You can lose your job if you will not learn how to use AI properly.
SPEAKER_01How do you encourage or educate them on what human oversight looks like beyond just, oh it looks good, let me send this email, let me send this board deck.
SPEAKER_00You give your work to AI and it executes for you, but you're still responsible for what you are doing. And you're responsible to validate it after the work is done.
SPEAKER_01I got one more question on the hackathon concept. What surprised you most about the non-technical participants?
SPEAKER_00People themselves didn't believe that they can do something. As soon as they just started, they started to believe. And uh from you know, uh boring and uh confused faces, after just a couple of hours, I saw the entire company in the office uh with excited faces. Nice. And uh, this is really new territory. What can happen, we don't know.
SPEAKER_01Hey everybody, welcome back to another episode of Using AI at work. My name is Chris Dagle, I'm the host. And before we get started, I just want to let everybody know that over the past two and a half years we've been running this podcast, uh, we've had a certain format that's been a little loose and organic, and me just kind of exploring conversations with uh AI experts, AI professionals around the world. Starting with uh, I guess maybe in the next couple of episodes, we're gonna have a very set format that I think is gonna make a lot of sense to all the listeners. We'll be revealing more about that in upcoming episodes, but I want you to know that um the show is uh in our effort to always improve the show. We're gonna continue to uh bring on fantastic guests like we have today, but also improve our format. And speaking about our guests today, our guest today is Dennis Romanowski. He's the actual the actual chief AI officer at Soft Swiss, which is a B2B iGaming software company that serves sportsbook, casino, and online gambling operators. Um Dennis is a rare uh guest type of guest that we have on the show because he's a real world example of a chief AI officer and the I guess the role of business transformation that he's playing, not just a technical or a data science role. And from our earlier conversations, Dennis has indicated that he explicitly says that adoption, change management, security, literacy, and department by department enablement are central to the job. And that's what we're actually going to be diving into today. The types of questions such as who owns AI adoption, how do we control cost? How do we prevent data leakage? The types of questions that I get regularly when I'm on site with clients. So, Dennis, before we get started, uh, you know, I'm I'm kind of interested in the journey that uh whatever your your uh we could do it briefly, but whatever your background was that's resulted in you now becoming a chief AI officer.
SPEAKER_00Hi, Chris. Um Hi Dennis. It's a pleasure to be here, right? Yeah, quickly jumping into the questions. I'm an engineer. Yeah, and uh spend uh maybe like 20 years in software development, but you know, my background is financial accounting, so it was kind of fun that uh I was programming, but I uh studied uh finance and accounting. And uh uh because of that, you know, I always worked it a lot with the clients. I understood the pain from uh IT technologies, yeah. And um when the era of uh AI came and we understood that it is a little bit different, yeah. We saw that uh you know we need a person who can understand uh the technical side but who can uh clearly talk to people from you know from business, from services, from operations. And uh this is I believe where I feel quite well, and I'm uh uh working different uh I'm working in different worlds now together, and uh I think it helps me a lot.
SPEAKER_01You know, that's one of the things since the very beginning of of me getting involved in AI, I saw a video, it was an interview with Peter Diamandis, if you know who that is. And this is March of 2023, and this uh YouTube interview with with Tom Biliew now has gotten probably three million views. But within the first minute and a half, Tom was like, hey, listen, for somebody who like a lot of the listeners, for somebody who's interested in going all in on AI, what like how should I be approaching this? And Peter said, I'm advising every company that I talk to to hire what I call a chief AI officer. And he said that to me it's not a technical role. It's not somebody who's who's building a large language model or fine-tuning a model or the data scientist, it's somebody who understands business operations and generative AI and is able to blend the two. And it sounds like even with your technical background, that's kind of the role that you found yourself. So, Dennis, I have a question. When Soft Swiss decided that they wanted to create this chief AI officer role, what was the business problem that kind of catalyzed them making this decision or that the problem they were really trying to solve?
SPEAKER_00In business, uh we look at it from two perspectives, from uh opportunities and from risks. As a company, we understood that you know the market changes very quickly. And um, today you have one situation, tomorrow it's absolutely different, and you need to adapt. And we understood that VI, you know, the speed of this change will be even is even impossible to predict. So we decided that somebody has to focus on it, understand this technology, and explain everybody how they can you know adapt better to it so that uh we can, you know, first of all, uh not be worse than our competitors and then be better than them. Yeah, can we like uh take uh you know uh more like market share in some areas of business? Can we do better? Can our service be better? Can we be more cost efficient? So basically those are the things that we had in our mind.
SPEAKER_01You know, and I I guess in the iGaming space, like for instance, if I'm a manufacturer, a new competitor, they're not a threat anytime soon. They've gotta get everything set up. But in iGaming, I guess with the right uh technology and the right backing, I could be a competitor to SoftSwiss tomorrow or within a week or a month or whatever, but not nearly as long as traditional industries. So that makes a lot of sense.
SPEAKER_00Yeah, yes, yes, absolutely. Because, for example, you know, now if you have uh a good uh set of tools like you know, uh Anthropic or OpenAI, you can like alone try to build uh a gaming uh website and run something. Of course, you will later see some uh complexities in how this world works. Yeah, that you need licenses, you need to talk to multiple different providers, but uh you know it eases the start in this business for some uh for some people for sure. And this is a risk, yeah.
SPEAKER_01So a lot of our listeners, or at least our our audience that we target with our our services, they're the lower middle market companies. They might not have the size of the team that you have at SoftSwiss and that sort of thing. What would you say that a mid-market company do if it can't afford a full-time chief AI officer or or can't find one that's qualified?
SPEAKER_00Well, um they can uh for example uh take one of the roles and uh just uh at this uh functions. I believe that, you know, in a smaller company it's easier to share this, for example, with uh uh CTO role or with uh chief operations role. You know, the the most important is to choose the right person who can understand uh well understand this technology, yeah, understand uh the changes that it brings, and uh who can you know help people to adapt because it's a lot about uh different kind of uh approach to work. Yeah. You just uh you come to to the office in the morning, you plan your work, yeah, and then you do this work for the entire day. Like if you work alone, if you work in the team. But these days, yeah, uh AI can do this work for you. So you have to do a lot of planning. You should be should be very careful about how you plan things, how you you know uh split your complex work into smaller tasks, run them kind of in parallel, yeah, and how you then uh verify, validate the work that is done by AI. Because AI can do it, but you need to still um validate for the quality. And uh this is the new set of skills that people mostly don't have. This is a rare uh set of skills, and people need to learn it and to understand how to incorporate it into their routines, and it takes time.
SPEAKER_01Yeah, this is an interesting concept because you're right. Like normally the planning mind is at the executive level, right? They're not as involved in the tactical day-to-day type of things, they're thinking more about longer cycles. But now, when any employee has the opportunity to delegate out to the tools, provided that they've got a good plan for like what they want done. A couple of questions for me come out of that. One is you you mentioned that you have to find the right person. What are some of the characteristics that you would say would like if you were advising a company and they said, Hey, out of our existing team, are there any candidates in here who you think would be um good for exploring that chief AI officer role or at least taking the AI leadership within the company? What characteristics would you look for in those individuals?
SPEAKER_00Well, uh, it should be a positive thinking person. Person should like to learn. This person should be easy to talk to other people, to understand their needs and to help them change. You don't have to be technical, really. Yeah? Yeah, but you need just to have some understanding of the technology, where it's going. You need to be you know ready to learn something new every day. You need to learn to experiment. Yeah, you need to take some risk, of course, because people don't want to take much risk, and you have to do it for them, help them to adapt. And uh I think those are the main things. Nothing special. This is a role like many other roles, but you know, it has some taste of uh novelty, that's it.
SPEAKER_01Yeah. So if if uh somebody's listening to this and they say, hey, we want to start to maybe identify some candidates inside of our organization, are there any particular departments? I mean, you mentioned CTO, you mentioned op. So those kind of where you would suggest they start that that search for this type of talent internally?
SPEAKER_00Yeah, I'm thinking about uh places where a product is uh being developed or where service uh uh accomplished. So I think about service and uh development departments, because uh you know there you have to deal with complex uh problems, you need to deal with different people, you need to solve different problems. So uh it's better if you know this person has some managerial experience, project management experience, and uh you know, working with multiple different tools, like to work with uh AI tools, like to work you know with different vendors, and uh you know you need to internally to work with HR, with security departments, yeah, and uh you need you know to get people together and uh solve problems for them. And some problems can be hard because you know they go through the entire uh company from through different roles and you need to get agree with everybody so that you know uh everybody can together use uh this technology safely, they can trust it, yeah, and somebody has to you know to help them in complex situations, like oh my god, I put something into AI, is it okay or not? Yeah, or that um you know AI produces some kind of results. Are they okay? Can I check it, or I should not like trust it completely?
SPEAKER_01That's good advice. I and that the reason I was asking that question is because we get that a lot. Like, who should we promote to this role? And the answer is it it's not a clear cut, there's no like avatar yet defined, I would say that that oh that's the ideal candidate for it.
SPEAKER_00So uh you know, you know what, Chris, I was also thinking that um it's you know in every company there are people who everybody trusts. Yeah, and uh especially when you talk about a new kind of role, yeah, and you endeavor. So something really strange. You need a person who everybody trusts. And uh if you don't have a clear choice, find the guy who everybody you know believes who can uh deliver things, who can uh talk to uh to other people and explain uh them complex concepts.
SPEAKER_01So let me just clarify: is that somebody that would have leadership skills? They would have like uh rapport-building skills. What would you say this are all of the above, essentially?
SPEAKER_00Yeah, somewhere somewhere in in between. Uh definitely some leadership skills are required. Because uh, you know, implementation of VI, it requires uh you know processes, project planning, uh explaining to people, yeah, leading some people, of course you need it.
SPEAKER_01So for a company listening, it's not as easy as going in and saying, hey, anybody want to do this? It's like there really needs to be some intelligent design behind the selection process internally. So now I want to move to a question that it seems like an easy question to answer. What's the ROI on AI, right? But I outside of a simple formula of okay, for this process, it used to take this much time or this many resources. We introduce AI and now it takes less. What's the delta there? Like that's easy to figure out. But those tend to be more focused on individual workflows or roles rather than company wide. And I know that in our initial pre-interview, you mentioned that that like ROI is not necessarily the word that you're using, right? You said that you don't like starting with ROI as that term. You mentioned something along the lines of economic effect. So help me kind of understand what's the difference between ROI and economic effect in the AI work.
SPEAKER_00You know, when people talk about uh ROI, they usually talk about something you know big. And um when we live in the world of AI, uh we you know first need to spend a lot, try a lot, experiment, uh, you know, burn some money for nothing, then for something. And um while we're doing this, we learn. But while we learn, this technology evolves. Yeah. And it evolves so quickly that you know if you try to build some kind of you know big project on based on AI, in six months it can be already outdated because the main providers of this tech will give it out of the box. Or your approach to how you build this uh piece of tech, it will be outdated because uh everybody went the other direction.
SPEAKER_01Yeah.
SPEAKER_00And uh that is why you know many of these big projects on uh you know complex, these bots, these agents, they fail. And uh this is because uh you know technology, this technology is still like forming, it's evolving, and uh for some time it will continue. And that is uh that is why uh you know it's better for me in my situation to avoid these kind of terms because people got used to understand uh uh something, how you know you behave uh about this ROI. But when we talk about economic effect, we uh talk about very simple things. Like uh we want to be sure that what we implement really um changes something in uh our work. Like if we, for example, spend you know uh four hours on something, yeah, we now spend like one hour, yeah, and for example, ten dollars for tokens. Yeah, as simple as that. And then um you start with these simple examples with ordinary people in multiple different departments in your company. You explain that we not just implementing AI for uh for just fun. Yeah, we want you to be more efficient, uh exactly you, not like uh your department and the entire company. We want you to be more efficient. I like that. And you need to get this economic effect by simple things. Like, for example, if okay, you didn't uh achieve like uh economy on uh you know on time, on cost, but you stopped hiring people in your department. Excellent! Yeah, this is also a great achievement. You started to do better service, you know, the quality of um artifacts, of you know, reports, of presentation that you do is higher, it's already an economic effect. And uh when you start to do like this, with these very small things, people start to trust the technology like more and more. And they uh don't uh then think that somebody will fire them because you know somebody will develop such kind of agent that can replace them. No, they understand that if they bring value to the company with this technology, then they have future in the company with this technology. And this way, slowly, step by step, they come to more complicated, more complex um ideas, how they can improve their uh processes, their workflows, then they uh come together into the teams and they figure out how they can improve the work of their teams, of their departments, and then the entire company. And I think uh with such evolving technology, it's better to go you know um bottom up bottom up to let first people learn the technology to achieve results on their uh you know seats, uh on uh on their workflows, and then build slowly uh up. And this is how we are doing. I would say that you know we uh do in both ways because we still have some you know more complicated projects uh with more costs with more risks, but uh there is not a big amount of them. But we have like hundreds of uh very small initiatives uh in the company in different departments, and now we you know we even uh try to select those ones who put those initiatives to production and could estimate the economic effect, and we price them. We have a special Slack group where tell about these guys and explain what they did so that everybody else also you know feel that oh, this is so easy, I can do it too. I think the future. So this is this is how we see it, this is how we do it.
SPEAKER_01I really like the idea of A saying, look, ROI will measure it at some point, but what we're looking for today is economic effect, the impact it's having at the individual level, and also being able to tell those folks the company is investing in you individually. We're not hiring the robot to come in and do your job. So I think that those are some really valuable lessons for the listeners. Let me ask you one more question on this ROI topic. How do you decide whether a use case, if somebody says, hey, I got an idea, I saw something in the Slack channel where somebody was doing something similar, maybe it was an HR operations or whatever. I'm in a different department. I think this would work in my department. How do you decide whether that use case deserves their attention or investment?
SPEAKER_00Well, I built some guides. How would you choose initiatives? Yeah, and uh we use it notebook LM to create videos explaining uh the differences of you know good and uh bad initiatives, and we uh made uh a lot of examples. And uh, you know, uh when we started this approach, I myself came to people and we discussed it what kind of initiatives can be valuable. And uh sometimes you know you talk to the people uh like uh ordinary employees, sometimes you talk to leaders of teams and you explain them, and then they help their people to choose the initiative. So it it works uh different ways. Uh what is important there is that uh you know you don't need to select the most uh efficient, effective ones, initiatives. Yeah, you uh select initiatives that can really promise some economic effect and that are easy to try, to experiment, to learn something on them and to deliver. Yeah, as soon as people deliver and they see even small improvement in their work, they become so much motivated more than you know than usual way. That's uh they can you know next time take something more complex and uh deliver it too. And they you know share this uh among others, and more people are coming. So that's that's the way.
SPEAKER_01So I I like this, but uh I'm starting to notice a theme that how I put it to executives and and you're experiencing the same thing. AI is not an event that occurs in the organization, it's a process. And so if you're an executive or you're you're a team lead and you're listening to this episode and you're ready to get results tomorrow. I mean, at an individual level, people can say, oh, wait a minute, this used to take me four hours, now it takes me an hour. That can happen pretty quickly. But as far as the company becoming AI emergent or whatever the term you want to use is, that's not something that's gonna happen if you bring in a consultant, they spend a day with a team, and now you're all good. This is gonna be exactly like Dennis is talking about. You start small, and maybe where you start isn't necessarily like tied to a strategic target or it's gonna really move the needle, but it's something where the user can get positive reinforcement from using the tool and go, oh wow, this works. They get encouragement, they get validation that this is gonna work, that they can do it. And then as you said, Dennis, they start moving into more, ooh, well, I wonder if it can do this or this or this. So I think that's great advice. One of the things that in the pre-interview that we talked about that I was uh particularly excited to bring to this was this idea of you uh you had recently run a hackathon, and as I recall, it was about 200 people across 33 teams, uh including non-technical business teams. Um a couple of questions there. Why did you do it? How did you plan it, and what happened?
SPEAKER_00Well, why we uh decided to do it, first of all, let's be honest, it's a motivation. First of all, we want to motivate other people. Uh we wanted them to know to that they can build something inside of the company. If you want, you know, to run some kind of cool you know startup initiative, you don't need you know to leave. You just need to go to Hackathon and you can find some other people and to. Together you can build something. Of course, as soon as you build something, you need somebody to review it. And this is why we decided to provide mentors on this hackathon. So we had mentors, some of them were from C level, yeah, some of them were from M level. And we also had technical advisors. So, you know, you assemble a small team, like three maybe five people, you have an idea, then you come to this, you know, uh mentors, guys, and they like try to challenge you. So, what you're doing, is it going to you know to really bring some economic effect? Yeah. Is it uh is it uh very expensive to do? Is this feasible to do? Will they get the data? Yeah, will the clients, the players like it or not? Yeah, and then as soon as they have kind of approval from mentors, they go to technical advisors, and those just help them to choose you know, AI technologies, non-AI technologies, cloud technologies, everything together. Yeah, so uh this year uh this hackathon was uh focused on AI. Yeah, but uh we still you know try to choose the ideas that are better for business of a company of our group of companies. Yeah, and um uh yeah, AI was just uh kind of enabler because thanks to it, you need less uh of engineers in the teams. And we had a number, maybe like half of the teams that had zero engineers in. There were some mixed teams and there were some teams with just engineers. And I would say that you know, uh, among those who won the hackathon, there were different teams.
SPEAKER_01What rules or prompts did you give the teams to keep them focused on real work? Was that where the mentor came in and kind of helped them guide them and say, hey, great idea, but let's steer that more towards production environment in the company? Or what were the guidelines, I guess, for the users or the participants?
SPEAKER_00As you said, uh, we start in Hackathon, we say that okay, guys, this is a preliminary set of ideas that you can follow, or uh you can choose your own ideas. But uh the goal is that they uh do something good for the company, yeah, and uh we will judge um based on uh how much of the economic effect these ideas will provide. So we had I don't remember exact percentage, like probably 40% it was the business, yeah. Uh it was like I think 20% of uh use of AI, 10% of presentation, yeah, and uh 10% for creativity, something like this.
SPEAKER_01Sure.
SPEAKER_00Yeah, so business had uh the biggest impact uh on the final result who will win. So and when you know uh our judges, uh when they put uh you know the marks for the winners, they uh had to follow it, and uh uh then uh of course the ideas who had biggest uh business impact, those ideas won. Not the ideas that had the best AI use.
SPEAKER_01Interesting. Okay, that's a good takeaway there. And just out of curiosity, uh what was the judging panel? What was their like their profile? Was it all technicians?
SPEAKER_00Mostly no no no, mostly C level guys from uh from all of the roles, you know, uh from uh HR, from uh B's uh BizDev, from operations, from uh we had the CTO there, yeah, I was there, but some other roles. Let's I think we have like uh we had like I think 12 people because uh we had 33 teams, so we need to split into separate groups to do it uh quickly. Yeah.
SPEAKER_01And this was something that what did it take place just over one day?
SPEAKER_00Uh it was uh two and a half days. So we started on Friday. Uh and um Saturday guys mostly working on it and day and night. Yeah, and on Sunday uh they uh completed their presentations, and until the end of the day, we just uh did uh a review of them and chose the winners and then had then had a celebration.
SPEAKER_01Actually sounds like a lot of fun. But I really like this idea of taking it from uh-oh, I'm sitting at my my desk or my cubicle or in my office, and I'm like, oh my gosh, I have to learn this stuff, and it becoming it being like a solo experience, that can be a little intimidating. But if you can get your teams, and even if it's you know three to five people sitting at a table, that isolation, the intimidation of isolation with this tool, like kind of goes out the window because it becomes this collaborative experience. And I like from being somebody who's dealt with a lot of change management, I can see a lot of value in the dynamic of a small group or a large group like this, as compared to giving everybody a license and saying, hey, go figure it out by yourself. So this is great.
SPEAKER_00Uh I want to add to what you said. It is important that there is support from uh the management of the company. So, yes, you can assemble one to three teams, give them like a couple of hours, but involve uh you know top managers of your company that they kind of review this work and they give feedback to the teams. This will be a good opportunity, you know, to do some kind of you know team building. Yeah, and uh uh what is also good that the teams will understand better their managers, how they uh you know make decisions, and uh you know it will also impact their work after the hackathon.
SPEAKER_01Yeah, yeah, yeah.
SPEAKER_00Because you know, it will make people closer, understand each other better, how they think, how they make decisions on business, and uh it will impact uh the entire business. So you know, even uh after the hackathon, you don't have that many great ideas implemented, you're still improving your team and collaboration in your company.
SPEAKER_01So anybody who's gone to a chief AI officer training has probably heard me say repetition is the mother of skill. And for me, just like with what Dennis is saying, yeah, that's great if you do the hackathon and everybody comes out with an amazing product. But that's unlikely to happen. But the repetition of going through that process is how they become better the next time or tomorrow they can say, hey, the thing that I learned, yeah, maybe our our idea for the hackathon didn't win or anything like that, but I learned something. I learned how to work with my team, I learned more about AI, I learned more about what my leadership is looking for. So I love those ideas. I got one more question on this. What surprised you most about the non-technical participants? The HR team, the operations team, sales.
SPEAKER_00You know, uh what surprised me is that uh we believed that we didn't assemble such a big uh team of participants. And uh people themselves didn't believe that they can do something. But you know, we did uh a lot of you know uh kind of uh internal PR to get people into the office. And uh as soon as they just started, they started to believe. And uh from you know uh boring and uh confused faces, after just a couple of hours, I saw the entire company in the office uh with excited faces, like uh super happy, like wanted to build something, like moving around, yeah, running, yeah, yelling, uh talking to each other. It was uh great. And uh Hackathon is not just for engineers, it's for everybody.
SPEAKER_01I know that when in our pre-interview you talked about it's not just enough to give people a course and expect them to adopt it, right? That there is a an active effort from the chief AI officer all the way down to run presentations, workshops, uh, private sessions with leaders, department specific proof of concepts and all those sort sort of things. So let me ask you for your opinion on this. What is the difference between AI training and AI adoption?
SPEAKER_00What is the idea of training? During the training, you know, you show some uh set of technologies, you show how to repeat something, but the most important is uh to show that it's not complicated, that it's it's possible, you know, to do it yourself. Yeah. And you know, do it live. Live presentation is the best presentation. And uh it's very important that uh in the world of AI, people can ask questions like smart questions, like stupid questions. Uh they can uh 200 times repeat about can I put my files into AI or not. Yeah it's okay to answer all uh every time. It's okay. We should we should do it. This is uh starting from training, but um, as soon as uh training ends, you need to um encourage people to come back to you, and they should uh bring their ideas what they want to do with this training, how they want to change something in their work. So, usually we end training with uh, you know, we try to give some kind of uh uh tests, but after that uh I uh talk to everybody, yeah, talk back in Slack. Um we have some groups, and uh I can randomly ask how do you guys use this or that? Did you change something in your life? And then I also talk to managers of these people. Okay, so we did the training. Where are the initiatives? What are people trying to do? And uh, of course, you also need uh support from your management system. We have OKRs on AI implementation, yeah. And uh so we connect all the dots that we do the training, we get initiatives, we get to the OKR. And this way uh it's the system, it makes people uh do something, it encourages managers to help people, yeah. And uh uh from the training, we get into real uh initiatives that you know uh automate workflows, improve the quality of artifacts, yeah, or do something like more complicated, like full or full automation of or some kind of piece of uh big piece of work. And uh this is how you do it. You build a system that uh helps its help itself to improve, improve, improve.
SPEAKER_01I can tell you, man, and some of our experiences, like early on when we were doing this, the company was like, Hey, we just we we want the training, we don't necessarily want a chief AI officer, we don't want somebody coming into the company. So just come out and do a tra do some training. And when we would just do the training, they say, Okay, great, we got it from here. We'd check in, you know, a month later, hey, how's it going? And adoption tanked. And we'd ask them, like, like what happened? You guys like you you learned a lot that day, you seem very excited. And without something like what you're talking about, where what do we do after the training to encourage usage and adoption? What happened was we heard this a lot. People were like, Oh, well, it was just easier to do it the old way. But it's not, it's just more familiar, more comfortable to do it the old way. And people without that tempo or that rhythm of of check-in or expectation on, hey, you're supposed to do something with this, they just kind of reverted back to how they were before. Maybe they were using it to for some basic stuff, but they weren't continuing the growth of their AI literacy. So I think that what you're suggesting here has to be baked in for those listeners who are wanting to do this. Like, don't just think that a one-day training or a lunch and learn is going to transform the organization. There has to be more structure behind it.
SPEAKER_00Yes, yes. You should have a management system and uh, you know, some uh maybe even in incentives, maybe not material incentives, but you know, like praising people, like telling about their success, like sharing their experience. So it works all together. And it's okay to start with very simple basic things and uh increase complexity slowly, then it will come. The complexity will come because people want to get rid of their routines in our mind. We don't want to work much.
SPEAKER_01Yeah, yeah, yes, that's true. Especially the stuff we don't like doing, because when we go on site, we we run some exercises that helps the individuals identify the things that hey, it's part of my job, but I don't really like doing that. And that usually for the listener, that for us we found that that's a pretty like low A, they know what those tasks are when you ask them, and B, they're usually nothing too complex. So with what Dennis is talking about, like you know, we don't want to work much, the things that we like doing or that we love doing and that we're good at, when somebody is in that role and they can spend more time doing those things that they love, the reason they would like their job, it doesn't feel like work, right? So that was a really good point. You mentioned this already that uh there are people asking questions about you know, can I put this information into the models and those sorts of things? So what type of, you know, in this policy access control monitoring environment, what should a company have in place before the employees start, you know, they they give them the Chat GPT business license or whatever, or that they start working with company data?
SPEAKER_00The best thing to do is uh to select uh kind of standard tools, explain what should be you know the simple uh the most simple configurations in them, and to write guides, like what is okay to do, what is not okay to do. As easy as that. You know, it's for a small company. And uh you can even start with now with private subscriptions, but then it's better to go into Teams subscription where you have more control, like uh you can very quickly you know switch off the user if something is happening unexpected, yeah. And you have some basic uh analytics on how people work, what they do. But of course, if you you know work in bigger organizations, you need enterprise subscription where you have like most of the control. Uh, you can uh, you know, in case of you have some kind of you know compliance and you need to to very quickly uh uh find the data, delete the data from your systems, like with GDP area in Europe. Uh this is possible to do. Uh so this is where you need enterprise, but you don't need to start with enterprise. I think a good team plan of uh tools like you know uh Anthropic, OpenEI, yeah, uh Gemini, whatever. So they are quite good. Quite good on that. You can easily start with this. What is also important is to set limits for people. People need to learn that uh they can burn their limits for nothing. And uh it's it's quite reasonable to start with what I like. I like to start with uh very small limits, and people will uh very quickly understand that okay, they burned the money out and they need to do something about it. They need to think how they spend their limits.
SPEAKER_01Oh, that's great.
SPEAKER_00And then I uh slowly add more, and more and more and more. Yeah, people come uh many times, and I uh slowly add more. I don't give too much, and this helps you know everybody to learn how to uh efficiently burn money. Yeah, some companies uh you know try to uh give unlimited use for their users and uh let them have bills for uh thousands and uh thousands and thousands of dollars. But you know, yeah, they can you know get some uh success here and there, but this will not scale after this period of um unlimited work. Then you definitely need to get back to limit, but you're not used to it. And this is I think it's a risk for even you know a smaller company. So you need to be very careful how you let your people uh use CA.
SPEAKER_01I like that a lot. You know, because one of the things, if you're a listener and it hasn't come up in conversations around AI yet, it's going to be token spend. If you're not hearing it yet, it is something that you're gonna hear. And it seems like, on the surface, hey, let's give our teams big limits and let them explore and that sort of thing. But I like Dennis's approach better because a couple things. One, if I limit the amount of token or or usage that they've got access to, they become more thrifty. They start to understand, whoa, oh, I can't do that. I have to I, as the human, can't be lazy and just rely exclusively on the model. And Dennis, one of the things that you talked about from early on was the skill of the like the AI-enabled employee, they need to be better at planning rather than doing. And that activity right there with limiting their token usage to me is like, oh, that forces them to create a new behavior when using with AI, which means they have to plan, they have to think through things. And then secondly, it seems like it's a great way for you to start to identify wow, John has come back to me four times in the past couple of weeks saying that he needs more tokens. Like, this is somebody, like maybe he's a candidate for an internal AI enthusiast in that team or something like that, or maybe he's somebody that we want to spend more time if he's gonna do the stuff, let's get him more training. Let's let's take him under the wing of a mentor. So I love that idea.
SPEAKER_00Yeah, yeah, yeah, yeah, yeah. You can you can understand that somebody's doing something wrong. Like, okay, uh, you want me again to raise your limit? How do you work? What do you do? Yeah, and then I understand that you know this guy uses like uh top model, yeah, with uh maximum reasoning for very simple tasks. And uh, okay, did you read the guide? Yeah, did you read the guide? Did you optimize uh how you set all these settings? And uh yes, it helps to solve these issues, yeah. And people uh do much better.
SPEAKER_01How do you encourage accountability in their usage rather than them just I'm gonna use AI, I'm gonna copy this, I'm gonna paste it and ship it, right? Like this idea of that the user being responsible for what AI does. How do you encourage or educate them on what human oversight looks like beyond just oh it looks good, let me write it, let me send this email, let me send this uh board deck.
SPEAKER_00I try to be smart and uh I don't say uh I don't tell people that they just uh you know have to do it because I say that um you know guys, uh you are needed. We can't do it without you. You know, AI cannot uh just live without a human. There should be always a human in the loop. Yeah, because uh AI make mistakes, and uh it can be you know uh not really good at many things, like you know, it can change the facts, yeah, it can forget something. Uh basically it's not a perfect technology that is evolving. So, what can you do? Yeah, uh you can keep yourself in the loop, and it means that you will not lose your job, but uh you can lose your job if you will not learn how to use AI properly. This is why you need to understand that okay, you uh give your work to AI and it executes for you. But you're still responsible for what you are doing. You're responsible for what you put into UI, you're responsible for how uh you plan this work, how you split it into you know uh into smaller tasks, and you're responsible to validate it after the work is done. So basically, you need it everywhere before the work after the work. So you cannot just fully outsource it. Yes, at some point yeah, the technology can come to at such a state where you can uh give uh it more responsibility. Yeah, but you know, but you need to advance yourself. You need to learn how to create you know very good instructions, you need to uh create uh agents that test the agents and validate their works. But still, this technology evolves and uh you cannot find super cool best practices that cover all this. So, in any place where you have you know some uh business critical information, you need a human. And uh that's how I usually explain it. So people are needed, but people who can uh work with AI responsibly.
SPEAKER_01You know, this is an interesting kind of flip on how I guess users in particular, oh, if I start using AI, AI is gonna replace me kind of thing. But the reality is, you know, as a leader who's communicating this to your teams, I would let them know that like by not using AI, you're putting your job at risk. By using AI, exactly you're more yeah. So that's a I haven't heard that perspective before, but it makes a lot of sense because ChatGPT is not gonna replace you. But if you're somebody who's a skeptic or you've got skeptics on the team, it might be time to sit them down and say, like, by not using the tool, that's how you lose your job. By using the tool, that's how you keep your job. So that was a pretty key insight.
SPEAKER_00Let me just add the one point to what you said. And uh we now uh we all you know uh going with AI. We start with uh tools as you know your personal tools. But now AI will become a team player, and uh all efficient team uh teams will use AI together. So, you know, um I see already a lot uh with uh engineers, but we also tried it in uh some other departments where you know everybody uh talked to AI uh to do some job. So you don't uh like you know send information into some other systems. You just write what you want to do into AI, and the AI just you know then passes this information through some systems to the other people. And they uh talked talk uh all of them they talk uh to each other only through AI. And AI here is uh as a you know medium that ensures that information is like structured well, yeah, uh that it it is uh you know proper, it validates it uh against some special rules, and this way you know this information is passing more clear between different roles. Then you just talk to everybody, to anybody, yeah, and uh they just you know have their own understanding than you, yeah, and uh you lose some information when this uh communication goes. So this is kind of you know experimental area, but uh I think the future is uh in such systems, and they make you productive because uh you don't waste time on meetings, you just uh write what uh you want to do, how you want to do it, AI checks it, it improves your um uh conversation, your talk, your information enriches it. Yeah, and then the other side can make faster decisions, can react faster, and the entire process of teamwork speeds up. And I think that's uh that's the future.
SPEAKER_01Well, Dennis, I appreciate the fascinating perspective from somebody who's actually leading AI inside of an organization. That's not we get a lot of AI experts on the show, but we don't regularly talk to people who were responsible for enabling the company, you know, uh holistically about AI. So this has been I I personally have taken a lot of insights from this, so I want to thank you for that. Is there anything that you like I I guess parting advice or any one thing that you would want a listener to walk away with from this episode?
SPEAKER_00Well, I would say that uh this technology is still evolving. Don't rush, yeah, be reasonable, but always see what's going around. And uh this is really new territory. What can happen we don't know. And it's a lot of fun, but it's also an opportunity and a risk. So don't forget that you need uh to really um you know grow this uh competency in your company about yeah, yeah, you need a person or you need a role.
SPEAKER_01Yeah, that's great advice. Uh, folks, anything that we've discussed as uh as far as tools or uh links or anything like that, they will be in the show notes, including Dennis's LinkedIn profile if you want to reach out to him and he have any uh specific questions about how they're doing things at SoftSwiss. So, Dennis, again, thank you so much for taking the time. I know it's after hours for you uh over there, and I really appreciate you uh sitting here and sharing some of these hard-earned insights that you've picked up from being on the frontier of introducing AI into uh organizations. So thank you so much.
SPEAKER_00Thank you so much, Chris. It was a pleasure. Uh I didn't do that many uh interviews uh on you know more like business uh side aspects. It was really interesting.
SPEAKER_01Great. I'm glad we had this opportunity. And thank you so much, everybody, for listening. If this was uh an episode that you found helpful, please forward it to somebody else that you think could benefit from these insights that we're sharing on this and other episodes. And uh if you have a second, we would love to get uh a rating on any of the platforms that you're doing. It just takes a second to share your feedback with, I guess, the world at large so they can see that this is something that um they can actually take something away from. So thanks everybody. Uh, we'll see you on the next episode of Using AI at Work. Thanks for tuning in to Using AI at Work. Don't forget to subscribe for more conversations about how to use AI at work. And a special thank you to our sponsor, Chief AI Officer, for empowering businesses with AI education and training. Visit their website for a free AI readiness assessment and AI strategy guide to help you get started using AI at work. That's www.chiefaiofficer.com. Follow us on Twitter at the handle UsingAI at work and visit www.usingai at work.com for free resources to help you harness AI in your role.