Ben Shaw: Fabio, thank you so much for coming onto the podcast. Fabio Faschi: Of course, happy to be here. Glad we can chat. Ben Shaw: I think the I I think the the question on everyone's minds is a little bit where where does AI bring what we know SAS to be in insurance? And that's also that's also the topic of the season. So I appreciate you coming on, sharing some of your experiences with me. maybe let's kick off with what's your background? how did you land in insurance SAS? Fabio Faschi: Yeah, sure. So I started in what I hesitate to call the traditional side of insurance, but certainly more traditional than you know what it is considered today in the insure tech world. But basically, despite having done a handful of internships at insurance agencies throughout high school and college, I didn't really make up a career of it until much later. but after university was actually working for a title insurance company, then was brought in under the wing of a very senior insurance veteran that was basically building the third arm for a small family-owned corporate structure that owned a mortgage company, a title insurance company. company and was building the P and C brokerage. And they said, hey, why don't you get licensed, come in under his wing? And then eventually you can take over the brokerage from him in you know three to five years. And those three to five years very quickly became three to five months as he retired. And then basically I found myself thrown into the deep end of the pool of, you know, having just been licensed in PNC and kind of learning how to build this all from scratch. was with that brokerage, you know, through through fruition, got it got it up and running, but then was headhunted by the startup world. Did my tour of duty through a number of startups, was with Policy Genius in New York for just under four years, built their entire PC division. and then you there, through the help of my mentors and advisors, I'd already stood up a small advisory that was doing some consulting on the technology side because that is really where I thought I was gonna be spending more of my time throughout my career. and then started really mingling the, you know, coming the two worlds between insurance and insurance technology. you know, it came up in the Advent of IntroTech 1.0 as they called it. So what we viewed as the high-end technology back then was the comparative raters, the CRMs, Ben Shaw: Cross. Fabio Faschi: and so all the implementations that came along with that, right? How to use Salesforce and HubSpot to build a brokerage, and how you tied in you know things like Easy Links and applied epic overhauls and all the technology that came with that. Very you know, it sounds almost funny to think about it nowadays because you know Ben Shaw: Mm-hmm. Fabio Faschi: how much it's changed in the last 10 years. but fast forward, yeah, I mean did my tour of duty through a number of different companies on the startup side. Eventually started spending more time on the technology side. And then that brings me brings me to where I am today. I spent the last five, six years consulting, doing a lot of work, specifically answering the question of, you know, what no one was even calling it AI back then really, but it was more so how do we think about, you know, technological automations and machine learning in insurance. and of course in the last three years that's very quickly evolved into you know, what is AI for insurance? How do we get it deployed? How do you know carriers think about it, MGAs, broke? Of course, right? There's there's no shortage of of how it touches financial services as a whole. And yeah, that's what brings me to where I am today. Ben Shaw: Very cool. very rare varied background. let's start with insurance brokerage. D did you come across technology while you were a broker? Was there certain things that you felt were already natural to the world of brokering when you started? Fabio Faschi: Yeah, so I've helped build three brokerages, right? And at different points in my career. and for sure, right, when w when I actually entered the industry the first time, I think my my the first thing that hit me as as a child of the technological era that you know grew up with computers and grew up seeing kind of the advancements of you know from from the first cell phone I had to then the smartphone, I assumed that the rest of the workforce had followed suit, right? And when you kinda start working Working in insurance for sure, you start seeing like, well, people are transacting via email to go to you know comparative raters. There were just some carriers didn't even have you know broker facing portals at the time and then so we saw pilot agencies for some of these first portals. Then very quickly that started to develop. But honestly by the time it hit fruition, it was like, okay, well I don't I'm not going to these individual portals. I want a comparative rater. And there was Ben Shaw: Yeah. Fabio Faschi: and there was always like there was always a a step further that these things needed to go, right? In terms of you know what were the limitations of the comparative radar? Whether they didn't support something or on the back end for your agency management system, like how were do downloads and so there's always there was always some other project in terms of like how do we tie and close this gap of getting things done. Ben Shaw: It felt it felt incomplete from the get go. Fabio Faschi: I exactly. It was always like you went into that conversation understanding, okay, where are my where are my limitations going to be? Right. And the only circumvention to that was gonna be like, okay, well, can like how do we build and close this gap ourselves? Is there even a way to do this? What will it cost? and or do we just swallow the pill and just kind of move on and say like that's just a known limitation no one's solved for yet? Ben Shaw: Mm-hmm. Fabio Faschi: but yeah, for sure, like and I think Mm-hmm. right as that was evolving and then seeing of course a burgeoning of newer players that, you know, whether it be an MJA or, you know, carrier offshoot that was basically saying, hey, we're gonna solve for this by providing a portal directly ourselves, right? The rise of even like direct to consumer carriers and how those played with independent agencies or not, or whether they were going directly to the clients. So you know, shopping for technology was always a very interesting conversation because in the back of my head I always had the mind I was like maybe I'm maybe I'm doing the wrong thing. Maybe I should be building the technology rather than working the actual pipeline of the insurance value chain itself. But yeah. Ben Shaw: That's interesting. that wait, when you sorry, just to to drill into that a little bit deeper. So so when you were on the broker side, you sometimes came across technology where you were like, you should be building the technology. Are you thinking for you as a person to move into the technology space or are you thinking the broker should own parts of technology? Fabio Faschi: So I I I definitely thought like both sides, and I think that was the first evaluation, Ben Shaw: It's like. Fabio Faschi: right? But I think candidly, I I think most insurance companies, whether they're brokerages or not, I I don't think there's a lot whole lot of value into building the technology and owning it yourself candidly. Ben Shaw: Interesting. Fabio Faschi: I I I think and I think now with AI it's funny because there's a lot of, you know, in all of SAS, not this is not an insurance specific phenomenon, right? There's a big impetus for sure, right? On hey, the cost to build has been so greatly decreased, everyone's gotta build now because now it's so much easier. and that's partially true. But I would say it's it's very easy to build something to maybe 70% of the way. And then that last 30% is really where you're gonna find out like, this might not have been worth building it. But more to the point is really the cost with technology has never really been in building. it's pretty, you know, it's pretty good to build things, and it's been pretty easy to build things in a static environment. Technology, the reason SaaS has always been a challenging industry is because its technology is the realm itself that is constantly evolving, much more than traditional insurance itself. And so you're always thinking about like what's my obsolete date for software? This is true 15 years ago as much as it is today, except today that timeline is invariably a lot shorter. And so you shouldn't be asking yourself, what's the cost to build? You should be asking yourself, what's the cost to maintain and evolve my technology? month after month from here on out. And we the the biggest thing about AI I would say is not even so much its speed or its accuracy. It's the incremental speed and accuracy from one iteration to the next at these as these models come out. That is where I spend an overwhelming amount of my time in terms of making sure that our evaluations are not only kept up to date for the customer and our end product accuracy, but also that we're able to keep up with the latest developments that are coming out of the major AI houses to make sure that we're Ben Shaw: Course. Fabio Faschi: able to offer the best models in the best configurations while also at at a at a better incremental price, frankly, right? In terms of like what's what's constantly happening now on a you know month to month, if not week to week basis, is that these models are changing. The underlying structure of this technology is changing at a much faster clip than any other software realm. has in the past. And that's really the that I mean that's really the biggest cost. It's not so much like building at a static point in time anymore and saying, I've shipped this Ben Shaw: It's Fabio Faschi: product and here you go. It's gonna be like how does this product keep up after a month, after two months, after a year. Ben Shaw: I I I almost feel like you're yeah, you're making me think back of when I started in tech fifteen years ago and the biggest pain point being exactly this sort of idea that you look at it as a project, right? That has a beginning and an end. Then at the end, once the project is complete, the the tech is developed and you forget the fact that you once you develop the tech a year later, it's all it's all r gonna be outdated. And I I used to work for this large UK bank. That did exactly this for their corporate banking. They had an in-house team that built this technology. And by the time they were done, it was a three-year project. You know, the technology was like it just didn't even feel right to use it anymore. And they ended up completely shelving it and taking something from a vendor instead. So I think these sort of buy versus build cycles have always been around. My question then is, how has AI changed how corporates think about in-housing versus externalizing? What do you think? Fabio Faschi: Well, I think it's brought more More of a spotlight to the build versus buy question, but I think at the disservice of some, it where suddenly everyone thinks like, okay, it's been solved for, now building can be done. but now it should change the appreciation to what build versus buy actually means because really it raises the question of like, okay, if you now are building, there is now to keep up that to keep that techno to your point, to keep that technology up to snuff with the alternative. Alternatives and the competition that is also going to keep developing in the meantime, there is a much bigger cost there. and at the same time, you know, the buy on the buy side, when you're buying from a vendor, you're essentially renting, you know, you're renting their engineers, you're renting their development, and you're still able to able to get a better bang for your buck than you were in the past because that technology is invariably being able to change a lot more. Like our underpinning product, for example, you'll tell you. I got further, right? Is constantly improving and the architecture that builds it. is constantly has to improve under it because it is far more flexible than re what you would say, you know, like you said, a software technology that you would buy off the shelf 10 years ago is in invariably built with a much more static architecture in mind because there's not this big development cycle on a month-to-month basis. You're thinking, hey, we want to put and develop the best technology that we can today, and we don't have to focus on the architecture being so incredibly flexible because maybe we'll sunset this. With a leg, know, turn it into a legacy product and do a 2.0 10 years down the road. That's not the way software can think anymore because you'd be invariably so far behind the eight ball on a year on an annual basis, right? so how you build that moat is really in how you think about the orchestration of your architecture and how you are going to evolve as the cycle of technology continues to develop year after year. Ben Shaw: I I I think that makes sense to me of how you think about further conserve their customers. I think from a if I'm an insurance company, don't I also need to think about a SAS product historically has all has always been somewhat of a of two-sided bargain, right? You you get a product cheaper because you're sharing costs with others, but it's still Fabio Faschi: Sure. Ben Shaw: i you're beholden to the you're beholden to the issues list, you're beholden to the prioritization. of new features by the SAS company and also most likely some of the key functionality you might want will never get built because it's just very custom to you as a large organization. So given that you know AI in theory provides a window of opportunity where some of that customization can now actually be done because it's so much cheaper using AI. how do you think about that if you know in in in the in the interaction between an insurance company and say further AI. Fabio Faschi: Yeah, and I think this is a real market dynamic. You could take it even outside of insurance, right? And it comes down to really like what is the profile of the client that's being served there. whether it's you know, that's the difference between software that's developed for a large institutional bank versus a private financial planner, right? That might want some automation done on their on their day to day. And it's the same thing in insurance, right? the you know, the rollouts we do at the enterprise scale would be invariably different than what we can. offer a small independent brokerage, a family-owned brokerage of of of five people. and so, you know, I would say the focus is depending on really what customers your your goal is to serve ultimately as a product in the market. And I think you're approaching it a little bit differently there. And I think the good news is like AI can can do a lot to service both sides, whether it's okay on the small scale, yeah, maybe you're you're offering slightly more out-of-the-box solutions, but you're able to do it at a better scale and at a better margin to service those smaller ones because you are able to provide those, you know, those those solutions that much faster. The question is, does this you know 10,000 small independent brokerages do they really need a best spoke solution for their services? The answer is probably no. And that's, you know, we've seen this also in other software in the cycle. That's, you know, you can get something off the shelf and it will service 90% of your cases. And those 10% of cases probably would develop you into more of a niche brokerage. and but the opposite is also true, right? There are some very niche brokerages out there that their goal is to serve a very specific person. Clientele. And so for them, they'd be better served working with a bespoke solution built for their specifics of hey, we only do, I don't know, the ENS cannabis side of insurance, and like there's you know, we work with five carriers, but for those five carriers, we need to have a very specific pipeline chain of how we get them their submissions, what they need to see, and and so you you know that's not going to be served well by the kind of off-the-shelf collection. and obviously for the enterprise, it's it's anything in all between, right? Because typically There, it's really just an accelerated or I should say augmented problem of that niche specialty where you're saying, well, hey, if you're gonna do you know, if we need a solution for our international property submission team and we need it for our you know global ENO team or our employee benefits team, like that's gonna be there, there's there's nuance that you need to build bespoke there. The advantage of AI, right, is that you can build faster to spec. And also, of course. Non-generative AI to bring in a layer of context that doesn't, you know, to that didn't exist in the past when you'd say, like, okay, we're gonna build bespoke, and we have to dynamically build the routing for every single one of these little edge cases that exist there. and because the enterprise world works on larger volumes. AI at the end of the day, right, is is one of the major pillars of AI is pattern recognition. The more data Ben Shaw: Course. Fabio Faschi: and the more data you have to work with, the better your result is going to be at the end of the day. So being able to train at the enterprise scale is invariably gonna lead to a lead to a better result, not just over time, but over even the you know, just the simple scale that they're working at on even the first deployment. Ben Shaw: My understanding you're right then that you're saying for the very niche players they may want something bespoke. But that doesn't necessarily answer the question of who does it in a bespoke way, right? Like there can be a service provider who knows how to use AI much more, may maybe better than that niche insurer. So that service provider could now service this customer because AI makes it easier to serve the long tail. Is that kind of how I would think about it? Fabio Faschi: Well, I think when we talk about the the service and delivery side, it's interesting. so does AI help with that? Yes, for sure. But I think, you know, these are complex deployments and so really it's gonna come down to still like how the delivery is done for any technology, right? It's gonna rely on people and I think like one of the things that I certainly hold true to like why I pivoted into this kind of realm from my background is because I I've been a practitioner in all these seats, right? Whether it be underwriting or claims or brokering or wholesaling or carrier side or MGA side, like I've seen all the day-to-day operations of sat in the seat, and that gives me invariably con like a context of like where to even start building with AI. Because at the end of the day, AI is reactive, right? And building with AI is going Be reactive to how you not just how you prompted, but where you're even focusing its its attention. and so how you build for a niche program in any in any industry, or how you build to where niche, I'll just call it a solution, because really has not again has nothing to do with insurance, right? You have to understand on first principles, like what are you trying to solve for, what is the outcome you're trying to solve for. And I'd say that's probably the biggest challenge in the AI era, is like the conversations we We have with our clients, a lot of time is is they might even underposition their ask of like what they really want to solve for. And they think like, you know what? We need to just get our team's emails in a better format. And I'm like, is that all you really want to do? Like, that's not like that's not the solution. The solution, right, is like What are you trying to do? You're trying to get your team data in a better format so they can pass it along in that format so that they can suss out X amount of information from that format. And so a lot of times that's how our conversation develops in terms of like we have to really think about like what is it that we are try what is the decision that we are trying to or the outcome that we're trying to get to at the end of the day, and kind of lead a horse to water to really go one layer deeper in how we solve for that. And that's where AI has unlocked the gate of well we have more capabilities but how you get to Ben Shaw: Mm. Fabio Faschi: the it unlocks the capabilities but how you get to the what we need to do what we need to make is is really the core Skill right that when it comes to the deployments. and the same is true for AI in in other realms as well, right? The example I always tell people is like you could tell AI to you know to to paint a a picture that you already like if you have a full picture in your mind that you want to create. And then you tell AI to create that picture, you might miss a few details at first, right? And so through that iteration, you'll say, no, no, no. These are the specifics of how I want this portrait to be made of exactly this. And I have to be wearing a hat, and the hat needs to be red, and it needs to be a wide brim. Like those are the kind of details that it's like, yeah, AI can very quickly iterate on those different versions, but until you give it that full context of all the specificity, like that's when it then fulfills your vision. But you still need to have the original vision of what it is that you're trying to accomplish at the end of the day. Ben Shaw: That's interesting. I when you say you have to have the vision of what you're trying to accomplish, where my head goes is one of the promises, apparently of AI, being that the person who has the vision of what they want to accomplish, now being able to be the same person accomplishing that vision, right? The the the baseline theory being you needed in the past, someone who does customer success, customer management, a product Fabio Faschi: Yeah. Ben Shaw: person who understands the art of the possible. A head of RD or a CTO who translates that into actual technical issues, the tech team that really wants to focus on these technical problems. And now that whole chain of people that obviously always introduces some inaccuracies, right? Could theoretically be replaced with, well, maybe not exactly one person, but maybe three or four people who have those capabilities combined in one person. How do you see that evolving? And have you seen any signs of this type of model already being adopted? Fabio Faschi: for sure. Yeah. on the on the SAS side, the level at which not just the speed and and capacity of what I'm able to do in a day-to-day, just because I'm able to outsource tasks that typically would then have to be delegated to someone just to whether or not I can, know how to do it or not. And again, I I've been someone that's been doing like the software cycle from you know from soup to nuts throughout my career as well. So like I know how to do the full chain, but some of you know other people that haven't had that same experience. exposure to be able to delegate that you know, even the first pass of something to accelerate to give you a a framework to work out of whether it's like, okay, like how do we need to think about the scope of work or or even like a contract review or like how do we think about the first pass of certain tasks? that previous to that, the only solution was go learn you you have to go learn how to do it. You have to sit down with someone that knows how to do it. so In sales, right, any team will rely invariably at different stages of the of the organization, right? On different formats of tribal knowledge. And I think that's like a huge accelerator that like you can canonize a lot of that tribal knowledge now and get a team up to speed in terms of like, yeah, this is where we want to be. And of course AI can you can even leverage AI to even give it, you know, give you its two cents on like can we optimize this further, right? in terms of like okay, we're we're doing this today, do you see do you see an outward room for improvement of like hey we're we're like yeah this is this would be an easy like low-hanging fruit of how you could adjust this to like why why you're adapting this on in every single one, why you're why you're changing the language like this versus like this is just a better language that you could use to always boilerplate like that. and you always have to take that with a grain of salt, but it gives you something to react to, right? And and I this is an analogy People always say is like you ac of course you can leverage AI like another person in the room, and the same way that you would take advice from whether it's you know an intern or your CEO, you might say, like, okay, well, does this actually make sense? Do I have more context than it at this point? Right? And people always make fun of AI because like it's very agreeable, and it will say, like, it'll be a yes man and be like, yeah, yeah. But but part of that is because people don't realize like they they assume that AI always has perfect context to the how they're envisioning something. And a lot of time then you'll give it that context and it'll say, yes, you're right. now of course there's also plenty of examples where AI is just you know blatantly wrong. But but that's a again, that's a matter of like there's always a a rationale that you're able to trace back as to why AI is able to say certain things a certain way. and a lot of it comes down to again understanding how AI really works, which again returns to like pattern recognition. And so that's you know, that's something that is just gonna continually improve and returns to my first point, which is like the shortfalls of where we're at today, and you mentioned like the person carrying the vision can now always execute on something. I would say it's like you know, they can't always execute on at the end of the day, right? Like I'm a very amateur developer, and so yes, I can get you know AI to develop certain small widgets and apps and things that I use and processes. But when it comes down to okay, is this optimized for development? Right. Nine times out of ten, right? An engineer will look at it and say, well, there was a much easier or structurally better way to execute this. But yes, the outcome works, but as we continue to scale this, like there might be a better way to solve for this problem, or like that's not the way that it's not best practice in terms of how something would you know would be done. And that's what matters in that last 30% I referred to earlier, right? Ben Shaw: So I love that. I think you were making two different points. I think your first point was we have found a way to make each one of us way more productive than we were in the past. I think you're good, you have good examples of using AI to brainstorm, and obviously you're still the human in the loop adds a lot of advantage. There, I'm thinking of the economist article that was published yesterday where someone was saying, hey, actually, if articles are written, By AI, why would you not like that? Because do you really think that every author of an article, every writer, is like the next Ernest Hemingway, right? That's it's a wrong way to think about it. So I I think it's challenging some of our assumptions of what what what we like in writing and I mean to your point about sales. My initial question, and I like that you brought it back to that, right? Was more about does the way in which we do work collectively change? As a result of coding specifically becoming much easier. And I I didn't when I said when I brought up the example that someone who has the vision can now execute, yeah, you're you're absolutely right, right? It it'll take them a while to be as good as someone who's done software engineering for the last 10 years. Fabio Faschi: Yeah. Ben Shaw: But let's let's think ahead. let's think five to ten years ahead. Do do we anticipate that how we hire people and how the balance of people between the different parts of an organization is changes and what one example I think of is having come from Germany, right? The automotive industry they they tend to have a lot of mechanical engineers, a lot less software engineers. One thing that made Tesla be able to, have so so so so much crazier multiples is because they were mostly software engineers. And Fabio Faschi: Mm-hmm. Ben Shaw: now the question is it will will it shift to maybe another line of work away from software engineering if that's no longer, the if that's no longer the bottleneck for growth. Fabio Faschi: Yeah, I I I think the reality is I think it will certainly shift it. I don't think it will be as black and white that suddenly a CEO or anyone can suddenly do because like, well, we don't need engineers because I have the idea of how to do it. Right. And it's like there's still like these are nuanced things that we're they're building that have a a certain level of depth that's like it's not good enough to It it's right out of the box. You can't only deliver a project vision just by outcome. or you can, but then what gets you from A to Z could be very different than your than what your end goal might be. and the example I'll give is something that I remember my you know Fourth grade teachers did as a thought experiment with us, and he told everyone, everyone's homework assignment was to write, you know, the instructions for how to draw a smiley face. And of course, everyone comes in the other day, right? And they come in and they say, All right, start by drawing a circle. And so he draws a circle, right? and like of of varying sizes, right? And then they say, like, put a dot in the top left. Top left where? Top left center, so then he puts a dot in the top left of the circle on the outside. Right. And then that and that's exactly right, just a small example of then like, and then draw a smiley face. And he draws it with like big human lips, right? And it's it's not like it's the instructions aren't clear, right? And that's a perfect example of, of course, just prompting in general, but also of a larger vision. That's like at the end of the day, right? Like how you're going to explain the vision for say you wanted to build a software platform today. You will invariably build a better product if you're able to specify the underlying architecture that is not only gonna get you that first version built, but then as you continue to develop it, okay, now I'm able to tack on more pieces without having to refactor the entire p product itself, right? And that is where you still need to understand the underlying what what have you built, right? the another example would be Ben Shaw: Ma maybe maybe, right? Ma maybe. I think the counter to that would be you just talked a lot about pattern recognition. At some Fabio Faschi: Yeah. Ben Shaw: point when you ask someone to draw a top left circle, Fabio Faschi: Yeah. Ben Shaw: right, that the the pattern would recognise that top low circle should be closer to what an eye might look like. And so the question then becomes, is there some model that will just at some point be much better at pattern recognizing than than than, the human would be? Fabio Faschi: And and I think, right, it's gonna come down to two things. It's gonna come down to yes, I think that gap will continue to shrink. will it ever be zero? I'm I'm less confident about that because I think even if it were zero, I think there's too much variability again in terms of saying the level of specificity that you would need to give to prompt an AI, for example, to build a software from top to bottom. Like if you tell it to just build a software that's gonna, you know, whatever, I don't know, do some sort of function, it will probably build you by basis of pattern recognition, the most like blasé vanilla kind of middle-of-the-road example of that. Which again, you might might not be able to then do anything with it. And so and you we see this, we see this all the time by the way, when it comes to AI, especially in these early days, like we see all of these paper cutter carbon copies of each other, right? And it's whether it's Ben Shaw: Yep, a hundred percent. Fabio Faschi: like you know, what I what they like to refer to in the online community, right, as like the AI courses. Suddenly overnight, because AI can give a very vanilla explanation of so many topics, people are like, ha, I'm gonna spin up the ultimate source of truth for how to learn about XYZ and it's like yeah okay but like you're really just copy pasting a lot of AI information Ben Shaw: Yeah. Fabio Faschi: right and it'll get you a very vanilla print and play that you can use for that and there's nothing there's nothing to it in terms of that additional layer and it's easy come easy go someone could very easily replicate it right and in the market dynamics of returning to this realm of SAS okay well if it can easily be built then it can easily be replicated. There's no defensible moat there in terms of saying, like, okay, what's gonna let someone else you know come and do it better than you? But there is a there is an incremental point where you move from something that was easily built to something that actually can fully address the outcome that's needed there. And again, this is the difference between me, you know, buying a Ferrari. And saying, okay, well now I own a Ferrari and I can ask Claude how it was built, and now if I want to tune it up myself, or if I want to rebuild a Ferrari, right? I mean I I I I know I know how to do it. Like yes, it will cost me a lot to put the parts together, but I'll be able to use Claude to help me build it along the way. And it's like, yeah, theoretically, yes. But wouldn't it be like isn't there an advantage to an engineer, an automotive engineer that's able to build like has studied how to build a Ferrari, has built other cars, understands the intricacies of why cars are built that way, to be able to build it that much faster than me, who's like, yes, now I'm armed with the knowledge, and maybe I'm able to get Claude to help me build it along the way, but like Claude might not know Ben Shaw: Yeah. Fabio Faschi: how I want to tune it, right? They not know what outcome I'm aiming for. Ben Shaw: Presumably, definitely. And I think I I think w how the way that you're now describing it to me, or you using this analogy, it it it just the consideration kind of fell into only one axis, right? Which is time. It used to be the consideration used to be a lot a lot a lot of different dimensions. It used to be talent, used to be IP. used to be it used to be time, right? It used to be a few other things. And really what AI in the software world, I'm not actually talking about Ferrari, because I I I'm not so sure that the mechanical engineering space is quite there yet. But definitely Fabio Faschi: Sure. Ben Shaw: in the software space, it's taken some of these other considerations away. And really what you're saying now is yes, I could maybe rebuild what's a good example, like Salesforce. But to rebuild Salesforce would take me X amount of tokens and X amount of hours. And let's fast forward that time. right. Fabio Faschi: Yeah, I would say it's it's probably short in the time. I think there's a few things to to call out there. that's like yes, you can achieve a much more reasonable a much more productive result in a much shorter period of time. the cost I think is an interesting factor because that's constantly f it's still fluctuating now and again the like the cost to outcome to the reliability of the like of to r reliability of that outcome is always interesting because that's really what needs to be evaluated at the end of the day. And certainly in the software space, it can be hard to tell at times. because just because so just because AI says something is functioning, you have to ask yourself, like, okay, how are you gonna measure whether or not it's actually working or not? but yeah, I mean th the the time theory, right? The the the time is the biggest increment for sure. and it and it does matter, right? Because that's ultimately like gonna be what you're evaluating it on. assuming assuming the first two parts are like correct. Assuming the first part of like that it actually built to solve what you actually wanted it to to to do. Right. Yeah. Ben Shaw: Very interesting. I think we're coming towards the end. Is there were there any other topics or any thoughts that you had in the concert and you know, and how the I AI landscape has you know, has evolved that you wanted us to talk about? Fabio Faschi: I think yeah, no, I mean I I think the biggest thing is is always I think in you know in the workforce generally it's an interesting dynamic, right? Because especially I'll I'll speak to you know insurance for an industry that hasn't been always technology forward. it's funny the amount of conversations we have, right, that can be reticent around AI in terms of one, if it's gonna take people's jobs, and two, if it's actually better than people when it comes to decision making in general. And had some very interesting conversations with, you know, underwriters in the past that ultimately they they Some of them have a very skeptical view of AI and have a very interesting view of AI because they're not even looking at it as an objective technology. And if what I mean by that is they'll say, like, no matter what the output is of AI, you'll still need to have a person looking at it to determine if it made the right decisions. Which I find a bit funny because particularly for you know for a practice like underwriting, that in theory. right, should be largely objective in terms of how we assess a risk and say, okay, insurance is an actual aerial science. If these factors are true, then this should be the outcome, right? And the output here. But some of them just don't trust it. And they say, at the end of the day, I'm able to figure out the better way to do it. And this is the same, and I see this through other practices as well. It's not just underwriters, right? It's the same thing with brokers. They say, like, we don't want to use comparative raters because I know the markets it needs to go to, because I've been doing this 50 years. I always find it funny because it's like, okay, well, if you know, if that is true, and let's believe it is true, whether it's the underwriter or the broker, you have a standard operating procedure, whether it's written down in your you know, written down somewhere or in your head, you have a procedure that You go through to determine where you are sending things. And that is what you could not only just teach to other people in the first place, and that is, you know, as a manager at multiple on multiple sales teams and multiple underwriting teams, like that's what I always encourage people to do. You find your best performers and you say, teach the other people around you, because like it's great that you're doing a great job, but like everyone else should also be doing a great job, and clearly it's learnable and teachable. Whether or not people are able to learn it as well as you is a different matter, but that's what AI. can do, right? AI says, okay, good. You found a great process. If you are able to explain that process to me, if you're able to explain the perfect process to explain the directions of how to draw a smiley face to me, then I can replicate it, right? And it returns to that core concept of pattern recognition. And I think there's a lot of skepticism but also misunderstanding that people just don't think at the end of the day that they just don't believe it. Because it's because it's different than how other technology has behaved in the past. But at the end of the day, right, it's able to, that is what the difference between generative AI and gener, you know, other AI that's existed for years. Like we've had artificial intelligence for years, right? Like you you you do you you pick an option between A, B, and C, and then it gives you a different branch depending on your option, right? And it's you've created AI that's able to react and react to different choices that are made. But now you do that obviously at a much grander scale, but now generative AI is able to react. to open-ended information and again return to that based on patterns. And I think there's just been a lot of hesitant hesitancy in respecting that, in understanding its true capabilities at scale. In under Ben Shaw: The difference between deterministic and probabilistic AI yeah. Fabio Faschi: Yeah. Yeah. Yeah. Yeah. And and not even just that. I mean, I there's people that also, you know, the difference between Claude, you know, Claude or Chat GPT is just chat versus, you know, cowork or work, right? And the other understanding like like no no no, this is something that's going to proactively be able to execute tasks for you based on triggers, based on time, based on what you're like it's able to actually proactively go do work for you as opposed to just being reactive, right? So I think there's, you there's way and it's pr and and then again, I would be shocked if this was exclusive to the insurance industry. talked to a lot of other, you know, CEOs across the other AI ventures in different spaces, and it's the same thing, right? People are always curious about two things that I think they they underestimate what it means to build versus buy in terms of like what building actually should mean. and then, you know, tied to that is really what AI deployment means, right? And I talked to a lot of organizations that even outside of insurance. right that will that will say no we've given our team co-pilot we've given our team claude they have everything Ben Shaw: Yeah. Fabio Faschi: they need and I'm like okay but like what what is like what does that do for them at the end of the day right it it in today's age that's Ben Shaw: Yeah, exactly. Fabio Faschi: the equivalent of telling your team they can use Google right like that so so Ben Shaw: Yeah, exactly. So far. Fabio Faschi: yeah I think you know we have to go I I think that'll change in the next five years I'm hopeful I'm optimistic that like I think people need to embrace it a little bit more I'm not an AI evangelist in the sense that like that's the only thing that matters in the world old anymore. But it's certainly something that can create massive benefits across the board. Yeah, huge efficiencies. Ben Shaw: Huge efficiencies. Yep. Fabio, thank you so much for spending spending time with me talking about the future of SaaS and AI. Fabio Faschi: Thank you, Ben. No, it's been a pleasure. Ben Shaw: Yeah.