From Digital to AI Transformation - with Gennadiy Vorobyov from Netpeak

This talk focuses on AI transformation and its applications across various sectors such as marketing, project management, and recruitment. The speaker shares real-world examples of leveraging generative AI tools like Adobe Firefly and Adobe Express for marketing materials and creative content. They also discuss using AI for summarizing meetings through tools like TurboScribe, illustrating how AI can enhance productivity and facilitate better work-life balance. Furthermore, the speaker introduces AI's role in recruitment processes, including analyzing CVs using a GPT agent to assess candidate compatibility with job titles. The presentation emphasizes the need for effective AI integration in business practices and the importance of educating teams about AI technologies.

Good morning. Uh let's check my clicker. Yeah. From digital to a transformation. Last maybe 10 years we heard a lot about digital transformation. Digital transformation blah blah blah and this and last two or three years we progressively listen and do and do whatever we can about AI transformation. what kind of AI transformation I going to uh share with you. It's all about some of examples that we

use in our work. It's like a prompt, it's like a tools, it's like a techniques how you can manage, how you can be more effective in your work, how you can uh uh take care of your work time and to have a real work and life balance. Yeah. So, uh we're going to talk about design and marketing using generative AI, AI in project management and everyday routine,

performance digital marketing, recruitment. Uh I going to share how my child that is six year old also uh do job with uh AI uh and some business analysis at that we also use. So I think yeah uh first of uh uh first of slides is going to be about AI in marketing in in design especially uh thanks to my colleague Mos that actually share this information internally

a team how she actually use AI to make some uh images and video creatives for our internal marketing for marketing of Netpic of digital marketing agency as ours uh first Adob Firefly how we going to use it and actually we should know how it works. That's why we use and we ask uh for example Deli about reference with prompt photo of an old man in style of

Lee Jeff. On the left you see the original reference to Lee Jeff uh pictures make it really nice but Firefly uh have some hallucination but probably it's about of poor of database that it learned from. After that we're going to switch to do express and this is really game changer when you want to make your own creatives when you use uh some deli and for example midjourney

it can generate some images with no sense with uh text that you don't type in your prompt. So using Adobe Express you can use it and you can make a lot of uh creatives with exactly text that you want be shown on the picture and after that you can correct it you can edit it and here you can see some uh prompt like a banner make me

a banner for tourist agency that offers minus 30% on Egypt holidays and this is all made by Adobe Express this is the prompt and this is the text that actually you can use you can edit and you could edit you can add your brand book images and etc. Next uh example from the same Adob Express results. Uh here is a test how we can build our own

brand from scratch. So the prompt is uh make me a brand generate me a new brand for bakery. So we see here we can see some creatives with new brand name half of and with nice pictures about that uh what we do in project management it's some tricky things and it's really some techniques that we use our project managers in their works uh how we can manage

our work routine and uh the first uh example is about how we going to summarize our meets when we make meet in online in zoom uh Google meet etc. So after that we should uh make an action plan. We should make a followup how we can automate that how we can use AI for this work. First we should record the meeting and make a voice memo. After

that especially for Bulgarian language special for Bulgarian language we can make uh very very nice text to speechtoext recognition using Turboscribe AI. They have a really nice free uh free trial with uh maximum 30 minutes video. So you can put the video or some voice memo and you can get a text transcription of your of your meeting. How I do in my in my work or in

my life when I make a meet with someone I just put a record on my phone. Of course the person uh should agree with that. After that I just put it in tubascribe and make a very quick summary how it's work in the steps. Uh if we use online meetings we can use Google workspace uh or some browsing browser toolbars. In offline meetings we can use voice

members on our smartphone. So we first we recorded the meet after that we do speech to text. In our example in our uh in for Bulgarian language we use turboscribe. We should put video text uh audio to text file video audio to make a uh subtitles. We make we can make a speech a speaker recognition and also we can work with a blo file with uh YouTube

files, YouTube links and etc etc. After that we going to use this like this how true interface looks like. So this is a list of our files that we going to transcribe. After that we make a prompting name the voices. If we are two speakers we can we can name speaker one is me, speaker two is Nikolai and etc. So after that we going to have a

really customized uh summarizing. Let me show how it works. So we have a webinar for Tik Tok cuts in Bulgaria last year. So here is a prompt. We are digital marketing agency. The speaker was me, Nikolai and then the next Nikolai and we put the document. We put the document with a transcription of the whole webinar and we put the presentation that we used uh during the

webinar and here down we see the summarizing of this online webinar. So the date uh who is the organizer and all the summary is here. It's made by in in one or two minutes. Yeah. So we have a webinar about 1 hour and a half. In 2 minutes we have whole summary of this Uh so how we going to use also AI in project manager? It's my

tool that I use last two months. Uh I use superhuman uh app. It's really effective app when you have a really big email routine. How many of you have about 50 to 100 emails per day? Yeah. Yeah. And every of you should deal with that. You should answer, you should sort it, you should filter it. Yeah. Of course you can say you can use Gmail filters, you

can use another filters, but these things really is game changer when you have a when you uh zero inbox paradigma when you use it. So this is a really nice tool. just try it and uh now it's cost about 30 or $40 per month but it's really give me additional four hours in a week I don't know what is that okay so the basics prompt about grammatical

errors that I think everyone already can use and everyone can make uh so some another examples with uh content creation it's uh obviously of course in in digital marketing we actually can use AI in all spheres of producement so let's skip that uh another thing that I want to share with you is uh if you have some scare to use AI Okay, don't be scared. Uh opposite

you should uh talk to him to to it to him to her however uh and to learn it because as you know it can make mistakes it can make some hallucination and etc. But let's see uh how I learn my child uh to help AI to be better. So my daughter, she's six years old named Darina and she want to draw a traffic light please. Yeah. She

make a voice uh voice memo from from phone and she asked like make me a traffic light and this is a result of uh traffic light. Yeah, we see some something not okay with this traffic light, right? And Darina asked me, "Oh, father, what what we going to do with that?" I said, "Okay, explain what is wrong and maybe you can ask uh charg uh the next

question was can you draw me please the same without two yellows? One yellow please." The response is here again is something wrong. For example, yellow is not is not on. Next try. Next try was why yellow is not showing. Could you draw me please? Red, yellow, green, please. And again the same Yeah. From again too yellow. Next try. But it's incorrect image of traffic light. Do you

see what's wrong? She asked the judg. Okay. Yeah, I see that is wrong. This is a more accurate version for you. It's Yeah, it's really more accurate according to JPT. Okay. So, but she's fighting with it. Oops. Technical error. So, uh what was the next uh now? Ah, okay. It's my slide. Yeah. Okay. You didn't fix it. your mistake. Yeah. Chip declare. Yeah, it's my mistake. Of

course. What we going to do next? And the next was, could you draw me please? The same, not the same, but one green, one yellow, please, etc., etc. Again, it was I think it was the final uh yeah, it was the final result of charg. So, uh the lesson is here is that yes, charg model can make mistake, can make hallucinations. So we should uh take care

of it and this is the signal of course that we should uh control it. What we do in uh recruitment also uh for example if we have a lot of uh CVS to analyze in in our HR processes how we going to deal with that? Uh we going to build uh we just built one GPT agent. and we just send and we just send the hyperlink to

our job title that we're looking for and checking after that how charge GPT understand this uh job title. If it's correct, we're going to upload the CV of the candidate and for what for to take a resume to take a summary and uh some points how does this CV corresponds to job title and the score after that we can make a score. It's not seen here but

for example for this CV corresponds to the job title up to 85% and after that we can make a score. Let's see. Let's look for CVS with only up to from 70% and up. So this is some filter that we can make automatically and of course it's not going to be a decision. You can you can check also another CVS. It's all our decision and all our

function. You can use uh this GPT if you want also for to analyze uh CVS by such way. uh another things that we made in our company in our group of companies uh because as part of Netpic group we have more than 1,500 people in the team and of course uh the HR processes are vital for us. So what we're going to do we prepared like a

digital portrait of every job title we are looking for. For example, if we are looking for seale managers or like managers or team leads, this is some typical cycological uh portrait that we should look should look for. So we going to use some uh typologies inology linguistics MBTI big fives with more than 100 criterias for MBTI. We looking for introverts, for sensoric, for logisticians, for rational guys.

In the big five, we are looking for characters like naroticism, openness, uh extraversion, agribliness and etc. So after that, we make a portrait. So our guys on this job title should have this uh this and these types of in character. Okay, what we going to do then? We have a matrix for for this job title. We have this matrix. So we have some numbers some uh zero

and ones. And after that we can make like a digital portrait of a job title and the candidate that should have it. What we going to do after that? Again let me explain you. So, so, so after that we make an interview, video interview, we record it of course and after that we again make the such summary that I mentioned in project management. We make speech to

text and analyze the text that people talk according to during their interview. And of course the recruiter see the two numbers for project manager like for project for manager title this candidate corresponds on 53%. But if we going to talk about uh ethonomic ethonomical candidate it corresponds for 74. So this candidate more individual player not like a manager. So we should look at him in for this

for this job job title. Um okay last four minutes 3 minutes and I want to uh share you some another interesting research from MIT uh it called uh it's discovering how minds and brains create language and the interesting things here that scientists from MIT analyzed that we can use AI in neuroscience There are some type of words. There are there are some types of sentences semantic structure

of it that activate our brain zones when we listen to some text in radio in TV in uh ads when we uh read it or when we hear it. So this research was very interesting and I decided why we don't use such such research to make art creatives for our arts for our text arts for our audio arts video ads radio ads and etc because we can

use phrases and creatives that activate brain zones uh that researchers from MIT mentioned how we do that uh actually oops uh semantic angularity. So we use uh unused uh phrases uh some grammatical difficults uh some uh syntaxes another ways and another AI models and our brains how we can use it for example I mentioned that I going to you I going to be today at balconic commerce

I just mentioned today I'm on the conference balconic commerce make me some uh creatives for billboards that you can make first was uh shop like uh thunder and sell like your dream. Another was sorry Bulgarian online offline and different different phrases that actually going to activate and stop your attention on this phrase on this art creatives. If you want to use it or try it, just uh

call me and I going to share with this GPT that we use uh just to try in your art creatives to to to to test does it really activate your brain zones. Another things that also we use in our business phone system called ring start actually we analyze all the calls that are made in your telephone system. After that we all again we make text recognition, voice

recognition, call recognition and we score all the phone call with some score from zero to to 100. And for example those those calls that with a score 90 and up should be controlled should be checked by the manager by supervisor because by statistics from all calls 10 to 15% of calls needs an attention from uh supervisor not all of the calls and this is really game changer

because you don't have to control you don't have to listen all the calls uh with a with a goal to improve improve effectiveness of the uh speech to text as I mentioned and call summary again using AI it's uh it was wow before two years and now it's just common task and common tool that we use already now today uh and last for me from me for

Bulgarian audience if you want to make this transition if you want to make this transformation we have made a free course AI course uh on Bulgarian uh from zero if you don't understand anything if you don't use AI at all to up to expert level uh six models so you can use it you can uh just scan this uh QR code or just remember the website ksova.ai

AI just on only on Bulgarian for currently and uh this is the way how you can use how you can make transformation transformation AI transformation for you and for your colleagues for all your company for your clients for your partners because this is a really way uh how we going to proceed with all challenges uh in our modern uh life. Thank you very much for your attention.

Uh it was my pleasure. Thank you.