Mark Vange on Autom8ly: AI-Driven Workflow Automation for the Future of Business

I don't know what the implications will be. I don't know what cool new things we will create, but there's going to be some subset of humanity that's going to continue to be creative and that's going to create these new things that we can't even imagine today that are going to appeal to and attract the attention of everybody else. You know, from a from the perspective of someone in the Middle Ages, the whole concept of a Renaissance fair, the ability to go somewhere for the weekend and pretend that you're, you know, living in Roman times or something, right, would seem crazy, but for us, large groups of us, it's not, right? >> Hi, my name is Demetri Bonichi and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI agents podcast brought to you by Jot Form and featuring our very own

CEO and founder, Idkin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show. Hello and welcome back to another episode of the AI Agents podcast. In this episode, we have Mark from Automately. How you doing, Mark? >> Doing great. How are you doing today? >> I'm living the dream. You know, you always got to be living the dream. Um, but no, I seriously I appreciate you for uh spending the time with us today. We're going to be chatting about automately and I think a great starting point would just be how did you get to where you're at um with automately or well first of all what does automately do and how did you get to where you're at with it right now? >> So automately uh we build AI automation solutions. We uh

like to focus on building the coolest AIS you know talking AIs and uh AIs that do you know anonymization. We have AIs that do SEO for large media companies. We have different solutions that we've built for folks and uh we're kind of uh of geeky so we like to focus on building the the cool AIs and finding partners in different verticals that uh help us identify you know what it is that the trucker needs from AI and then we help build it. They help sell it and we kind of go to market that way in different verticals. >> Makes sense. Okay. So um how long have you guys been in business? Um, Automately got founded in early 24, so about a year and a half for Automately. Uh, but I've built and sold seven companies over the years. I've just kind of have always built

technology companies. That's kind of what I do. So, you know, uh, I've been doing AI for 20 years because I spent a lot of time in video games where, you know, it was more common. And then uh kind of in 23 when we kind of turned the curve on GPT and just kind of you know the price of compute and the price of the chips just got to where it suddenly became um reasonable to get this stuff done in in a business context. I got really interested in what does AI look like post adoption. you know, um I'm uh old enough here to uh to have experienced a few technology waves and it's uh you know, it's always the case that technology moves really fast, but adoption moves really slow. So, what I always try to figure out is what's it going to look like

in four or five years when we actually go through all the adoption and all the pain and suffering and we're actually using it each and every day. And that's kind of what automatically is is uh our experiment to try and figure out what that looks like when uh the world catches up. >> What do you think that will look like when the world catches up? So you know right now the narrative is really around GPT and agents which for a lot of people kind of suggest autonomy and I actually think that of course that will exist but I actually think a lot of AI and a lot of the way it's going to happen will be what I call cooperative AI. So these are AI agents that work side by side with people rather than um you know replacing people um as it were. Um

I I think there's a few reasons for that. I think at a very simple level people don't want to buy from bots. People want to buy from people. Um, you know, I I joke with folks that if you're my plumber, I don't care what brand of wrench you use. I want to know that you'll be there at heater explodes. Right? If you're my lawyer, I'm not just buying the work product you do. I'm buying your insurance. I'm buying your license. I'm buying the resources of your firm. I'm buying your liability. Right? I'm buy I'm buying all of that stuff as well as the contract or the piece of paper that you could produce. And so that's not going to get replaced, right? So I believe that actually this kind of cooperative AI where where it's an acceleration tool for people rather than a replacement for

people is going to be where most of the growth and where most of the eventual adoption will be. The other cool part about cooperative AI is it lets us get there much further, much faster. And we can go later around kind of like our methodology for how we implement AIS if you want. But the fundamental reality is, you know, it's a very hard to get to 100% accuracy, but you can get a lot of leverage from 50% accuracy or 70% accuracy or 90% accuracy if you have a workflow that accounts for it. And if the cooperation between the AI and the person is structured in such a way that the person injects that extra 10% or 30% of their knowhow, their experience on top of the busy work that the AI can get accomplished. >> Yeah. No, that that that makes sense to me. Do

do you know what the reason is that we're kind of behind um or not behind why why we're closer to 50% maybe or 75% than we are to 100% accuracy? I look I think that there's a few different reasons for it. Um I think that the first reason is measuring success. So step three of our ninestep plan is always kind of figuring out how do you even measure success for AI inference, right? Like if you decide that you want an AI to do something sounds simple enough, but when you actually ask the question of okay, how do you measure the level of success that you achieved in doing this? often it's very difficult. So if you can't measure something, it's very hard to achieve good success in it because you don't know what success looks like. So that's kind of a big barrier. People think

they know, they intuitively kind of, you know, they know it when they see it, but when you try to actually put some qualifiers around it or like quantify that measurement, it's really tough. And so what happens is you get part of the way there, then you get some feedback from the consumer, the the human experience, right? Then we can incorporate that feedback back in and get to 51% next time and then 55%. Right? Getting to 100, if you don't know what 100 really looks like in technical terms, is almost impossible. >> Um, I think that makes Yeah, that makes sense to me. And what has been the hardest part of you uh trying to explain that to the average person dealing with um you in in your work? Because I can imagine that's kind of a bit of a struggle to to explain cuz the

AI industry right now almost feels like this like amalgam uh like esoteric like solution bot, you know, and nobody really practically understand or well not nobody. people who aren't in our bubble don't practically understand the implications of uh what requests mean and how hard they actually are to um be perfectly accurate. >> You know, I I think a lot of the problem is actually us uh the technology people. Um you know, we're we're used to 15 20 years of selling SAS solutions with the word being, you know, kind of solution, right? Um so we're used to like okay, I'm going to sell you a tool and so this tool must work, right? So it's like here's the solution, but actually that's not what we do with AI. AI is statistical. AI is not deterministic. AI is never finished. And so we set ourselves up for

failure if we go into a client and say, "Hey, I'm going to give you a tool to do X." Because implied in that is perfection. Right? when we come in and we talk to potential customers, we have this language that we use around, you know, think of an AI as a colleague rather than as a tool, right? If you hired a new person, the first thing you do is you give them, you know, all your manuals, which if you're lucky, captures 70% of what they need to actually know about their job, right? And then they shadow somebody for a while, and that person goes, "Oh, and by the way, the forks are kept here. Oh, you know, by the way, don't forget to take the, you know, the slop bucket out every night. That's not in the manual. And then, um, then you start giving

them little jobs with close supervision. Then you give them bigger jobs with a little less supervision. And over time, you kind of set them loose until they're ready to contribute. And in some jobs, that could be six hours. And in some jobs, like if you're a nuclear power operator, that's a fiveyear cycle of getting to the level of confidence where you actually leave the guy or the gal at the console by themselves. We talk about our AIS in exactly the same way, right? So when we deploy an AI solution, yes, we take the you know the the documentation, yes, we ingest some past communication for modeling, but then we also select the tasks we start with so that they are subject to close supervision. As I mentioned earlier, very early in the process, we identify what are the measurements of success for the task. And

then as the task is performed, we develop what we call a confidence metric. And that confidence metric is then used to drive the level of independence or autonomy that you can then give the AI. But then you base your decision about how much autonomy to give the AI on an actual metric that's derived from what you're achieving rather than cross your fingers, hope for the best, right? >> Yeah. No, I think that's that's a fair point. like my understanding of these tools is one thing and then the the gap I do feel like it's fair enough that we as the tech people do need to kind of provide an accurate representation and most of what I think we're being sold actually doesn't really probably come from people in your position it's more of the Sam uh Samman open AI the Microsofts of the world that

are trying to claim like it is uh something above what it is like for example You were probably familiar with the operator um cap. Yeah. Catch the operator like that product's fail rate is like 60% or something stupid, right? Yeah. But there was that wasn't mentioned, you know, in any of the uh news press about it or by OpenAI or the, you know, companies that would necessarily be meant to, you know, talk about it. So, I think that's a fair point. it's actually kind of on us to uh dial back the uh frame of reference as to like is this thing actually going to be able to do whatever we want immediately. So good point there >> and I think there's actually two layers to that, right? Like one is is the tool capable, right? And and so obviously GPT has gotten better, less hallucination,

this and that, but it's still not perfect, right? Um but the other thing we're experiencing I think that's underappreciated is and we saw this in the roll out of cell phones too, right? the capabilities that you can have for yourself in your own life with AI far exceeds any capabilities that you can have in an enterprise setting right now. So on top of what you're talking about, right, where so you know, I I can go and I can generate video and I can use GPT and I can do all this cool stuff for myself that if I work for a Fortune 500, I have trouble doing in a corporate context because there's all the other complexity. there's the liability and the disclosure and the you know um trust uh in in the data and things that that an enterprise has to worry about that's actually

layered on top of even the performance right so just like when we started having smartphones right we all carried smartphones but our company said no smartphones because we don't know how to deal with them and we don't know how to you know track the email and and keep track of the text messages and the liability and blah blah blah right so you know I I think there's two modes to that failure. One is yes, absolutely even for you using GPT yourself, it's not as predictable or it's not as completely accurate as you'd like it to. But then layer on top of that, you know, the the the investor relations guy who has to have an embargo for 30 days before the quarterly earnings, how he feels about like GPT being available to people inside the organization or like the head of compliance or the you

know head of security, right? So there's there's huge failures across the entire system. So when we go in and say we have a tool that will do XYZ like that is so likely to fail without a little more information to your point that you know we're kind of shooting ourselves in the foot when we go in with that kind of tool mentality I feel. >> Interesting. Yeah, that's a good point. I feel like a lot of uh a a lot of people don't appreciate the personal aspect as much in our space and in like general business space because we are so focused on what it can do for business. But at the enterprise level, there's a lot of this red tape. No, you're not you're not wrong at all. I I actually realized recently I've just gotten back into just fitness more and more and

like just on a personal level, there's so many people who are selling this or that app for example for like macro tracking. And I'm like, why would I do that if I can just like open up 4.1 or 40 and enable uh web search and just be like track my macros for today. I'm going to keep texting you throughout the I'm going to keep sending a message of what I ate throughout the day and it'll be like 90% accurate and like for all intents and purposes I'm I'm fine with that considering it's what my goals are. So, but for a company, right, that 10% at an enterprise level can be the the make or break point of of efficacy or not. >> AB: Absolutely. And and if you're a hospital, right, you need to be a little more accurate, right? If you're if you're a

nutritionist in a hospital that's you know minding somebody who's very sick uh yeah you need you know 49's confidence. >> Interesting. Yeah. No that's a that's a good point. So with automatically how are you helping to uh how are you helping uh people nowadays? what what I I guess dive into some examples of of how you are um bridging this uh gap to kind of like modernize businesses um with secure and uh you know transparent AI. So I'll I'll I'll give you an example. my favorite one. Um, so we uh focus on building really cool AIs and I I don't believe that we can be really good at cool AIs and really focused on a particular vertical because once you start getting into a particular market or vertical, you start having to dive into the nuances of that market and you kind of lose contact

with this very rapidly evolving AI in you know universe. So our go to market is we actually like to find partners you know people who are in different verticals who really understand those verticals um who have the contacts who are at the trade shows they may not know how to build an AI but they understand how the AI can add value into that space and we partner with them to offer services into that space. So one of the people that we found is in trucking which I love because it is conceptually the least techy of all industries right you kind of think of it as being very blue collar right and and and and with the help of this of of this person this partner we identified a couple of use cases which are incredible um but my favorite one is it turns out that

all these truck drivers it tends to be a lot of momand- pop operations you know they're sitting there in the truck all day long driving from point A to point B and they spend spend most of their time on the phone calling all of these different brokers trying to find the next load, right? So, these brokers are sitting taking hundreds of calls a day. Um, and less than 5% of those calls will actually resolve into a deal because that guy won't be there in time. He's got the wrong kind of vehicle. He doesn't have the correct licenses. He's, you know, there's a myriad of of nuances that make somebody the right fit or the wrong fit. And so we um the one use case we identify is basically having a a a phone answering AI that can ask you know the first 10 questions that

the user is going to ask anyway and filter out 90% of the calls so that those brokers can focus on the 10% that might actually result in a real deal and do a better job of it and also free up some some of their time to do more selling, right? Because that's the other thing they're trying to do at the same time. So there's an example of of a use case where we've kind of identified with the help of of a distribution partner something that brings real value, real incremental value to a particular industry, a particular use case. It doesn't have to be super accurate. It's not really uh a make or break. It's not life or death. It's just helping filter through some of the most obvious um time wasters. uh we did another voice bot or through through through a different partner we

identified another use case and this is for managed service providers. So these are people who provide like IT services for small to medium companies and we were talking to one in particular and he told us that 70% of their phone call volume was I forgot my password. Um, right. The agents hate answering those calls because they're boring. Um, the business hates answering those calls because there's no upsell opportunity for I forgot my password. Right? My email server is running too slow is an upsell opportunity. I forgot my password is not an upsell opportunity. It's just I need to reset your password. And the people who forgot the password don't want to wait on hold for the agent to free up. So literally a voice agent that can go through the checklist of questions that the human agent would go through and just oh if you

need to to change your password let me help you right everything else goes to the human agent you know another great use case where we can we build a voice agent that will do all tech support yeah we probably can again we're you know getting to 100% is really hard. This is 20% of the work 70% of the call volume. The caller is happier. the agent is happier and the company's getting more valuable time out of their agents, right? So again, like identifying these opportunities and helping companies be more successful by delivering like properly scoped problems. Don't need to boil the ocean. Don't need to be 100%. just need to find what the right efficacy, what the right confidence level is, understanding that it's not going to be perfect, but also it doesn't need to be perfect in this use case. >> I have an

interest I have a question I think that might be interesting then with that. Um I liked your uh phrase by the way, not boiling the ocean. I haven't heard that one before. That's pretty good. Uh I I think with this kind of adjustment that we're having in the workforce, it's always a focus it seems like from the general public that work is going to be decreased uh in regards to like uh actual jobs, right? Like this is taking jobs. Um because how you framed it was interesting. It was almost like those that are working are going to actually have more uh enjoyment out of their job. They're going to get more out of it. They're going to do calls that aren't ridiculous like I forgot my password. What would be your commentary on how I mean that whole situation? Because I do think it's an

one of those more hot button interesting topics where everybody's always jockeying for the uh the high ground of what is what is the right way to look at this? Is is AI going to steal our other jobs? Is it going to make way for better and more interesting jobs for people? How do you think it's going to impact the job market in the next 5 years? >> So I think that of course there will be displacement, right? Like this is this is a fundamental unpleasant fact that there will be displacement. Some people who are doing some work it is going to be displaced. Um, this place doesn't mean that you're not going to be able to find a job, but this place means you may need to update your skills or learn to do something a little different than what you've done. Um, but I

think also there are two or three fundamental truths that that remain. A, people want to buy from people. People don't want to buy from robots. Even worse, there's very little value in robots buying from robots, right? Like there there's really no point there. And but but that's kind of what we're what we're almost headed for in in some people's conception, right? Like if if if if my AI is booking all the travel from your AI travel agent, then you know like I'm still buying the travel from you, right? So it's not displacing that entirely. Um >> Oh, that's a good point. Yeah. The the simile I like to make is, you know, I can bring a forklift to the gym and I can lift any weight, but it's completely missing the point of me going to the gym, right? People want to buy from people.

People value content generated from other people. Um, I think this whole um, you know, the the the prevalence of Tik Tok, the prevalence of only fans, the the the whole kind of media landscape only supports this notion that ultimately what people want is connections with people. It's not just about buying the service. You know, you're sitting in Chicago. I'm sure there's within 20 miles of you, there's got to be 3,000 lawyers, right? Somehow all of them make a living, right? It's not that their services aren't fundamentally funible, right? They have to write a contract. But you pick a lawyer based on relationships, based on the way they conduct themselves, the their philosophy, their, you know, family connection, whatever it is, right? But it still comes down to you're buying a funible service but you're buying it from a particular individual who you like or you

don't like or you might even switch to a different one but you're getting the same product ultimately right so people want to buy from people want to deal with people uh AI creating work for AI I think is kind of you know missing the point so I prefer to think of it as as empowering you know I don't know if you recall a time before our answer to I wonder if was let's Google it, right? Like you had to go to like the encycl the encyclopedia bratannica or like order the book and wait for three weeks for it to arrive before you could answer the question. How empowering, how much faster, how much more productive have we become since we could just Google it. And for me AI, especially the generative AI stuff is basically like the next phase of let's just Google it, right?

The trap is the trick is that if you just rely on AI without critical thinking, without taking on the responsibility for that 20, 30, whatever it is percent imperfection in the AI, then you're going to end up with egg on your face and you know your brain's going to atrophy. But if you actually use it to power your creativity, if you actually use it to enable your ability to do things faster, then it's incredibly empowering, right? Like we could be digging dirt with our hands, but guess what? We use shovels and we use back hoes. Why? Because we can move more dirt faster. You know, if you start viewing AI as your mental shovel to move more mental dirt faster, you can just get more stuff done. You know, if we do get to great coding agents, how cool would it be if you could

literally go build me a website to do X and suddenly X exists, right? How many more X's could we build? And yes, some of those X's would not be successful. So what? Right? The point is you as an individual could do so much more and get so much more out of your life. You know, I am colorb blind and I I I joke I'm ambisoninestrus. can't draw, right? But now I can literally talk to an AI and generate the images that are in my mind, whether they are diagrams for business or whether they're just just pictures that that that are for, you know, my soul because I want to see this picture realized. Five years ago, I couldn't do it. I would have to hire an artist. So someone will say, well, you know, but now that artist is unemployed. But the truth is, I

would have never hired that artist. I would have just never generated that image. >> Interesting. Yeah. No, that's actually a good point. Um it's almost like especially for small businesses, they can't even afford to hire people out of these instances for these uh things that they want to produce. So rather than hiring somebody, which they wouldn't have, they would use a tool, which still in essence, especially if it's a smaller AI company, they're still paying a person to be fair, which I guess is actually to your point. I had never considered it from that framing and that's why I've been asking this question as a theme throughout I've had a couple podcast episodes today and I like asking this question because everyone's come at it from a different angle. But no, I I I think that's very that's very true. We're we're still required to

make a transaction with a person. I also think when it comes to the jobs uh side of things, some of of what people do in their dayto-day, they they've almost taken a a bit of an approach that they value it very highly. Not just because they had half the paycheck, but it's like, oh, well, this is what I do for a living. I don't necessarily think people who were on the assembly line for like Ford or otherwise prior to some of the more difficult or mundane tasks being quote mechanically automated um had a different response but I think now if we look at it we don't have a negative response to it occurring but since now it's all knowledge work like everything we do is quote knowledge work for for the most part the immediate response is negative But just because it's all on the

computer or it is digital does not necessarily mean that the skill or ceiling that you're trying to reach for yourself is actually anything else than equivalent to plugging in whatever it was in the widget on the assembly line. Like not to be rude, but like yeah, if you're just picking up a phone call and dealing with people asking for what's my password, like I don't think anyone's actually being fulfilled out of that. So, I thought that was an interesting point you made and it's quite interesting the dichotomy between we all accept well yeah it was stupid people are like putting their hand on the assembling line and whatnot. It's like well we we could also make that same assessment if we're being a little bit more objective about the situation here with some knowledge work jobs. >> I think there's an important distinction that you're

making or that you're implying between a task and a craft, right? A task is just a task. A craft is the thing that people pride themselves on having developed a special skill for. And actually some people turn tasks into crafts, right? Like a lot of coders view what they do as a craft, right? Like I can make more beautiful code than you can. Is an expression of it being a craft as opposed to I'm just, you know, building a house. Um, some brick layers view what they do as a craft because they can make it perfect and they, you know, that's the guy you want laying your brick. Let's face it, right? But some guys are just like, slop on the mud, throw on the brake, slop on the mud, right? So I think that where where people get worked up and understandably is when

they've spent lives or years honing a craft and becoming really good at it and and having their kind of self-image and their pride wrapped up in their ability to do that craft and then that craft is reduced to margarine through automation. Right? And you see that in musicians and you see it in artists and you see it in writers and we've been seeing that for a long time. But now you're seeing it in software development and next year you'll see it in legal associates and the year after that you'll see it in junior doctors. Right? the the the the question is how do you keep the human sort of need for for differentiation and for kind of pride and for fulfillment still engaged when you're automating a part of the task. And I think again where I really believe in cooperative AI being the answer

is you get to do the stuff that's inevitably not the stuff that people focus on and view as being like the magic ingredient they add, but you can accelerate all the work that needed to happen before there, right? Like nobody wants to spend, you know, five days mixing all the paints so they can paint the cyine chapel. But guess what? have to until that stuff got automated and then they could just squeeze the tube and the paint was ready to go. Same thing. This is an ongoing kind of process. >> I absolutely know I completely agree with that. It's funny. Yeah, I'm trying to imagine that. Or even the you know and I do think there's some beauty in the mundainess of some things but only for art really like only for art. And I think the you're you're right on the difference between tasks

and art. And um I remember there's an interesting uh there's an interesting paper written by Burch and Russell in like 1910 um or 1912 no it would have been sorry it would have been after the first world war so closer to 1920s um where he essentially made this uh argument. It was called impraise of idleness. His uh entire uh premise in this article was that uh we've arbitrarily been blowing each other up for the past like seven years. It's going to be great now that we have uh robotics and like automation in uh what we've advanced even during the war because you know half of the workforce wasn't really even working so we can chill more now is basically his whole point. He's like oh so we can relax a little bit more. Um I think it was like kind of the first introduction of

anybody even mentioning like a less than 5 day work week or 6 day work week at the time probably cuz they probably only took Sundays off I think um probably in the 1920s. Um, and it's funny cuz people, it's especially in American culture, consistent focus on, well, we got to work, we have to maximize, we got to work, we got to maximize, we got to work. This could be the actual adjustment towards uh more artistry and less tasks. I think if if this goes well, I'd hope so, right? Like if it doesn't, I'm trying to think we will never get out of this cycle, right? Like, you know what I mean? Um, you know, I I think that we've what we've always seen and this has been a continuing trend uh for millennia is that as we expand our scope of attention, you know, as

we can focus on more things, more things have to go on autopilot underneath that, right? You know when when when you were for 100 thousand years we were scrabbling just to generate enough calories to survive didn't have much time for anything else right when you actually started seeing some some tools suddenly we have the time to start painting cave walls right it's I I I I'm I'm eternally optimistic about the capacity of humanity to create new cool things but you can only do them if you can liberate the time and energy that you spend on the thing before. So, you know, when people um freak out about, you know, it's going to take all the jobs, like again, I just I just don't believe that's what's going to happen because, you know, in the 1950s, like we didn't have yoga studios and we didn't have,

you know, um hair, you know, blowing salons and we didn't have like all the stuff that came out of the the generation of more free time and more free wealth based on automation, based on the growth of the economy. And I don't know what the implications will be. I don't know what cool new things we will create, but there's going to be some subset of humanity that's going to continue to be creative and that's going to create these new things that we can't even imagine today that are going to appeal to and attract the attention of everybody else. You know, from a from the perspective of someone in the Middle Ages, the whole concept of a Renaissance fair, right, the ability to go somewhere for the weekend and pretend that you're, you know, living in Roman times or something, right, would seem crazy, but for

us, large groups of us, it's not, right? because we have the time, we have the free income, we can actually spend time doing this thing that's not fundamentally productive because we have spare calories, you know, like it comes down to spare calories uh and and how how you choose to spend them, right? >> Yeah. No, I think that that's a very good point. My um I mean for full context, um I mean I'm 27 uh and there's a common um so I have two both sides of my family are like first generation immigrants. So, I've heard stories and I do think there's a funny difference between people who have that experience of like close familial bonds of on the ground experiencing World War II and or the Great Depression. And when you don't have that, most people don't maybe see it the same way. I

I I I was lucky enough to be presented with as, but like the concept of people thinking that it's even anything close to as difficult as it was to live 60 years ago, 50 years ago, like I have stories I could tell for days of what I heard from the the people who were on the ground in that time. And and and I think it's it's AI actually is another ability for it to take another step forward. You know, that's why I bring for forth that uh in praise of idleness thing because I think as much as maybe Bertrand had a point uh or well some would say, "Oh, well, we're not where he we're at yet, where he wanted us to be at yet and it's been 100 years." It's like, well, also unlimited PTO um have you heard of Uber Eats? Uh you

know, these types of very basic things that we just take for granted now. Like, oh no, my HOA costs too much. I'm like, "Yeah, also I have a pool." Like, you know, it's like there are so many different like, for example, the uh AC in my building for the building was down, but the entire central AC for like individual units were fine. And like just these basic things, you know, that we've we've gotten to with technology progress. I just like to ask questions about where do you think it's going to go um for individuals in the future because it's it's what the one thing that is like esoteric is I don't think we any of us have like a proper answer to the question right you you watch the movie um Back to the Future and uh it was pretty far off but you know

pretty still you know but you know we might hit that in 2100 but um I I guess my next question would be what what what excites you most about um this uh relationship between like humans and I get and AI in the the collaborative agents that you're you were discussing earlier. What what kind of pushes you to keep wanting to work on this every day? >> So, um look, I I I'm actually a first generation immigrant. English was not my first language. Um and I I was born in the Soviet Union. >> Um yeah. Um >> yeah. Um, so you know, I I I I maybe have a little bit of a different uh trip uh than than most folks have had. Um, and you know, I I actually um fundamentally think that my job on this earth right now is to show another model

of working with AIS that's not a zero sum. and and I believe that that's I genuinely believe that that's where we're headed, but also it doesn't mean that we have to get there. You know, I've been fortunate enough to have multiple exits. I could be just sitting around sipping pinadas and and you know, playing hockey and having a good time, but I also feel like I have something to add to the conversation in terms of showing a model that I think is less destructive and less disruptive and and more enabling. And that's actually mostly what gets me excited, you know, in the morning. You know, we we've built this one solution, for example, that allows somebody in an emergency room setting to enter details of the patient presentation. In other words, like what what they're complaining of and and what are the symptoms that we've

been able to ascertain and it goes through the history of patients, you know, like billing and and and diagnostic history um and does all the um inference and basically can tell the person at that hospital, you know, here are the three most likely things going on. here are the tests you should do now to kind of evaluate if it's one of these things or something else. And what that, you know, does that help the, you know, the 30year experience doctor in a hospital room in Scottsdale? No, probably not. But it probably does help the, you know, the slightly qualified medic in some um, you know, reservation medical emergency room where they don't have access to that 30-year experience doctor or that field hospital, you know, in in in Kenya in the bush where they're trying to keep folks healthy even though they don't have access

to that uh level of medical care. So, you know, I think an I think AI can can help enable a lot of those capabilities in new and very powerful ways if it's channeled and if it's kind of implemented in a way that's thoughtful and that understands and deals with the realities of real world, you know, the security, the reliability, the privacy, the legal frameworks in which it needs to operate, like all of that stuff that maybe people who haven't uh um lived through as many of these sort of technology shifts as I have don't quite appreciate yet. >> I mean, yeah, it's probably double layered cuz even um what year did you uh leave the Lavia? >> I left 74. >> Yeah. I mean, with everything that was going on, I' i'd imagine that from a technological standpoint, you probably had like two wo moments.

I mean, moving here, I'm sure under the the you know, under the curtain, like everything was just kind of like late, right? if I'm not mistaken. Yeah. So, you come here, you're like, "What is this?" um tangibly uh seeing that jump. Yeah. Wow. That's Yeah. It's impactful. Yeah. Like, one of the most grateful things I am is that like um I have like immigrant parents because it's just like it's a little bit more humbling when you talk about everything and and you just hear like uh my mom's from Greece, my dad's from Italy, and the stories you hear are just like, "Okay." Like I remember my no no when we were talking one time about like what was school like and in his broken English he he said like school? What's school? The Germans blew it up. And I was like I was like how

old are you? He's like 10. I'm like 10. So yeah, you know, now now we're talking about all these like esoteric future capabilities with uh with TAC and it's it's it's just like a crazy uh crazy difference in in reference. Um, but it's it's it's amazing and it's it's really cool that you have that background because I think it um speaks to the appreciation you probably have for um where we're at now. >> No, absolutely. And I, you know, I I I I I'm still full of wonder in in the sense that I I think um we've still only scratched the surface and I think that we are going to see things that will seem miraculous uh in you know in in hindsight. Um you know the quote uh any any sufficiently advanced technology is indistinguishable from >> it's indistinguishable from magic. Yep. >> You

know, put AI in front of somebody from even 30 years ago and it's going to seem almost magical, right? Forget about 100 years. >> Dude, I'm a nerd and it looks magical to me and and I I pay attention to it every day. So, like >> I you know, I I'll I've also shared this experience where I've built something with my own hands and then I run it and I am mystified that >> Yeah. You you go, "What is this?" What do you think the the most interesting improvement for personal use is going to come out of AI and AI agents in the next like 5 10 years? >> I think that the ability to personalize learning is going to be the greatest asset that folks have, you know. Um, so my wife's decided that she wants to learn Italian right now. Um, yeah. Um,

five prompts of chat GPT equals, you know, a year of of Dualingo, right? >> You know what? That's true. That's accurate. I agree. >> And you can learn anything, right? It's it it's all out there in a personal context. Again, corporate context, enterprise context. There's a few other complexities and that's kind of where automately lives. But in a personal context, they the the the empowerment and the ability if you allow it to have the conversation rather than thinking of AI as as as as a tool um is where where the real power is. >> What are your thoughts on um I don't know if you've I mean I it's funny I was actually going to talk about teaching and and learning in languages. So, it's very funny that you uh brought that up. Have you interacted at all with the um uh pro version of

chat JPT? How they keep expanding the uh length of like being able to talk to it? It sounds better and better. Have you have you seen that people are using it to like train themselves on language? >> Absolutely. And uh that's actually exactly what I'm when I was speaking about uh when I talk about the >> not it wasn't written. It was it was Yeah. >> Yeah. It's it's it's spoken. It's verbal. um you know so so I've uh I've I've actually made a made a point of of always of uh playing with AI in other languages because I'm you know I'm fortunate enough to speak a couple of them. Um so I actually interact with with AI in in in various languages just to actually in in part to understand what the difference is between English AI and other language AIs um because they're

not really the same. you know, we we do tend to be very English centric in the AI universe. Um, but also, yeah, it's it's it's very cool. Um, the ability to converse, um, brings about a different kind of thinking. You know, I I know that when I'm exploring ideas, even coding ideas, when I'm trying to solve like hard problems and trying to kind of really think through architectures, I I I I walk my dogs and I talk to OpenAI and work through architecture because I have the imagination, but I don't necessarily have the minute um technical detail of all of the different tools that, you know, these cutting edge architectures are touching on. And guess what AI does? So, it's the combination of my imagination and the AI's access to knowledge, especially now that it has kind of live search tools and it's not limited

new in 2023. Um, that's that's a very powerful, very enabling uh capability. >> I think you you're you're on to something there. my my like main thing that I've I've really enjoyed in the last little bit is like even just uh so obviously chain of chain prompting was a thing and then 03 kind of threw that out the window for for from what I can tell because it's does its own like reasoning right reasoning models threw it out the window I should say and um yeah I remember mentioning to the co-host of the podcast Idkin about seven months ago when we first started saying man I can't wait for the instance where chat GPT from an API standpoint or any of these tools language models from an API standpoints it gets access to the internet it's about to go crazy you know it's going to

be incredible and then now we're 7 months later in and and we're already like uh hyonic we've already had ad adapted to the fact that uh not only obviously on the front end it was available I think last October you could like web search but people don't appre appreciate like this thing can keep running and look look stuff up and go through. >> You just expect it now because it feels natural. >> Yeah. Yeah. And it's only been half a year and I'm and I'm I'm just still baffled that like when people when it was released I was like this is like does do people not realize this is it? Like this is the moment where non-stop this thing can learn or what? And uh yeah >> so we uh at my house we celebrate New Year's instead of Christmas. So for New Year's of

2023, so like two years ago, I tried to use image generation to generate happy new year's cards for everybody just as an experiment. So that all those models were trained basically in 2022 and in early 23. And so every time I tried to generate a happy new year card, it said 2022 on it, even though it was 2023 because all the images that it was trained on had 2022 in them, right? That's only a year and a half ago. And now the image generation has gone so far like that's not even something that that you could contemplate happening now. And now I'm using images to generate, you know, technical diagrams with words and arrows and and all that stuff. Then a year and a half ago, I was trying to generate happy new year 2023. Right. So to your point, the rate of improvement of

this stuff is is incredible. It's going to take us a long time as a society and especially a as as an enterprise and corporate society to understand how that blends into our work, how that blends into our legal framework, into our copyright framework, into our, you know, um, you know, our ownership of IP framework. all all of these all of these sort of concepts that now have to catch up with the capability. Right? So we are in a very early days. It's going to take a minimum of five years for all that stuff to kind of pass through the meat grinder and come out as something at the end of it. And so I kind of view my job as to kind of try and imagine what that looks like, but also I can't wait to see it. >> Yeah, I'm really excited for it

too. And I I know everyone's probably really excited for it as well. If you want to learn or if you want to please just as we close it out, could you please give us a uh final plug for everything that you are doing over at Automate Lee so that everyone can check you out. >> Sure. Thank you. Um so we work with uh enterprises and and companies to help them understand how to bring AI automation to their processes. As I said, we focus on cooperative AI where we think of AI as a colleague rather than a tool and help you understand how to implement it in ways that can bring value today rather than sometime in the future where uh when it becomes perfect suddenly. And um we do it u with full recognition of your needs as a corporation all the way through to

being able to do fully encrypted inference and uh and other uh cruel cutting edge things that that allow you to be be compliant and use some of the ai tools that your employees use in their everyday lives and want to use at work but can't. >> Sounds great. Well, make sure to go to automately.com. That is spelled a U- T O M8 the number ly.com. Thank you so much for watching this episode and we'll see you guys in the next one. Peace.