AI in action: Real cases and practical tips - Vladyslav Donchenko | SaaS Founders Podcast

Hello everybody. My name is Dimur Odenko. I'm the founder of SaaS founders, helping founders of software development companies to find newbie clients via LinkedIn. And today we are together with Vladislav Donchenko. Vladislav is the CEO at Cebu it. He is helping startups grow faster with the cost effective adoption of AI capabilities on scale. And today we're going to talk on the future of IT industry, of different useful tools, AI tools that we use in our business processes. Hi Vladislav, thank you for joining our conversation.

Hi, thank you for being here. I really appreciate that we're having this conversation today.

Yeah, awesome. So let's start with the future of IT industry. I see that some people are afraid of AI in the software development. They are a little bit aware. So now they see that AI is not perfect. I cannot generate a perfect code for now, but they're afraid that AI will generate more and more advanced code and will substitute people like software developers in future. What's your point of view on that?

Basically I'm doing the same research all the time. I'm talking with the different people with the different perspective from the different occupations. And the thing I notice that in the majority people saying that AI is the tool and it helps people and really professionals use it at the most potential and really it makes their life easier, their work better and so forth. So on the people who really afraid of it is people who still doesn't use it or afraid to use it or do not know how to start using utilizing it. So from my perspective, AI is just a tool and it is just a set of tools you need to get used to with your daily life. Like previously it was, I don't know, computer and the other step it was Word, Excel, other tools. Right. Which is mandatory. No one asks those type of questions on the interview. But remember back in the day, 20 years ago, it was a normal question. Right. So AI is just another tool and it cannot replace ultimately the human because human cannot formulate or articulate the right business requirements at the end of the day which are can be seamlessly integrated and then implemented. So we're still humans. The alternative is AGI. So if Elon Musk will sue OpenAI for getting closer to the AGI, then probably we still have some time to enjoy this wonderful time helping us as a tool.

Absolutely. I know that you focus your efforts on the startups and you're helping startups to implement more AI solution. Could you give me several examples of how startups and the founders of startups could benefit from AI? Because There are huge ocean of opportunities can work with the ChatGPT, we can work with different other AI tools, whatever. So how AI can help startup founders?

Basically you need to start from the objectives. So AI firstly it's not a silver bullet, it cannot tackle absolutely everything. So we need to understand what are the areas, what are the components of the solution can be covered with AI at the first place. Because if you need very discreet logic, very I'd say the it exact set of outputs with 100% projected quality. Let's say that AI probably not the case. If you want forecast things, if we would like to analyze the things, if you would like to measure things between each other and do some kinds of data capture and data analytics, those type of things, then AI is basically a tool and you absolutely must implement it into your solution. Because automatically if you are not, you are staying behind. Your competitor does that. And it only a matter of time, once he will get the share on the market because he's doing it more effectively, more efficiently and cheaper at the end of the day because, because definitely not developing the model yourself, but utilizing its power and capabilities is a very cheap thing to do.

Okay, everybody say that AI is the future. Can you give more specific examples? For example, the startup founder needs to, let's say test the product market fit, right? Maybe create the MVP first. Yeah, test the product market fit after then to provide some customer support to the existing clients right now. And there are a lot of, let's say a lot of type of work, type of jobs that should be done. And you know, as the startup founder, they're always wearing a lot of hats, right? So here you are selling, here you do the marketing, here you do the software development. Okay, how I can help it, how I can help to solve this? There's a huge amount of the questions that appear at the same time.

You should remember that you still have 24 hours at the end of your day and you cannot do everything. So it is simply impossible to just pursue all the goals at the same time. So you still need to have people you rely on, the trusted partners whom you can talk to with experts in your field and then basically together jointly build up the solution you are trying to achieve. So first, if you get the funded as a seed round, for instance, you already have an idea, you already have a perspective, you already have the people who believe in you. And this is great. The other step is to elaborate what is the most optimal way of implementing it if you are already, I don't know, established organization like huge Marketplace in the U.S. which we helped recently, right. That needed first client support. And we've just implemented the very fast, to be honest, simple solution that tackled the first line, first level of support which helps clients get refunds, get replacements, I don't know, complain on all the things, request for the help and support, give advice on the products, give advice on the marketing stuff and so forth, so on. So this is, this is the thing that quite easy to implement, but it is the main functionality, the side functionality is data. So and being data company is the crucial step because when first line support is just a human, is just a human. So you need to supervise this person, you need to do the trainings, you need to do the regular checkups, you need to motivate the person and so forth, so on. And it is sometimes it's very, very hard to capture the data, how, how it behaves with the clients, what the outcomes of the conversation, where their follow ups was it sorted in right manner and AI at the back end. Basically capture this data, analyze it, provide the insightful information out of it and making sure that you see those issues, you can address them. If this is a mistake of issue or conversation being escalated to the L2 client support or it was handled by the L1, you still have those, those type of things and it is quickly analyzed and basically provide very insightful information that you can use in your marketing, in your product development, in making sure that you are driving your business in the way that it is data driven not by, not by your assumption, but based on values, facts and data we captured. So this is very good thing to have.

Yeah, sure. So as a founders we're responsible for first let's say closing of the clients, so lead generation, marketing, sales. Then we're responsible for delivery process meaning support and let's say fulfilling the sale obligations, etc. And then we need to follow up them and just to understand that everything's working fine, they're happy, happy to stay with us. Right? Exactly. Can I give some examples maybe of let's say of the works that you've done before, let's say from your expertise, how you helped to handle let's say some of that processes and where the usage of AI was the most efficient one in your cases, let's say back for your from your practice.

So you can use it basically almost everywhere. Almost it means. Right, as I said, when you don't have a discrete, discrete logic and very predictable 100% outcome. So the another use case was the voice identification for the KYC process for one of the FinTech clients where they implemented the another layer of security and they had face recognition, they had two FA identification, they had passwords, they had very close perimeter cloud implementation, they would like to do the voice identification at additional optional layer of security. And because right now there are, as you probably know, lots of fraud being made when AI calls, I don't know, your relatives, your friends and saying hey, I need urgently money, I got in the car accident, I didn't have time to talk, something terrible happened with me, can you immediately send me money or something like that. So right now the voice can be also hacked so and really easily hacked. And we tried to implement this solution with the, with the custom model that able to capture the autogenerated voice versus the real human voice. And it was really quite a challenging exercise because in reality when you are working with the live data, not the test data, there are background noises, there are, I don't know, lots of, lots of issues with the connections and etc. Etc. Etc. But during the number of test, test and deployments we did to retrain the model to fulfill this goal, we've achieved very significant results in predictability, almost 100%. It's not 100%, but it is almost and it was deployed to the production and basically put in place.

Oh, that's fantastic. So it means that I heard that search engines and other, let's say platforms, they can distinguish the original text, let's say human written text with the artificial one by using some expression etc. So and now I see that it's the same situation with the voice. So it means that the same situation with the video?

Yeah, with the video. This situation is that to make a very good quality video setting for myself there are tools in the market but to make it at hog and based on your, I don't know, prompt for instance, it takes a lot of compute time and it only a matter of time. One thing model to create really great results for the video will be, will be trained and basically they will be a break breakout even in the video. So you won't be able to distinguish who is att your interview, who is attending your I don't know, candidate interview with who is in front of you. It is AI person or not and you probably saw some cases in the media about it. So yeah, that's true.

Okay, I, I think it's like internal wheels. So some people will try to, let's say to extend the technique to make let's say the, the video better but some people trying working on the same technology but to use it to protect, let's say the identity to identify if it is fail. So no. So it's like the sword and the shield, right? So everybody's defending, so it's like growing.

Exactly. And, and regarding the defending, I have a very, very good thing to know that usually the startups only focusing about product and the features, nobody looks into the security of the solution. And security unfortunately comes with the cost that you need also take into the account. When you are planning your budgets and you're selling, pitching your solution to the clients, it's only a matter of time. Once, I don't know, big booms in the media will appear regarding the AI tool to be hacked already the other vulnerabilities available in the market, it's just a matter of time. While the hackers will exploit and basically produce the the attack vectors on those solutions that are basically mainstream. And you also need to think about that at the minimum we implement those security features which are easy to implement the security by design, which we also always pursue whenever it is possible. Making sure that the solution we deliver is the minimum covers the security essentials. But to do the advanced security coverage, it basically comes with the cost. And sometimes startups fail to get an approval for that or fail to get the budget for it. And from, from my personal perspective I'm being scared because we also work with IoT things and simply we are plugging IoT to the backends which are connected to the, to the ML and AI solutions. So if AGI will be released by OpenAI and Elon Musk will fail, imagine what can happen next. Those IoT stuff will be immediately exposed to the next level AGI. And this is basically a terrible story I would like to avoid. So I'm really helping start up to understand that this is a priority and from my perspective the investors should also take care about it. It is a sound reason to to increase the budget on the security. And from my perspective it is a must have. While your competitors will fail there, you can be a very successful one. Proving the evidences, what you did to do that, what you've implemented, how you measure that and how you ensure that your solution is secured. And we are providing the whole set of evidence to make sure that. We're doing regular pen testing, we are securing the environment, we're securing cloud resources on premise resources, whatever infrastructure you have out there, we are making sure that it follows the best practices.

Yeah, that's good. Okay, I understand that you are a good specialist in that. So if anybody needs let's say a service like that to check up the security and to, let's say to audit the, let's say the platform and the solution they're using. They can ask you. Right?

Exactly.

Yeah, sure. Very nice. Okay, I heard that and I know that startups, founders, they are looking for different solutions to get new clients. What tools do you use, let's say to find new clients and maybe what approaches, let's say give you, let's say some results in that process.

Yeah, let's get back to my answer. I still as a CEO of my company, I have 24 hours in the day. I cannot wear the multiple hats. I need to trust people, I need to rely on people. And with this perspective, I'm heavily reliant on you because I am your clients and I'm really happy clients because I see what you are doing. You're helping me with lots of activities and you are making literally my day easier. I can give you one example. I'm going to the conference next month and I have list of attendants out there and I've identified the target audience I want to talk to, I want to get introduced to, I want to have a conversation regarding their solution. Our possible partnership immediately gave it to your, to your engineers, your specialists. And they already started to get me prepared for this conference. So in I literally today I have a contact request saying that hey, let's book a meeting on 17 July. What's your calendar? Is it busy for that day? No, it's just started, I don't know, a couple of days ago. And right now I have a meeting in my calendar. And what did I do? I just sent a message. I just send a message to you guys. Is this awesome? You need to trust people. You need to work with the reliable partners. And for that case, I would just give a word to you how you do that. Tell me this story.

Oh yeah. Thank you for your trust. I really like working with you, with you and with your team. But I really see that, you know, people are, people were fascinated with the team plates on this specific approach. It worked for a certain period of time that that's even sometimes it works now. But you know, it's easy to copy and paste the templates and people start identify it so it's not genuine no more. It's not a genuine approach. So that's why, that's why we're talking about A.I. a.I. Help us to, let's say to with our, let's say with our approach to make it more customer custom to make it More custom and unique because you know, basically to send the non genuine message is simple, but there is much less reply. But we have developed, that's my part that I would like to share. We developed a AI based tool that researches the profile of potential clients, researches the posts, the recent posts, maybe 1, 2, 3, 4 posts or recent posts and then it adjusts the templates that we provided, let's say the certain logic of the conversation and it adjusts these templates to this specific person. That's why people feel the human touch, they don't understand because it's of course it's verified with the person. So there is a human is watching and feeling if it is a good fit of the message or not. Unfortunately it cannot be done automatically and that's why people review it and if they see if it is good, they send it also manually. Just not to be banned, not to be restricted with the LinkedIn because LinkedIn tracks who is clicking on the button. How often do you do that? So that's why we do it manually. So we. I think that the approach of joining the best possible efforts of the human with the human vision, with the human eyes, with the human's perception and automation, based on AI, on the research will give a really good benefit of that. Because this combination will be much stronger than the person itself. Because the person, it's not. It's impossible to research all these profiles with, let's say with enough depth to get the understanding of that it's from one side and AI, unfortunately yes, it's good. It helps to do, let's say the preliminary work, the preparation work, but the humans should supervise it, to send it properly. So that's why I'm happy that you got this, let's say replies, you got these opportunities to talk to the people. And of course we are also investing heavily in development of AI, especially in our field and lead generation and in our project. It's a top secret, but we're using this tool with the old team of lead generators and that's why you see people feel that it's good, people see that it's, it's helping to connecting and create trust. That's why people engage much more with the other projects. We see that the engagement level, like the number of replies, it doubled and for me it's the best answer. Why should we implement AI in the different platforms? Because. Because it generates the results.

Exactly, exactly. Generates the results. I see them. And also it's great that person do it manually because physically at the end of the day he Studies how AI answered the question so it benefit to his or her progression as a salesperson. So AI really powerful tool and a person behind it really gets elevated in his career with the AI. So yeah, it's a win win scenario. I like that. And this is the type of, type of business cases I propose to my clients. AI should not replace everything, it should contribute and help you do your business effectively. Yeah, yeah.

So to summarize, we could say that not the person and not the AI only can be the best solution. But only let's say the joint efforts of the people with the human vision, human behavior, human perception and capabilities of scaling it with AI can give a real benefit. Yeah, I think it's very clear. And the more we engage with AI technology, we see the strong parts of the technology, strong sides and the weak sides. We need to compensate with the people. But generally the results are really awesome. I love, let's say, let's to see the statistics and compare it with the manual work, solely manual work. I'm fascinated with that and happy to see that. Okay Vladislav, thank you. It was really interesting for me to hear your opinion on that as a practicing, a founder of a company helping startups to implement AI in their field. And yeah it proves might some, let's say my experience and things that I see. So I would be happy to communicate with you also after about that. And if anybody needs some advice in terms of the AI implementation, please contact Ladislav. I think he can give you the very specific tailor made advice how to benefit out of the implementation of AI. Thank you very much.

And at the same time, if someone needs a really professional and qualified sales service. So do not hesitate to reach out to Dima because I'm his customer and I really enjoyed the process of working with him and his company helped me a lot to, to really elevate my business to bring it forward, to make it more robust and competitive from the sales perspective. So you should, you still have one head and you need to choose wisely with whom you're working and it's great to have people, talents and companies like you on the market. Thank you for that.

Thank you, thank you. It's awesome. Thank you. Hope to see you soon.

Have a good day.

Bye bye.