre you want. So

[00:58:25] Whitney Lee: Okay

[00:58:25] Ben Savage: do in, in the tool, I'll s- I'll send you guys the little explainer video I created for it. That's one of the other things I've been like, how do I get AI to tell some of this, convey this information that makes sense to people? Um, but you can choose whether you want pop rock or inspirational or hip hop or country or whatever will fit your brand. So you can go feed

[00:58:46] Whitney Lee: It,

[00:58:46] Ben Savage: and it will generate a song

[00:58:48] Whitney Lee: and it gives lyrics. Like what kind of lyrics would it give?

[00:58:51] Ben Savage: Um, I have that on a different device. I don't have any here. Uh,

[00:58:56] Whitney Lee: Okay

[00:58:56] Ben Savage: it, it's-- They tend to be overly inspirational and overly... If you, if you went to a cult, I think that's kind of the lyrics they might create, but that might be the right thing for a corporate theme song

[00:59:07] Whitney Lee: And, and really the idea when you make something like this is just you're learning the te- technologies underneath and entertaining yourself.

[00:59:14] Ben Savage: or

[00:59:14] Whitney Lee: Yeah.

[00:59:15] Ben Savage: "Hey, I had this idea. Can I glue these things together? What happens?"

[00:59:19] Coté: That, that reminds me, there, there's, uh, somehow, uh, I came across this, I don't know, guy who makes music, as so many guys who make music do, named, uh, named, I think it's Louie Zong, and he has this one album called Business! Exclamation point. And it's, it's, uh, it's like it, it's kind of... I don't think there's any lyrics, but it's, it's, um, I don't know, uncanny.

[00:59:42] You listen to it and you're like, "That is business." Like, the, the, those are, that's totally the sound of business happening. And you know, there'll be like, uh, I think, I think the, the, the, the, uh, the song, he of course has, has song names, and the one I always remember is, like, The Big Meeting, and you know, it's just like all of the, uh, I'll, I'll have to, uh, I'll, I'll, I'll have to put it...

[01:00:06] There, there's, uh, he's got the full album in YouTube, so you can go listen to Business, and, uh, you know, The Big Conference, The Big Meeting.

[01:00:15] Whitney Lee: now too.

[01:00:16] Coté: Yeah, yeah. I think

[01:00:17] Whitney Lee: Holistic Detective Agency.

[01:00:20] Ben Savage: Yeah, like I love playing with that. I've been lately also doing some experiments on like how small can I run stuff? Like what can I run on an NVIDIA Jetson, the little eight gig, uh, thing they have.

[01:00:29] Whitney Lee: Ah.

[01:00:29] Ben Savage: Um, we're just doing more hardware 'cause so much of my background has been building physical products at points, so I'm like, "What can I make it do?" And I've made it program a silly badge, that's about it so far. I haven't had enough time to experiment with that stuff yet. But that's--

[01:00:47] That's where you're going next?

[01:00:48] to do it

[01:00:49] Coté: I, I...

[01:00:49] Whitney Lee: some dollhouse m-models?

[01:00:51] Coté: I, I think, I think there maybe there is an emerging hobby in the, in the AI world. It's called something like, uh, you know, productivity entertainment. And, and it's, it's sort of like there, there's this con- there's this concept in, in the role-playing game world that's sort of like, uh, prepping time is either not playing, so you shouldn't spend so much time prepping, or prepping is actually part of the fun of playing, so you should spend a lot of time doing it.

[01:01:16] And like, you know, it's these activities of like, "Wow, I'd really like to organize my photos," or, or like, you know, I, I would really like to come up with business songs. Like, and, and it's-- I've accomplished something that is entertaining myself. And of course, there's the learning aspect in everything, but it's just sort of like it's either this or like watch some TV, so this is probably fine e-even though objectively it doesn't do anything.

[01:01:46] Whitney Lee: I'm all for gamifying everything. Yeah

[01:01:49] Ben Savage: recently build a tool to go through a massive set of drives I've got with all my MP3s and music files I've ripped from different CDs and bought and all the music, and go figure out where are the duplicates and then figure out how to organize it for me, 'cause every time I start, my ADD kicks in after about three folders and it's like,

[01:02:08] Coté: Oh, that's the dream, to have the, the, the one type of Tupperware, uh, in, in, in there and have it all organized well. Well, I, you know, now, now that I've remembered the business album, I need to get to listening to it again so I can, uh, I, I, I can remember things like Calcutron 8000 and visualize it. Uh, great tracks.

[01:02:29] But tha- tha- thanks for being on. That was fun. Now, we've mentioned many places, but g- g- give us, uh... Where, where could people go to check out, like, what, what you're up to and, uh, and your stuff on the World Wide Web, the internet appliance?

[01:02:42] Ben Savage: so, um, LinkedIn, our website, Veritas Automata, that'll get to wherever we've got our podcast and all the socials for that, and those are probably the two best places. I'm a bad social media person on my own. so at some point I'll actually remember to post something on LinkedIn. I think it's been like a month since I last did,

[01:03:00] Coté: Mm. Yeah. I, I f- I feel like I've been running an experiment over the past two months to see if social media matters for me anymore, and so far the answer is no, it's cool. You don't have to log into LinkedIn and see whatever is going on there. I mean, I, I, you know, a- as far as, like, a business front, not y- not business front like an illegal front, but the, uh, uh, presenting yourself as a business, that would be a whole other story, but just, like, as an individual, I don't know.

[01:03:28] It's not much going on there

[01:03:30] Whitney Lee: that was one of my AI projects was to get all that pretty automated.

[01:03:34] Coté: Yes. I, I was

[01:03:36] Whitney Lee: if, if I was like in a car crash and died tomorrow, my LinkedIn account would still post every day for the next month. It would

[01:03:43] Coté: As, as I was saying that I w- I was thinking like, you know, W- Whitney is kind of active. I mean, she's active, but the she part, sh- there's just active.

[01:03:54] Ben Savage: It's just she's embraced the whole digital twin concept. She's

[01:03:57] Coté: Exactly. That's, that's right. Well, uh, there's no segue from digital twins that I can think about, but this has been another episode of Software Defined Interviews. Uh, and it is... Normally I, I would tell you, uh, exactly the URL to go to, and I can actually tell you. It would be softwaredefinedinterviews.com/130.

[01:04:20] I'll put a link to the business CD, to the, the URLs we mentioned, other things like that, and, uh, thanks for listening. And with that, we'll see everyone next time. Bye bye.

[01:04:30] Whitney Lee: Bye

[01:04:31] Ben Savage: Bye