Vidar Hokstad | The Lethal Trifecta Killing AI Adoption

0:00College football's back start trading on polymarket. New customers use promo code ball 50 deposit $10 to get $50 since the latest start trading. 18 plus C polymarket's terms of service for more information about restrictions and eligibility. The polymarket app is an independent affiliate of polymarket US and not CFTC regulated. Trading always involves risk. Complete loss can occur. The $50 offer is issued as nonwithdrawable trading credit for use on the platform. Offer exclusively available to first time depositors. Good afternoon everyone. This is Robin Ayyub from the localization fireside chat. And welcome to another recording with this particular episode today I think, number four today. But we're recording episode 276 believe it or not. So we are busy today with with podcasts so you'll see me drinking a little bit more water than usual this time of day. I want to thank you all actually for joining me on this recording. Pretty interesting conversation we're going to have here today with my guest of honor. We'll get to the introduction in a second. I did spend most of my career watching companies mix the same

1:05expensive mistakes all the time. I mean companies do that. That's normal. You make mistakes. You learn, you move on and then you correct those mistakes hopefully. And generally revolves around hiring for the title before they understand the problem. So we need X we go higher them. Whatever the title maybe, CTO, CFO, etc. But we don't understand to the depth of what we tried to correct. Today's guest has spent the same the same 30 years in the industry or in their his industry. On the other side of that equation showing up at the first technical high as the first technical higher. The co-founder and the fractional CTO who walks into a chaos builds something that actually scales. He founded, he co-founded his first startup at 19. Congratulations at 19. I don't know I don't recall what I was doing at 19. But I was up to no good. I think he has been inside the VC fund. And he's now he's naming the thing that most AI consultant was say it out loud. That there is a

2:11lethal trifecta. And you heard me talk about that on on this channel before. A lethal trifecta widely killing AI adoption in most companies. They're already living inside of it. So companies are living inside of that right now. And if you've seen the statistics out there, if you've seen and this is for the audience, you've seen the reports that are out there. McKenzie, Gartner, everybody's reporting on this now. Most AI adoptions are failing. So about 80% of them are failing. I'm going to go use the 80-20 because I don't have the I'm routing in up for you guys and routing in down because I don't have the exact number. But I think it's higher than that. I think it's in the 90%. The projects that they're actually being deployed and they're not seeing the results out of that. So why is because we don't understand the problem. So I want to really dig into that conversation today with my guest of honor and a guest of speaker on this on this podcast today, episode 276. Vida Hockstad, he joins me from London, England. And Vida, welcome to the

3:11localization fireside chat. I really appreciate your time with me. Thanks for having me, Robin. So one of the things that we talk about on the channel, which is very interesting to the audience, Vida, before we get into the topic of the hour, is the story question. If you don't mind, if you don't mind indulging the audience, what is your story? What got you here? So yeah, so basically I come from Norway originally and I was studying in Oslo and I was in the computer science lab talking to a couple of fellow students about how much better our intercollection was at home. And I mean, at that point, this was 95. Most people didn't have internet at home at all. But we were frustrated about how bad offerings were. So we decided we were going to start an internet provider because how hard

4:16could it be? And that's why I got started at 19 because I didn't know how hard it could be. And I had no idea about running a business. And you know, typical bravado for 19 year old, except the directed world starting a company because well, they were frustrated. And you know, it turned out that you could bring it to a surprising degree and get much further and I would have thought with what I knew afterwards. This is the thing. I think the best thing for me about starting at that age was that I had I wasn't afraid of starting because I didn't know what could go wrong. And we made every possible mistake. We definitively spent much more money than we should have of the little money that we raised. We got found in that one of my co-founders found an investor in a bar. That's a high list. And so like a Silicon Valley kind of an environment.

5:18Well, yeah, except it was much smaller amounts. We were talking about 40K, 4050K or something like that, about 30 years ago. But still tiny amount of the money. And we spent it very, very quickly on getting equipment and getting a sorting out of an office and getting everything set up. I mean, we managed to launch and we didn't manage to get customers in there. We also managed to start price war with the largest incumbent because we had no idea what you were doing and we figured, hey, you know, our cost basis is this. So we put our margins as low as it possibly could, not realizing that our competitor who was priced about four times that much. I hadn't probably seen that much because they couldn't go lower but because they were milking the what they saw as a relatively limited amount. So when they saw us getting customers at that low price, they probably did three days later. They dropped their prices and spent millions on

6:22driving customers to their offering at that price. And we found out the hardware in that, well, okay, you start a price war, are you better head of watch test or are you going to be released from it? So that's basically how I got started and what's followed on from that was that here we realized that well, it was pretty easy to start a company making a success out of what other hand is a lot harder. But it didn't scare me enough not to try again basically. So I spent a lot of time doing other startups after that. And you know, basically my biggest sort of takeaway from from having done those startups is that you know, when people think about startups as like high risk, you're really shouldn't because every single startup I've had that didn't go well. I walked away from way in a better position than I went into it. Because I gained experience that you wouldn't do in a normal job, I gained contact that I wouldn't

7:26have in a normal job because I had to do all kinds of jobs and talk to people in businesses I never would have talked to as a regular, if I've gone through to regular development jobs. Now, Vidar, it's very interesting at the age of 19 starting a company. But for those audience out there that they are contemplating, starting a company and they're not at the age of 19, I wish I was at the age of 19 now contemplating starting a company. There is a sort of a, I wouldn't I don't know what if you can reflect back on this one, I appreciate it because I think there is a bit of a curve that you get at the age of 19 because you don't know the ramification, you don't know the details, life hasn't really pushed you around kind of thing to put you through the experiences that you know today. And that's a good thing. I don't think it's a bad thing because when we don't know, we make that step that we're not afraid of the unknown at the age of 19, we just do it. Do you think at that point it was like looking back, do you think that was a good thing,

8:28we're a bad thing and how do you analyze that? I mean, I think it was absolutely a good thing. Am I, I got to the very different perspective on what I was looking for in my career, and what I had going into it. I was originally I was planning on going into academia and you know, studying computer science, get my PhD, do research. And I'm sure I would have enjoyed that as well. I really love thinkering with stuff, but at the same time, it will be a lot more limiting and there's a lot of things that I've done that I wouldn't have. You know, a lot of opportunities to travel and work with companies that I've owned all the wise have and learn very, very different things and a much wider range of of subfields. Welcome to the newest grocery outlet location, your couch, no parking lot, no shoes required. And the commute is approximately zero feet because

9:30now you can shop Instacart or DoorDash and get your grocery outlet favorites delivered right to your door with the same great item prices as in store. Your couch has never been this productive. Grocery outlet. Big deals now delivered. I think, you know, it's it's a great age to do it because I have no dependence to worry about no no mortgage to worry about. No nothing. So it's easier in the sense, but at the same time, you know, I've done this for a long time and I feel like I've had enough experience with it to see that. Well, yeah, sure when you're older, you will want to have a bigger financial question to fall back on. But you will also have a have a long time and opportunity to build that up, hopefully, if you have been if you've done this for a while. And it's not that you know, well, it shouldn't be that scary. It's not that much riskier than taking a regular job. There is a slightly higher chance

10:32that your company will end up folding, but you don't have control over that in a regular job either. And you have a much higher chance of building up your network, getting contacts that will help you get further. So almost all of my jobs, almost all of the companies that I've been involved in wait, where I let I co-founded or been brought in as an early technical artist has been because of compacts that I made in previous job. So my second job actually, I was it was the lawyer of our investor that had had to three of us to to accompany to one of his new clients after we've decided to sort of close up shop for the first company. You know, this is very interesting. The lawyer of your investor had hunted you. Yeah. So we had paid the investor must have been very happy with you. No, we had both we had both of the out investor at that point. And you know, things weren't really going. That's what we had sold off the dial up internet part. We both are out of the investor and

11:34we were doing some some small, you know, it basically at that point, they had morphed into a console from the sea. Yeah. And we weren't, you know, we weren't making a lot of money. It blew living off it, but you know, it wasn't it wasn't fantastic. So at that point, it was like our investor was out. So he didn't mind. So that's the kind of thing you end up that end up happening. It's you know, you might fail completely, but more likely your company will kind of keep on you know, going functioning in some form and on older and you need to make the choice. Well, do I cut my losses at it point and move on to something more professional or do I stay and try to make it go of it? And a lot of the time it is better to just decide, okay, well, this didn't go big. So let's move on. One of the, you know, one of the, as I hear you're speaking about the, you know, starting up a company at 19, I interviewed somebody on the podcast a few months back,

12:35you know, he's a young young man from Montreal. And during the conversation with me, like when I never met the person, similar to today, we never met. And you know, first time when meat is recording a podcast, the same idea with the individual. And I was taken back by how young that person is when he showed up on the screen with me. And when I was inquiring with him, when I was just having a conversation with him, I found out he came out and he told me, look, you know, I just left high school. I'm not going back to school anymore. And I just started a company and says, excuse me, can you, can you repeat that? And so he was giving me the information. He was giving me the, you know, what he did. He created an online company and he already got his first investment. I can't remember the amount of dollars. Somebody actually invested in his company and he's not even 18 yet. And as I said, you're not going back to school anymore. He said, no, I'm doing what I want to do now.

13:39So I'm not going back to school. I go back to school. There are, and today I'm assuming back when you did the first invest, the first idea that you started the first idea is different than it is today. Perhaps the tools that exist today is different than the tools existed in the back then. And perhaps if you were saying, okay, so if I have the right, if they have the tool that I have today, back then, maybe, you know, because you touch on a point, it's very important that every startup they go through that is starting company is one thing, scaling it and making sure that this company, you know, is solid, sustain the adversity or maybe crises or make it resilient to grow further. That's a different, that's a different kind of an approach to managing the company going forward. And today startup environment is different than when you started. It's very different. I mean, it's absolutely very different. There's much more matured. There's a lot more opportunities to raise capital if that is what you want to do. But

14:39at the same time, with AI, for example, you also get into a position now where it's easier and ever to start a company by yourself and do it's rapid. So your choices are very different. Back when, when we were starting, you know, just getting an office, getting the equipment we needed, and so on, that we needed to have capital to start with. And for some types of businesses, obviously that's still the case. But at the same time, you know, I have a bunch of AI agent in a chat room that I'm tasking with different tasks every day. And they act as if their, you know, as if their coworkers take carrying out tasks for me that I would have had to hire people to do before. And so you can get to a completely different scale before you actually need to raise money to hire people. Another, I think also, I mean, means that in a sense, the risk is lower

15:40because you can test ideas to a whole different degree. You can, you know, launch, launch at the site, a bit actual functionality in a fraction of the time it used to take and see whether or not it gets an interaction. Before you start taking investment that hiring people, doing all of the things that are high risk. No, you've done this six time, no, Vidar, right? So six startups, six plus startups or more probably. And, you know, that that involves a lot of stuff, like the funds, the exits, the learnings, the failures, the successes. And then you came back and you built a consultancy, a fractional CTO today. Like why now? And, and, and is that, you think there's opportunity right now in this or it's not really, yeah, so it's kind of not so much of why now is and why now again? Because maybe why not to it? Yeah, my, the thing is my company is, I first set

16:41up my company over a decade ago. But what keeps on happening in these kinds of situations is that you keep on starting for a client and the client will come to you and go, oh, but why don't you go full time? And, you know, that happens regularly. It's just happened again. And, you know, I have to have the conversation of, well, do I want to do that or do I want to continue doing this? And, well, in this case, I definitively want to keep doing this to some extent. So the question is, well, do you, can you make this work together? And, you know, every time it's been a different conversation, one of those times was when I ended up going to work for, for this venture capital from where I had built their systems to begin with and I went, oh, you know, we can't justify continuing to do this, contract basis. So I do we need to find somebody to hire full time or you can come work for us full time. And I got it off, I couldn't refuse. So that's the nature of that kind of business.

17:47It's also very tempting when you find something you're excited about working it to sort of taking it on full time. Now, what's your opinion, Vidar, on fractional work in general? Because I'm not sure if you knew, but I'm a fractional CRO and also now fractional managing director. So what I do is I help people go to market in or, you know, evaluate, increase, you know, all that stuff that that CRO will do. I do it on a fractional basis. And also on the managing director's side, is a lot of, I've noticed like in this multimedia and size companies where the owner is still involved, they're getting tired. I mean, they don't trust it to hire somebody or probably they don't have the fun to hire somebody where I couldn't come in and help them out. What is your general opinion on that fractional market now or fractional economy that I see out there? You know, I think

18:47is more relevant than ever because it's gotten so easy to start companies, but it's hard to get, it's hard to scale to the point where you can hire whole people because that's expensive, especially if you need somebody very experienced. And part of the reason why I've been doing so much, contracting, consulting, work, I've also been exactly that I've got to a point in my career where I started getting too expensive for most people to hire full time. And at the same time, it's, it's, you know, it's a safe car for me against getting bored. When I have my price in many different, you know, my fingers in my different price at the same time, I work with, work with different teams and that kind of thing. So I was quite, quite enjoy doing that even when I'm doing like sort of mostly full time on all clients, I like having my fingers in a couple of other projects to have something different. And I mean, thankfully, I feel the market, the market

19:53is quite healthy for that kind of thing now. And even when I'm talking to clients like now, and I have somebody who wants, would like me to go full time, you know, they're quite open to being flexible because they recognize that it's better value in having people who are constantly learning by doing all different things as well. The question regarding frictional work, do you think like the, there's more acceptance to that? Like, I feel like because the other day I went to a networking event and there was a lot of fractional workers out there. But you notice, you know, one thing in common is the age. We're all at an age where, I'm not saying everybody, I'm saying the majority. There are many people that they are, let's say, either on the edge of retirement or already retired and they're using it as a second career, etc. And they still have something to give, some energy to give. You don't like like your retirement, that's for sure. You're too young for that. I was wondering if you feel like in that

20:56fractional dynamics that's happening out there, age is playing a factor. Welcome to the newest grocery outlet location, your couch, no parking lot, no shoes required. And the commute is approximately zero feet because now you can shop Instacart or DoorDash and get your grocery outlet favorites delivered right to your door with the same great item prices as in store. Your couch has never been this productive. Grocery outlet, big deals now delivered. I think so, I mean, when I was younger, I, you know, even though I was doing startups, I wanted to have this, that I was looking at the affordability of it, even in a startup. You're looking to go to their job and work on that single thing. But I think as you know, as you're getting older and more experience and senior, you get to a point where as I mentioned, as well as the physical factor in it is hard to,

21:56the pool of companies that can afford you get smaller. And so you have to make a choice of all, either you're chasing bigger companies, which I'm not really that interested in. I quite like working in like small, medium companies where you actually get to have a say in all everything. I like putting into things. Or you find ways of doing multiple things. And if you, especially when you're working in those smaller companies, and you're senior enough to have gotten expensive, it's easier to justify your fractional work. No, Vidar, like when a company perhaps call you and say, hey, we need you as a fractional CTO to come and help us out. What does day one look like? And if you don't mind walk the audience through perhaps, how do you diagnose perhaps the tech stack that you've never seen before? I mean, this is a brand new company for you. What do you start? So I think that the first thing is to figure out, well, what do they mean? If they're hiring a fractional CTO or a fractional, I don't worry about the titles because some people will

23:01call it CTOs and people want to tech lead, whatever. And it really depends on the size of the team. What that role looks like? Because if you're looking at a small team, a CTO role is glorified team lead. It's somebody who is going to be doing a lot of hands on development. If you're looking as well at a larger team, it's purely strategic role and you will never see a text editor or, you know, no idea. And so the first thing you need to just figure out is what's the dynamic here and what is it that they actually need help with? This is the strategic partner, it's the hands on part. And for me, that's kind of part of the fun of it because I'm currently, my biggest client is one where there's a lot of hands on work, but there's also some strategic strategic work and it's not labeled in the stock as a CTO role. There is somebody with the CTO role that I'm speaking to. And that's fine. You figure out where you slot in because you've gone for hire. So one of the first

24:06things I will say is, well, I will take on whatever role you actually need me to do. And irrespective of the title, that might turn out to be very, very different. But especially in small tech companies, they often have this idea that they need somebody that's going to be, you know, a big, as CTO for a big company, as opposed to what they actually need right now. And that's actually a one of the reasons I think that these kind of positions, the fractional roles are the most important because you will hire somebody who has probably been several places around that size ladder. And, you know, they might have like me worked as the first technical hire up to the point where you have a pretty decent size team below you. And they will be able to look at it and see, okay, well, you're a five person team and you're not going to be looking at somebody who's

25:06who's going to be negotiating the Fortune 500 companies because you're not the size where those people are going to want to get in a room with you. They're not going to be interested. You're going to need to figure out the smaller stuff first. And that might mean looking at how do you raise money so that you can hire enough people that you can actually go after those big accounts. And the technical bits around that, but it might also basically be getting better down with the developers and figure out what actually how do you finish this prototype. So you have something you can show to investors six months from now. I do like the idea of it or of what you just said, you know, I'll be whatever you want me to be within the domain, obviously, within the category that you're trying to work with. A lot of people they stick with, I don't do this, my, you know, very narrow focus. And again, it goes because you're because individuals who work in fractional environment, they can push the envelope as much as they want to and they can retract as much as

26:09they want to, right? So they can open the possibility to the customer or their customer as much as they want or they can retract it as much as they want. There are a few things though, and I'm just using myself as an example that at this point, I'm using the your words I can do whatever the customer asked me to do, whatever they want me to be, I will be. But there are things in the sales environment I will not do right now. Like, and for instance, an example, I will not be a lead generator. I can introduce you, but I'm not going to be the guy who picks up the phone and call 200 times a day. I won't do that. I've done that when I was a kid. I'm not doing it now. However, I can settle that up for you. Now, in the process, I mean, do you run across something like this? Like, I mean, you're not going to be the tech guy who's unplugging things around. Are you do you do that? Well, there will be cases where you ask to do something that's just not cost efficient firm to have you do. In that case, yeah, I will tell people that, yeah, I could do that for you, but it's going to be very much more expensive for you to have me do this. And I have also done the call,

27:13call calling actually in my first company because I hated it. I absolutely hated it. You know, in a situation like calling, yeah, no, exactly. In a situation like this, it's not necessarily so obvious, but I am notoriously introverted. If I can get away with it, I will sit in front of my computer and not talk to people. And so in a situation like this, it's fine because it's a friendly conversation. But if you're doing call calling and people don't really want to talk to you unless you call in some of it first, it was awful. But you have to do it. And so I will do even that if necessary, but I will be very honest. It's, you know, clients that, well, it's not going to pay for you to have me do that kind of job. That's a very valid point. The cost is very valid point because I interviewed somebody on the on the on the podcast about fractional work as well. And what did he say to me? He said, Oh, if you are closing every opportunity, that mean

28:18you're not charging enough. And that resonated with me. Are you charging enough for, and I'm always asking myself the question, am I charging enough for the services that I'm offering? And the point in case some point, if somebody's going to say to me after I, you know, I send them my invoice, hey, I want you to do call calling. You know, if you want to do call calling at the price that I'm charging you, go for it. But I don't think it's efficient for you. Yeah. Yeah. I mean, that case is very valid. Well, in that case, more, what's more valuable to them is getting your advice on, well, okay, you should hire somebody at much lower level. And to me, I, I, one of my favorite things to do when I get, when I get approached by recruiters, for example, is when I can tell them, sorry, but, you know, I'm too expensive for your client. You should be looking for somebody at this level. And change their expectations. Yeah. Yeah. No, you've said, like, tech teams are often costing the company far too much. Have you run across where

29:23there are some rooms for cost efficiency and some of these teams that you were involved with? Vidar? Oh, absolutely. I mean, I'm kind of like halfway far in myself in positions in Pulse where I was like, okay, well, I'm dumb now. You should hire somebody cheaper to do this or you should, or I should go and you should keep one of those other developers that you were planning of hiring because, you know, if somebody needs to cost cuts, you need to look as well. Okay. Do they need somebody who can, who can do all of the things I can do, or do they need somebody who can just keep the lights on and do something more simpler? And, you know, it's much better for me in the long term as well to be the one to tell a client that, well, you know, at this point, I'm not needed anymore. hire me again when you have a problem where I'm actually cost effective. As well as, you know, I'm looking at teams sometimes and I'm seeing them. And a lot of the time it's not even the people. It's just that the structure they are,

30:27they are building for something. They think that they are a 100 person company when they are a 20 person company. And that means they have roles that they shouldn't have. It's not even necessarily that they need to fire somebody. They need to put somebody in the right role in change roles because you can find people that are actually capable of doing the right thing, but they've been put into leadership role, strategic role that should not exist because they're not large enough yet. Not kind of thing. No, Vidar also another discussion that some of these companies are having when it comes to their IT stack is the platforms that they're on, like AWS, GCP, Azure, all that stuff. So when they are locked into one platform, it sounds like I've seen it before in my days when I was working for a large corporation, when you're stuck with or it's not stuck, when you commit to one platform, scaling inside of that platform becomes costly. How do you manage that? How do you handle this

31:29conversation either to move away from this platform or to engage in a different conversation to perhaps reduce the cost or rely on a single platform? So I think a lot of the time it's a question of, well, what's your priority right now? And sometimes it's fine to stay on say AWS because it's not a big enough deal right now, but you should at least think about it because some of these platforms are all of the big cloud providers. Basically, I have learned a lesson that I hadn't that 19 that you don't start a price war when you can't actually come out of it better than you went into it. And they are the big cloud providers are in a position where they collectively, the top four or five of them own enough of the market that if they start dropping their prices, often have sufficient margin that they can often drop the prices enough that no matter how much of their main competitors market share they take, they won't recover the lost earnings. And so there is, it is near interest to drop prices to a point to a level where they're actually

32:37competitive with the smaller providers that don't have the money share. And you're paying for us. But the question you first have to ask is, well, does it make sense to move now? When you get large, it almost certainly will make sense to move unless you negotiate a really, really good deal with them. And this is part of the reason why people stay with places like AWS is because they're looking at AWS and saying, I, at one point, naturally, it's also customer, I don't think you remember whether or whether there's still a customer like AWS, but they're not paying list prices. Nobody who's paying more than six figures, possibly at least seven figures, three WS every year are paying list prices. They'd be stupid. Even at like seven figures, you can get, you know, tens of percent of discounts below the market prices, often 50 percent or more. If you're at Netflix size,

33:37the margin AWS are making on the earth very, very low. For the very simple reason that, while if you're looking at lower tier providers like HETSner, for example, a German provider or OVH, or any of the smaller providers, you can get things there at third to 50 percent of the price of AWS. For certain things like bandwidth, you can get bandwidth at around two percent of the price you're getting at AWS. So you're two percent. So they're willing to make a deal, right, Vidar, like they're willing to work with their customers? They will absolutely make a deal once you get big enough. And that's the thing a lot of people think that while they're self-service, so while, you know, you're assuming there's not the grocery thing, but at the moment, you're big enough that you have like half manager, you're big enough to start asking for ban for deals. And the more it looks like you are able to move, obviously, the better leverage you have. If your platform is tied to every AWS

34:42service on the earth as long, they're going to be looking at it and thinking, well, you can't, it's going to take you three years to entangle from this. So your discounts are going to be according to that. If everything you're running is running, running on tooling, that is trivial to lift and shift somewhere else, you can go to them and say, hey, you know, we have our disaster recovery set up in Google Cloud already, or we have our disaster recovery running on Hetsner. And we know that that's, you know, we can see it's a much cheaper provider for us. You are going to get that price from AWS. That means that maybe it's not necessary for you to move. But having the ability to is going to unlock a lot of potentials and savings that way. And having the ability to move is also a great risk management move because, well, who knows your account might get locked? It happens. It's not frequent, but it's a risk you should be aware of. Or, hey, if you're hosting

35:43somewhere unstable, AWS have been unable to restore data for some of their customers. And all of their mid-leaster data centers after it got bombed, it happens. So, you know, it might go to the bombs, it might be a fire. You should be prepared for those kinds of things as well. So it means you should at the minimum have backups outside of your main cloud provider. And if you can stand up a requirement, an environment with somebody else as well in a reasonable amount of time. And it doesn't need to be that you can stand it up tomorrow. It might be that it will take your week, but if you have the process is ready and you have backups and you have the data, and you know how to do it, you stand a much better job at, at both surviving bombs and surviving and negotiation with the night of yes, a counter-executive. Does it go back to the architecture, Vidar? Yeah. I mean, yeah, so there's a lot of simple things you can do there, which basically,

36:44I mean, the first thing is look at every service you're using and ask yourself the question, is there an open alternative? And so it isn't even necessarily that you can or that you need to use services that are open at your main provider. You might use AWS Aurora, for example, as your database. And it's fine because even though that is AWS-specific, it is compatible with Postgres or MySQL. So you can then test your setup on one of those databases and confirm that, yeah, you can. So it's fine to take advantage of special capabilities. Your cloud provider have as long as there are a close enough analogues that you can make your platform agnostic to them. So that is my main sort of architectural advice is make sure you can move. And the upside of this is that, well, if you make sure that you can set up your entire infrastructure

37:54on commonly commodity, common commodity platforms, it makes your developers job so much easier as well. Being able to stand up a copy of your environment on their dev machines, for example, is invaluable. It makes things so much easier. So your first starting point might just be that. You make sure that you can stand up a copy of your stack on a development environment. If you can do that, then you can move out of your main hosting provider as well if it becomes necessary. No, before I move to AI, I still have one question for you on the topic of certification. Most of these companies are chasing the ISO certification when it comes to IT infrastructure. And there are many of them. The one that comes to mind would be the ISO certification regarding security, regarding finance, and the financial market uses another certification. I'm just forgetting the name right now, but you're more familiar with the whole ISO certification

39:00when it comes to that. So how does that fit within all these platform movement stuff? So it's our kind of extended helps because being able to show that you can bring, for ISO 27 and now, one, for example, being able to document your recovery, though the time to recover, the time to bring your platform back up again, the time to the ability to restore backups and those kinds of things are important. Now, to some extent, it might be easier to document the basics if you are on a platform like AWS because there are platforms built around extracting evidence from your setup with those providers. But at the same time, you can often, having a platform that is completely ready for a lift or shift is another provider makes it also quite easy to document that, yeah, you have the ability to restore because, hey,

40:02you have your backup in a separate location, it's locked down, and you can bring your entire platform up from scratch and sort of here's evidence because you can actually demonstrate that the entire platform can be stood up from a backup or nothing else. So that works out quite well. Now, there are some areas where it is very convenient to be in one of the big cloud platforms and that are around things like access control and auditing and stuff like that because all of the big providers have solutions for that, that your auditors are going to be comfortable with. So it's a trade off. I think it is more important to be ready to be able to shift the parts of your infrastructure that gets costly on the big providers out than it is to actually do it. Because if you are spending 2K, 3K, 10K even, amongst, it might not matter to you, depending on

41:06your margins. This is another discussion. For some people, it's okay to spend 100K a month on AWS because their margins, because the hosting environment costs like 2% of their technical costs and most of it is the engineering team. And if your hosting is 2% who cares? Unless your margins are really slim. But for another business, let's say you are doing something that is really bandwidth hungry. Chances are that it's going to drive your AWS build Triduruf and you might be in a situation where your hosting costs are 99% of your technical costs. It becomes critical to be able to move out. But for most companies where the costs are smaller, it's not going to be a question of, well, you need to move today. It's a question of,

42:08you should be ready to move once the proportion of your costs taken up by a cloud environment is getting to a point where it's painful. Now, does this also apply to SOC 2 type 2? Or are you guys in Europe don't have to do with that? Well, we do have to deal with it because so many of the companies will end up having a subsidiary in the US or a high sufficient exposure to US clients that still need to deal with it. So, yeah, I mean, it's a lot of it. It is very much similar. It's generally almost all of these certifications are more about while documenting what things are like and having a basic structure in place. They are generally not that onerous in terms of what they are actually requiring you to do. As long as you are documenting everything and as long as you are almost with your customers about it. So, you might have a recovery type of trajectory, for example, that says

43:11that while it will take us a week to get back up again if the system, you know, if the data center gets bombed. But you need to tell your customers about it. And there are certain things obviously you cannot just get out of that way. You cannot tell your, you cannot get away with telling them that, hey, we are not going to protect personal information. But most of that also comes down to fact that while you are doing the sensible things, you are, you have proper, you know, you are keeping things patched, you are documenting your password policies, you are enforcing them. And most of those things will fall out naturally once you're actually going to a checklist and making sure that you're doing all of the sensible things. Welcome to the newest grocery outlet location. Your couch! No parking lot, no shoes required. And the commute is approximately

44:12zero feet. Because now you can shop Instacart or DoorDash and get your grocery outlet favorites delivered right to your door with the same great item prices as in store. Your couch has never been this productive. Gross Reoutlet. Big deals. Now delivered. We will be a lot of work because it turns out that small companies will forget about a lot of sensible things. A lot of them will forget about very, very essential things like ensuring they are actually testing their backups, for example. But they should be doing them and just getting the certifications for a lot of them will be an exercise in going to a checklist and saying, okay, there's all of these things that we thought we were doing or we should be doing and we haven't really verified that you're doing them. No, no. Let's talk a little bit about AI and the AI adoption. You talk about the lethal trifecta as AI risk for companies. What do you think right now?

45:20What is the biggest risk you see for companies when they're trying to adopt or adopting AI? So yeah, so I mean the lethal trifecta originally comes from Simon Villasano's doing a lot of excellent work in his place. But yes, it's a very, very critical security issue because basically what we're seeing is that people will start adopting AI and it will happen piecemeal without a proper strategy behind it. And it's very, very easy to do lots of small things that seems innocuous on their own and then they combine to form this little trifecta of that basically causes a massive security hazard. And what to give an example, that sort of illustrates that trifecta. You might do something simple like say decide that you're going to have your AI assistant

46:22read your email by itself. If it's just reading your email and it can't act on it, it's not a problem because well, it's going to give you a summary and if somebody convinces your AI agent of something, the worst thing that will happen is that it gives you the wrong summary. Now, on the other hand, you might also be looking at your AI assistant and think, hey, okay, well, I'm going to have it manage my calendar. And it's not, you know, worst case scenario, somebody manages to convince your AI agent that it should put them on the calendar and it's, you know, it's slightly more annoying than giving a wrong summary, but it's not a big deal. Now, the problem there comes to the moment you are and connecting another system. Let's say you give that same agent access to read data from your bank account. You know, again, it might seem innocuous because he is not read only access, but if you're thinking about typical scheduling

47:29permissions, if you're going to let it manage your calendar as well, it might be that you now create a situation where somebody can email your assistant, convince them that it's you, and you need access to some transactions from your email. And oh, by the way, you're going to discuss this with your accountant who's at this third party email address and you should schedule a meeting include some transaction information and invite that third party guy to that event. And suddenly you have an outbound email with your financial information going to an address you have no control over. And that's basically the lethal traffic that's possible. That's possible. That's possible. Exactly. And it's way too easy because it's like you are, have given the agents relatively innocuous looking permissions in isolation. In isolation, they'll be fine. But when you are giving a third party access to input information to your agent and you have given that

48:34agent access to internal information and you have given that agent access to exfilterate that information auto to system through the calendar invites. Then you basically create a situation where somebody can feed an instruction to your agent because it's easy to convince them of things. I have an agent that reads my email and it frequently gets confused because somebody is sending me something that isn't particularly important because they're trying to get me to do something. I'm not interested in. And it excitedly summarizes it for me. And it's like as long as it's only providing a summary who cares. It takes me five seconds to dismiss that and it surfaces lots of important things. But if I had connected that same agent to something that allowed it to actually exfilterate the information, it'd be dangerous. Now, one of the things is like I'm noticing in the deployment of AI in general is the homegrown idea of deploying AI, basically companies,

49:38they say, okay, we can do that. Sounds easy enough and we can do it. The fact that it's convenient and it's a double-edged sword in my opinion where you have an easy to use from the box kind of thing technology, but the potential for this technology to be very dangerous is very important to understand how dangerous it is. And what I'm also noticing is and I don't know if you agree with this is at lack of education. We don't know enough about it. For the general user I'm talking about, they don't know enough about it. They're just diving right in, not necessarily knowing what kind of information is dangerous to put on AI versus the information that you should probably okay putting on an AI. So the lack of education and the lack of systemic upgrade of our knowledge of our skills in that domain is unbelievable. And there is the, I'm thinking there is a sacrifice between convenience. I want to get this thing done now versus I want to wait, study it, analyze it,

50:41see the impact of it, then deploy it. And there is another one we talk a lot about on, I run a part, I run, sorry, I run a, a networking event. It's called the Global AI Exchange that takes place on Fridays. And we talk a lot about this and you welcome to join the Vidara if that's interesting to you. I'll send you the registration. We talk a lot about how people are using AI. Now what are the use cases? And we noticed that a lot of use cases are fraught with problem because we're overlaying AI on top of bad processes. So when you put a new technology that affords you a lot more new ways of dealing with things onto an legacy, what I call not necessarily a bad process, but this is a legacy process. It was not designed for the new technology that we are trying to deploy. And people are overlaying technology on top of that AI on top of that. It's sort of exasperating the process and it's making the process a lot more visible now

51:43that there are issues with it. So have you seen this where companies are overlaying technology, AI technology on top of what they, what I think they should do is revise the process first. Absolutely. I mean one of the really frustrating things of working in software development is how often we have outdated documentation. And so it's often the actual process that is actually being followed. There will be some variation of the original process tweaked by institutional knowledge of all of the part of the documentation you should ignore and that nobody is looking at. And the new point on AI agent that it and the agent will read through all of this documentation and think it all applies. And so you get the situation where you have a process that there's no resemblance really to the actual documented process that you know, according to your repository.

52:51Welcome to the newest grocery outlet location. Your couch, no parking lot, no shoes required. And the commute is approximately zero feet because now you can shop Instacart or DoorDash and get your grocery outlet favorites delivered right to your door with the same great item prices as in store. Your couch has never been this productive. Grocery outlet. Big deals now delivered. It's a massive, massive problem. People will apply AI and think that oh yeah this is all going to work right. It's going to read all our documentation without realizing that yeah it might not even have you know it likely didn't work in the first place and that is why people are working around it in all kinds of ways. But those revisions haven't made it into the documentation. And I actually had one of the startups that it about 25 or 26 years ago now involved setting up the dot name top level

53:53domain. And one of the things that we had to do in preparation for that was to handle a sort of land rush process, basically pre-registrations that we needed to handle fairly according to our contract with the object I can handle that was the tour of different top level domains. And we had to go through that process so many times with excruciating them all the detail to make sure that people actually adhere to the process as written and that the written process actually because if you are giving a person a process they will infer all kinds of things and fill in all kinds of blanks in a way that an AI model won't or it might be inclined to interpret things the different way that what a person would have been filling in the details. And it's incredibly hard to get a process so well documented and streamlined that it's actually something you can follow word forward. And so if you point an AI at a model at a workflow that is intended for a

54:59human that might have years of experience within your organization, I mean it's going to get work just as bad if you point a junior person that is like first day on the job at this process and tell them well you can't ask anybody. The straight, the straight. Now I work in the localization industry that's the language industry for short. And I've seen like multiple waves of technology probably can attest to that year in the business. I've seen like things like translation memory, then machine translation, then now we have large language models. And each time the companies that suffers the most were the ones that either moved too fast without governance or too slow without courage to move. So you got you know too slow and too fast dynamics going on. Now with your frameworks in mind with framework in mind, what do you think are the one thing companies in a language services space should do in the next six months or now? Well that's an interesting one. So I've

56:02kind of been doing stuff adjacent to that quite a few times because it's something you get getting involved in. And actually I did contract work on a large language model for a company who had it at the sideline besides their translation services as their main business. And yeah, I mean large language models on one hand are fantastic at doing translations if they're closely monitored. But at the same time there's so many risks there in terms of of not understanding the context correctly or the models thinking they understand the models sufficiently and then making changes to the source material when they're translating things that they shouldn't have. So on the upside I've seen things like I mean as I mentioned earlier I'm originally from Norway and I periodically tested these models to see how good they are translating to Norwegian and to Norwegian dialects and they can do that shockingly well. But then you

57:10come across cases where they will suddenly substitute a word for what that is, you know, mean something different and suddenly different or that that might be a correct translation but that is unsuitable within the context. It's an incredibly hard problem to solve because even with you know even with inexperienced translators you will have run into that. I think you need to you're still at a stage where you cannot let them do this and monitor and you will need to do a lot of quality control and a lot of work in making sure that they have sufficient context around what's what is actually being translated and I think that's a hard one because a human will come today and have a lot of they take the easy way out. They just give you less context. No, you know I mean we're coming to the end of our conversation there Vidal

58:15and I want to make sure that you know there are a few couple more questions that they're a little bit you know a short. So if you had to leave every founder out there or CTO and maybe a CTO and the localization industry or tech leader in the audience maybe a concrete thing that they can do differently when it comes down to AI and AI adoption. Let's say starting next week on Monday when they come to the office. Let them let them go through the week okay now then Monday comes and they have a fresh idea. It's a mind maybe a mindset shift maybe an actual action that they need to do. What would that be do you think? So I think the biggest mistake that people do with AI today tend to be that and that sort of separates the ones that are getting a lot out of it from the people are we might feel they are getting a lot out of it but they're still spending too much time is to stop babysitting your your AI set concrete measurable goals and start it on a long running

59:22and then close you believe that running in a separate window don't sit there watching it and because that is like watching over the shoulder while somebody who might not be quite at your skill level do something and then second guessing everything they're doing. You obviously don't you want to keep an eye on it if it has access to do something dangerous but you publish it not giving it access to do something dangerous in the first place you should be running it in isolated environment by kind of thing. But wait until it's done and then the review what it has done review the outputs rather than every single task because it is not going to do everything right but a lot of these mobile are getting very good at eventually getting there. So the question regarding AI adoption in the localization industry you're right I mean the the human needs to continue to be in the loop kind of thing and need to be involved and those are

1:00:22the guardrae and we're seeing this quite a bit in every technology like I had the opportunity test drive a Tesla in Canada for instance for the audience that's pretty interesting because in the US for instance you can go drive a Tesla there is there is a version of like actually the cars that you're driving in the US they don't have supervised a full self-driving mode for instance. In Canada you're not allowed to do that they have to have a supervised self-driving mode and most cars right now they have a they are in that capacity they are in that type of technology. You still need as a driver you still need to be self involved you need to be involved in driving the vehicle similarly in AI in the deployment of AI you still need some sort of an involvement in deploying these types of technology and the reason I'm saying that is because not that you need to be so micromanaging this you can if you want to but why deploy technology if

1:01:23you're going to micromanage the technology like you do and that becomes the style of an individual. I've worked with some individuals managers that they absolutely want to micromanage every single task that their staff is doing that's not management anymore you might as well go do the job. If you are going to deploy technology you may want to give it a bit of a room but you may want to also be involved as well does that make sense with our did I get that right absolutely and I think it's about managing risk right if it if it has the chance to do something really dangerous like in self-driving car that's not quite ready please. You will want to have somebody actually sitting there monitoring but for a lot of tasks we can give them an isolated environment we can have air go off and do things and the only output is a set of changes and those changes should be reviewed but you don't need to sit there and watch them making the changes in case they make a stupid mistake midway what matters is better that changes but that mistake is still there when

1:02:27it's finished and there are so many techniques to actually look at well okay you you're not going there's no point in looking at it before if you're making changes the software system for example there's no point in looking at those changes before it passes your test suite because if it doesn't pass your test it is broken yeah so why sit there and watch it while it's fumbling around trying to get something to pass your test suite so I think that is that is that is it's a matter of sort of managing the risk and once you have a finished change that you want to push into production then yeah then you have a risk you don't have a risk when it's working in a isolated virtual machine and can't do any damage and at that point they are getting so cheap that in some cases you might even be looking at all okay well there's an opportunity risk if it does the wrong thing and you're not getting any value out of it well okay so run three of them have three of them try three different approaches to solving your design problem for example and you triple your chances that one of them might do some things might do things the way that you're happy with and you know that for some

1:03:33types of changes that's going to be too expensive but for an increasing number of things you can look at a sensing okay it's almost throw away you software is just some extent becoming a a question of okay which individual changes are throw away you can have it do five different designs you can have it do three different attempts of fixing our abog and it's as long as you're not sitting there and spending your time watching it that might end up being cheap enough because it's a matter of well tokens that's the less a sell you cost to that horse unless you spend unless you're using billions of them like I am yeah now where should people go to find out more about what you do and how do they connect with you where do they follow you and interact with you yeah so LinkedIn is is easy my name speeder hoxdow I'm the only be the hoxdow in the world so I have nowhere to hide but also my website's hoxdow the consulting look absolutely and for those individuals who

1:04:35are looking to connect with vidar and you happen to be connected to me if you want an introduction please reach out to me happy to do an intro to vidar that be an honor and pleasure to do that so as we come to a close here vidar anything else you like to add to the conversation I think to go back to the start basically don't be afraid of trying new things because I think that's too many people get sort of stuck doing what they're comfortable with because well it's the least chance of failure but I learned a lot more from my failure stand for my successes because when you failed you you know you know that well okay that was the right thing when you succeed it could have been luck it could have been hard work it could have been just happening to be there at right you know at the right time so you get a lot more signal out of your failures than your successes so you know the more you try things and fail the better you will learn absolutely and on behalf of the audience and myself I want to thank you for being such a good guest

1:05:38for us today and for helping us record episode 276 well thanks for having me and for the audience I want to thank you so much for joining me on this podcast I really appreciate it and if you're watching us on youtube thanks for watching and if you're listening to us on the podcast channels thanks for listening in and if you're checking out our blog posts where I archive these conversations in a small calls note version of the conversation a small pieces of information where you can browse through them through them quickly do your research etc thanks for doing so as well and if you haven't downloaded the app yet go to app at l10nfiresightchat.com and have the app with you where you can see all of the content gets pushed out regularly each episode generates about 30 pieces of content I hate for you guys to miss it go check it out and I want to thank Vidar Huckstead for joining me today and I'll see you guys at the next episode thanks everybody see you next time