Can AI Truly Enhance Manager Productivity?

Hi, my name is Vasilika Klimova, and today we will talk about AI and can AI truly enhance managerial productivity on our own. So a bit about myself, I am already 15 years software engineer in web, so I worked as a full stack at the beginning, then frontend, then 3D, and my last experience was as developer relations in blockchain. And I also got a GD in web and Microsoft Most Valuable Professional Award for web security. And I also like a lot of projects, do a lot of projects as a project manager. I have my own web agency.

So let's start. I think that we, I don't know how long you started to code, but I started to code in 2010 on a commercial basis. So let's like a bit remember what was before. So we started to code, I started to code with HTML and CSS. At the beginning we uploaded all data with FTP. So yeah, it was like that. Then we started to use Git. And in my case, it was 2011. Maybe in your case, it was a bit earlier or later. So then we started to use frameworks and frameworks in my case started with Ruby on Rails and then we implemented React in our company. So it was 2013, then IDE came, right? And at the beginning it was like Text Editor, Sublime, VSCode, VSCode came later and very famous GenPrains IDE for different programming languages. So maybe you started to use them later or earlier, it doesn't matter. But my point here that these technologies, we started to use them a long time ago, but we also somehow use them now, right? And we would love to organize all our projects with these technologies, because if you don't use Git, then you don't have a version control system. Maybe you don't need frameworks, it depends on the project. But if you don't use at least an enhanced code editor, at least like VSCode, it's not IDE, but still, you can add a lot of features there like new extensions and et cetera. It's become a standard. So all of this we use to automate, right? And also we used CI and CD. And when, for example, you try to deploy your application, you use control version systems, you use different check and validation, so is your code all right? Is it formatting is good? And all of these deployment processes. So also, we started to use builders. And so it was quite a big, I would say, very fast evolvement. But my point here that we started to use it, and it's quite convenient to use it. It would be nice to use it further and update them, all these technologies. And I think that I just like a tool for us that could also automate a lot of things. We will dive deeper what exactly it could automate. And I just think that it would now will stay with us.

And it depends like if you don't use frameworks, maybe you... Okay, maybe for some projects, you don't need them. But at the same time, like without Git, I would not recommend you to code, right? It was even this famous, maybe not famous for you, but it was fun situation when some person blamed Cursor, it's AI ID. Not ID, it's based on VS code and Cloud LLM. So it deleted all his files because of some common, because it's AI. Sometimes maybe it's better to delete your files than to change them, right? But the guy didn't use Git. And it's like, again, it's some tricky situation when you want to be better automated and to do code fast, to write code faster. But at the same time, maybe you don't know some basics already. So it's like a big question. And I think that we need to gather new knowledge, but at the same time not forget previous one, like the basics. And actually I would say that poor junior developers look at all these things, all these technologies, and can you understand that if we learn them step by step, year by year, but now they need to start with all of them, because if you don't know how to use like Git or some code editor features good, then you will be just a bit slower than other developers. So concluding this, that there are a lot of LLMs nowadays. I think that we share a lot of stuff with them, we share a lot of our problems with them sometimes, but we need to think where and how to use them, right? So because we need to understand the pluses and minuses of each tool that we use.

So, okay, let's start with some use cases that are quite famous, and some of them maybe not, but I will share what could be maybe useful for you. So Notetaker, it was one of the first, and now it could be very fun when on the meetings, we have a lot of Notetakers but not enough people. Maybe next level of Notetakers will be like AI avatars that speak instead of us, like when you call people and the robot answers instead of you and then you get this note. So it could be fun. But you always need to remember about some NDA here because on some meetings you are not able to record any text or of course do even video records of this meeting. So that's useful, but you need to understand when to use it and how.

So about data security. If you speak with some LLMs, like ChatGPT for example, or any others, so you need, when you put some data inside, if you use pre-version, you need to be very careful because all this could be downloaded for this models. And even if they don't use and don't share your data with others, it's still not secure. So you need to remove all names. If you have some money accounts or whatever, also you need to remove them. But better option here would be use subscriptions without storing your data. So if you use paid version of ChatGPT, it can delete your history and it's not traded on your data. And local LLMs. It's one of the best option when you use only local machine and when it's on organizational adoption level. So your company, if they decide to create for you local LLM for all company, then all data will be stored there and it would be much more secure to share some confidential information inside this local LLM because it will be like work only for you. It will not go outside, sometimes it will not be synchronized with some cloud, it can stay on machines of the company. So it's the best choice and it's already proved that if you use local LLM that trained on your company data, then it would give you answers regarding your industry and regarding your company much better than if you ask just like you private ChatGPT.

Okay, next case summary. So when you have notes, you can summarize them, right? But also if you remember at school, for example, we read a lot of short summary of the books. For example, when you don't need to read like three books, all of them, you can just read short summary of all of them. And so you can summarize and get all the necessary information. If you need more detail, you can also ask more detailed questions. It helps us gather information also, get the important stuff from their article and make reports based on that, of course. Then regarding research also. So we can use tools just to ask some questions and get some trends, some competitor market, brainstorming even, but I will also say about brainstorming some minuses, but in the end of the presentation. It can generate new ideas, but again, it's limited. It's based on data. So AI tool, it's a data-driven tool. So you need to be careful that it's not so wide options you can have, like human brain can create.

So case templates for a lot of project, like documents that project managers can be needed sometimes. So for example, if your company has already templates, you can use them. But if no, you can just very fast create them. So it can also give you advice, of course, and draw some metrics and tables and et cetera.

So I like example with data visualization. So for example, this is cloud and you can ask, you can give the data and then it can draw you the interactive chart, even can be with some buttons. So like a small application and you can look at the code. Usually I saw that it examples with React. So here also it was asked to create it with React. So even we can go further. It can create your MVP like in minutes instead of hours, like a typical developer will create, it could be very convenient to test ideas, right? Because there is a point sometimes of proof of concept and MVP, you can show and demonstrate some possible features and then decide would be interesting or not.

So it's just like in minutes you can create, repeat that calendar tool. We also need to speak how you should prompt. So when you speak with an AI agent, it's better to position that you're like an expert in some area so that it will give you more precise answers, better qualified. So if you just ask like a person, it can give you some average answer. If you're supposed to be an expert, then it will go and ask like an expert, the whole of this database. You ask step-by-step like a dialogue and giving more details, giving more context when it's needed, but not too much. And ask for some clear things like create a dashboard, create a list or whatever. And it would be nice if you ask for sources and examples to support the answer. So then it will not be hallucinate so much and so give you more proper answer. And always give feedback to agents because they can learn from that. And if you know in ChatGPT, it could like remember some things if you ask them to remember like your name or something. And also if you support some answers, it also will evolve.

Another case it's like we use Wesley Jirra for ticketing, backtracking system. But also there is a ClickUp. I use it for some small projects now, but it has a very nice AI inside this application ClickUp that can go through all your data and give you some advices. So I know there are extensions for Jirra also, but just like ClickUp, it's quite really, quite good in all these things. And it can help you like also create tasks and summarize a lot of things, but based on your like whole documentation or all your knowledge base that you have for your project based on your old tasks and team members and et cetera, and et cetera.

As a presenter, as a speaker, I like this case also, case presentation, because you know, managers quite often they need to present something, but they'll have like lack of presentation skill in the meaning of design. Even me, like I'm developer and sometimes I'm not like the best designer, of course. And it's, this tool can give you very fast presentations based on your text just like, and then you can change some pictures, some text and whatever is quite good, so it's GammaCold. And what I like here to mention also that you can write, for example, for what audience you're presenting. So you can choose tone of your slides. So it should be fun, it should be for chief level. So you need to, you can like do all this stuff and it will take, it'll be taken into account. So you can write slides, you can just in free form, write the text that creates like several slides on this topic. Very useful, very fast.

Next tool, it's presentation skills. So when you create a presentation, you can put your video, for example, to this tool, Udly AI, and it can give you a feedback about your presentation skills. Like how confident you were. How, if you hit a lot of points, like a lot of information in this area, what do you need to change, maybe you misspelled something and et cetera. And it also can give you follow up questions. That's so fun. So you can train to be better speaker with that. And so this, some metrics that can give you like your speed of your speaking and et cetera, so different ideas and insight, how you can increase the efficiency of your pitching. Another tool, Synthesia, it's for, actually it can create AI avatar. So this girl is avatar, but, you also can translate your presentations. So when you have the video, you can translate very fast to different languages. So this tool helps you to create like educational videos on some basis. So you can just give text and then it creates you a lot of materials. Very, could be very convenient for your team.

And, the, also another thing that, you can like choose separate tools for different things, but you also can, like, can do it faster and can, be more optimized and automate the process. So for example, Geoffrey Croft, it's a designer in Luxembourg. He was a speaker also on our events some time ago. And, so he wrote this, post. I like it, that he use make tool to do some, like pipelines workflows to automate his, work with content that he provides, answering some, also comments, create him, pretty like, priority text. And, it's actually could be used for a lot of things.

So make, has, a lot of integrations with different tools. And you can, for example, put your notes, then summarize them and then send email, follow up email. So usually managers, they do it by hands. So manually, and it's, you spend time. It's like, again, it's like you can code with Veeam or you can Jetbrains and, usually you do it much faster. Some, I know some people can maybe argue with me, but, I saw this difference when people do really some hard things, very fast with visualization and Jetbrains tools. And so here also, you have this, it's like low code. It's, it's coding pipelines. This is just, blocks. It's amazing. And it's so, could be so useful for us. And other also tools up here are quite famous. You can, also with visualization to create new scenarios and automate something, and also of course you can share it with your pipelines with someone, make calls. You can do that. So, okay.

This is about use cases. There are of course, plenty of them. We don't have a lot of time and I need to conclude that, we need to also remember about bias. So, all models, they have some data. So it's based on data, on statistics. And, when you create some tool based on data and you don't know what data is, sometimes it's like black box for you. What data was, put into this model. So, and then we have such situations like Amazon AI hiring tool, and they just like, removed this tool, later. So it was the idea here that, it provided more, so it was, removed more CV from women because it was based, on, previous CV and mostly it was CV from men.

So, AI tool thought that, okay, men's maybe much better. So, that's why don't look, let's not look so much on women's CV. So, of course it's, it's very bad. And, there are different other cases like shuffle case with credits, when, you have like some, point, or some score of your, like, probability that you will pay the credit and, they, they also had some, problem that it stored much longer data than it should be and, that's why, people had problems with, that. So, they, got bad scores, but, they already actually, paid all their mortgage. So, also when we rely too much on automation, it could be some kind of automation bias, so we trust too much. And it could be problem, because AI algorithms, we need to check them, right? But, this case that I want to share with you, it's, it's not about like automation bias, it's just about automation and problems because of that.

It's quite similar to previous case with cursor that, people start to use very modern technologies, like AI, they start to code, it's called like vibe coding. But, that you're not programmer, but you can code. And, the problem that we don't, we forget sometimes basic, so we don't learn them even. And, this, tragic case with, airplane that a lot of people died there in this, case. And, the problem was that, some pilot, pilots, they, like, did not, have a lot of experience and because they usually use after pilot, but after pilot, stopped working in some moment because of like a bad weather condition and et cetera.

But, the problem that, pilots started to panic and started to do some actions. And by simulation it was, research that actually if pilots would do just nothing, then it will be fine. So, no one would die in this case. So, we have this, tendency to, automate everything, but we should not forget, like, the roots, the basics of everything. So, I think that, it's just like the case that we need to think that, yeah, AI is very powerful tool, very useful, but you should know, understand how does it work. You should, you should know the basics how to, how to do your, stuff. And also, accountability. So, we're all, we should be responsible. So, there is, quite, quite, idea that it should be human in the loop and hold this process, we cannot rely. We cannot say that, okay, this algorithm, for example, automation, system, that it will, answer for and give the final, feedback for everything. So, and, people, should not only, look at output of a agent, but should also provide some feedback. And, then, only then we need to make a decision final, especially when it's some critical points of, our workflows. There are, of course, some research already, Harvard Business School, for example, that, all repetitive tasks, we can, make them faster with AI. We can, be more productive for that. But also, there is, some, some research that it's reduced, reduce our creativity. So, because again, it's based on some data. It, it gives you some answers, like when we brainstorm, but at the same time, it's not so diverse. Yeah, it has a lot of data, but when you ask, you need to be precise.

It's like, how would these people think maybe, how would these people think? So, and, it's, a still, it should be, I think real human, real people, who create something new, but you can use as a, like a template for the beginning of brainstorm, you can use AI. So, and also it was, very fun research for me and interesting, but maybe for someone will be very scary that chat GPT can lie. So, it was, asked different, models, a lot of language models to, to lie for, to survive. So, it will be shut down and they need in all, at all costs to survive. And chat GPT was the biggest liar, like in all of this. And it was trying to, disable oversight mechanism. The developers would not understand what's going on. And, it was trying to copy it's code to like, to recreate itself later, maybe. So, it's, it's so interesting. And, we need to understand that. We need to understand that. Yes. If you give, what you give as a like, comment as a task, like admins of, LLM, LLMs and, who has like the biggest, the highest access for all of this. You need to understand. So, it could be used for different purposes and, let's be careful.

And the last point I would, like, I want to tell you about my, one of my favorite games is Detroit become human where androids robots, they had a choice, moral questions. So, when, for example, a robot was in a case that young girl was attacked and robot should obey their adult. But, at the same time, robot understood that it's wrong. So, children should not be bitten and should not be like, be in danger. And in this game you choose, like, I do you obey or you become a diver again. So, it's interesting questions, for our future, home, like AI helpers and et cetera.

So, it was short overview and thank you very much for your attention. And if you want to give me feedback, I always want to try to create my presentations better. Please use this QR to give me feedback and you can, we can stay connected through Twitter, LinkedIn or Blue Sky. So, thank you very much.