 AI became one of the hottest topic right now, and everybody want to break into AI field, become a product manager or engineer son. However, there's a very little training out there for everyone to effectively break into AI product management. Today, we're going to demystify AI product management by head of AI from Save the Children International Nicholas. He's going to dive into five different factors tell you what the most important AI tools how to become an AI product manager was important AI courses. People can start to join for free starting from today. Wait until the end of this video where we're going to share with you the generative AI use cases everyone must master before they break into AI. Hey guys, this is Dr Nancy Lee, a direct product and future informs about how 100 people plan their dream PM job offer in fan companies and unicorn startup and continue to get promoted as a product leader. In this channel, we cover tech trends and free product management training, like subscribe and check out new videos every Tuesday. Hi, Nicholas, welcome back to the channel again. How are you doing? Very well. Thanks. And you? Awesome. Yeah, I'm doing very well. And this specific video, we're going to talk about you as the head of AI in Save the Children International from your real life experience, how to help other also become an AI product manager. And also, let me please allow me to do a brief introduction who is Nicholas. And actually, Nicholas is the head of generative AI at Save the Children International, which operates in more than 100 different countries globally. And Nicholas itself is also Brazilian who moved to UK with visa sponsorship from his new company and very exciting global career. And actually in our last video right here, he died into specifically how he become the head of AI in the international company as a Brazilian citizen. And everyone should check it out right there with all different kind of career growth strategies. And in this video, we're going to dive deeper regarding AI product management, lots of technical terms and also lots of free courses recommendation and different kind of use cases people must master from the head of AI. Awesome. So Nicholas, why don't we do this? Why don't you give people an overview regarding what is AI product management? So in simple terms, the AI product manager is someone that knows the PM tools, the product management tools, but is prepared as well for the new challenges that arise from working with AI. So someone that doesn't want to use AI for everything, but someone that tries to avoid AI as much as possible, do it with complexity and drive real impact and move real pointers. This is awesome. I love your quote, use AI as much as possible, but avoid AI as much as possible when it's too complex. This is awesome. Let's dive deeper a little bit. So can you give us some examples regarding how AI is used in product management today? Given all companies put AI next to the name, like Notion AI and kind of AI, and then they become the AI company and raising 10 times more market value because they add AI to the company. So tell us more, how actually AI is used in product management today? So first of all, I see three levels of implementation of AI in a product. So the first one is simpler is to use AI to improve an existing feature. The second one is to create a whole new feature that wouldn't be possible without AI. And the third level is to build a whole new product that would be imaginable without AI. So the AI PMs really shine in the third level. For the first level or second, you maybe just need a new developer and off the shelf tools. So why don't we dive deeper regarding each level? Can you give us one example of each? Let's say the first one is you add existing product, improving existing features, you add AI. So give us one product that's in your mind right now. In the first level, you can improve features like Gmail does detecting spam messages and filtering out for you. So it's a way to improve functionality that already exists. You could have a set of rules that detect if the email spam are not, but you can use AI and make it better. So it's easy to implement the risks are very low, the efforts low. And but it's easy for the competition to catch up. In the second layer, you can think of, for instance, AI generated content for marketing, something that is happening right now. So tools, commercial tools are already that launching some features that help you create new content out of description that you will provide. Awesome. Such as the email marketing tools or social media posts. Some of the social media posts were created by AI and some of my social media posts were created by AI and LinkedIn posts, different things. Hey guys, if you can guess which one created by AI comment in the description of my social media posts. I love to see if you can tell or not. Yes, those are new features for email marketing tools. What about the third one? On the third level, on the third level, you can think of products that could be completely new. For instance, self-driving cars without a driver seat. So you just go there, sit and go to whatever you want. It's not a car that was adapted using AI, but it's a new feature that you get a new whole new product that allows that is selling you a new value but would be this new value in these cases that you are able to sleep while you travel. This could be an example. This is perfect. You said true AI, PM, are mainly in need for the third type, not the first two, am I right? Yes, yes, because in the third type, you need to, the risks are much higher, effort is higher. So an AI, PM must be able to understand how the data strategy, the AI strategy and the product strategy align to create a sustainable advantage for the company. I think this is where the AI, PM truly shines. So you need to be sure that you're building something that as you get more data, your product gets better. And as your product gets better, you'll get more customers and you'll get more data and with better quality. So you need to build these wheels, these snowball effects in the product. Yeah, we call those the flow, like flight wheels and turning to snowball effect and you get more users, more data, better better data to train the models and more users. It's a loop and the wheels just get bigger and bigger. Thank you for experiencing this, making it much simple for people to understand. So now, tell me, understand, now we know three types of different kind of AI being used in product management. What are the top three skills to master for AI product managers, given lots of people want to break into AI PM. So what are the top three skills that's necessary? Okay, so the top three skills, first of all, and we almost never heard here about that. This is AI UX. It's very important to understand how your final user is going to use AI, what level of explainability is necessary for them to trigger the right decisions so that your user is able to receive more value, to draw more value from your product or your feature. So the AI UX is extremely important. I had a work together with the International Partnership on AI that was related with understanding how people that don't have any AI literacy deal with AI models. So is this safe for them to, to receive the predictions directly? It depends a lot on what their role is, what are the tools they have at their disposal in the moment they are using your product and the legal factors as well. For instance, if you send a detection of a disease in XM to a technician, legally, they can do do anything too much elaborated. So maybe it's not necessary for them to destroy the solution. So first skill is to get more in touch with AI UX and within that you'll see the concept of explainability. That's very important. Second, know when it's better to use off the shelf, so AI solutions built to customize the AI solution in-house or hire a company or even wait for a big tech to launch it. So I surfaced the first AI machine learning wave, which was the AI for images. And I felt a lot in the trap of developing things in house. And then six months later, a big tech delivers the same feature. And then two more months, their solution is better than ours. And we had to migrate for theirs. So a key rule, a good rule of thumb in this matter is the functionality is something that's general that can be applicable for a lot of companies, regardless of their domain, probably a big tech will launch the solution. Or if you build the solution, they will buy your company. That's that's a good strategy as well, if you want. But if you have something that's very specific, for instance, in the case of Save the Children International works specifically with children. And solutions that are specific for this public in specific locations are solutions that probably a big tech will never launch. So it's worth to to build in house or try to to hire a solution if they it's ready array. A third top skill for an APM is you don't should spend too much time learning algorithms. I think it's more important for you to learn and what are the tools you have at our disposal, and how the pricing works and how it gets its increases over time. So yeah, can you agree on that? Isn't two and three kind of similar number two is more thinking about what kind of what kind of tools you think might be caught up by your competition by big tech. So it will be replaced soon. And our third one is understanding the features and also calculate the pricing. Tell us more. Okay. So the third one is the skill of understanding when is the way to complicate things more or less. Based on what tools you have at your disposal and how is the pricing going to to follow that. For instance, we had a product that the cost per month with AI models was $200. And then after understanding better, the peak hours of the product, we found out that that the peak hours are very narrow. And we migrated to a solution that was able to scale down the models. And then we reduce it 80% the costs, just with one simple this decision. And I think this is a critical skill for an AI PN. I see. So basically, making a business decision for the type of AI product and AI tools you will be using and also leveraging the user data usage, so that you can make the best technical decision knowing which one to choose, but driven by business insight. Am I right? Yes. Yes. That's what I think. Awesome. Awesome. Okay. So okay, great. Awesome. Now let me ask you this question. You know, lots of people always ask, hey Nancy, do you have any recommended AI courses? There's so many different kinds, some are free, some expensive, you know, so tell us. So what's your answer to this? There's so many free AI courses, also $1,000 AI courses. So which one do you recommend? For AI, I would recommend starts learning what the clouds have for you. So Google Cloud, AWS, Azure, they all have a lot of courses on how to use their off the shelf solutions. And when you start learning that then you can deepen as you needed depending on your product. So for AI in general, this is a good start. And also try to be in a community such as the PMA in that you can exchange experience with people that are actually using AI in practice. So I think it's more important to to get high, high level, high quality information than doing general courses for AI, which is a already an old topic. So I consider it's very old right now. Generative AI specifically, there are a lot of levels of implementation, I have mapped a five level framework to understand this craziness of the generative AI landscape. I hope soon Nancy and I will launch a course on that to help you. But so I mapped out these five levels and found out that most of the hype around generative is on the fifth level, the last one where you have the AI systems, detecting, creating, coming up with complex plans and checking this plan, prioritizing this plan and then following it to achieve a larger goal. So this would be the state of the art in the implementation generative AI. And the most simpler one is just simple prompting, you'll write your question, get the answer back. But in the second level, just with meta prompting, you can get a lot of good results. For instance, I don't write my prompts anymore. So I ask AI, a generative AI to to read some articles on the internet on how to write great prompts, and then write the prompt for me for summarizing x y z for doing x. And then it comes up with a big text one, one paragraph. And then I feed this. So you can do a lot of these things, which I'm calling meta prompting. I don't know if people are all settled on this term. But you can do a lot of things just with that. I think 90% of the problems we have right now that people are looking into gen AI, you will solve with the first and second levels. What you actually need to learn right now is how to make prompts because regardless of the level, there are prompts involved. For instance, in the most complex level, you have prompts for for designing your agents so that they can think about their question, they can raise questions. But in the end of the day, everything is prompting. And the first thing you need to learn is that and a good course I found out is the deep learning.ai. So the team from Andrew and G they have this course that's just one hour. It's the best course I've done. I have done a lot of them that go very, very deep on reinforced learning, human feedback, fine tuning, but you don't need to over complicate. Start with prompt meta prompting and then if you need more help, ask me or ask Nancy and then or ask an expert, but we'll probably don't need to get to these deeper levels soon. And knowing that you'll be able to understand, well, this company is tackling level one or two, this or their solution is just a set of prompts, a beautiful plot prompts, or maybe it's better to do in-house. But this other company is is now with a solution that is very, very smart that comes up with plan and then it's a much more sophisticated and high risk application. So maybe in that case, it's very useful to hire this company. Right now, most of companies are just in the first and second level of selling overpriced prompts. So learn prompts, you are good to go right now, you need more complex stuff, follow me on Medium, follow Nancy, and we'll get there together. Awesome. Yeah, I like the several height that you have. Hey guys, there's a one hour free prompting classes by Deep Learning AI. And we're gonna put all the recommended classes in the description of this video. They are all free and including all the AI tools by different AWS cloud, Google cloud is all there. You can start learning right away. And regarding five layers of JNAI, a Nicholas and I are actually planning to create another course. If you're interested for free to join our waitlist, once available, you'll be the first one get to know and also check out different median articles start learning for free right now. So that that's where you start to get a big grasp of AI. And of course, find different communities such as PMA communities, any community out there learn from someone who already started having experience, maybe from engineers or plan manager, anybody who started using AI, that's actually the best way most practical way for you to start learning starting from today. Awesome. So Nicholas, tell us more regarding different kind of JNAI use cases we should pursue right now. We'll maybe choose not to pursue some bad use cases you can tell us. So first of all, something that is important for people to understand that no one's telling clearly is that the large language models are just machines that you feed text and output text. I know you know that already, but they will truly grasp what this means. This means that everything a large language model is spinning out is an hallucination. Everything is a hallucination. It's not like it has critical thinking and it's smart and US and it will answers correctly 40% of time and 60% of the time it's hallucinating. No, 100% of the time it's hallucinating. But that doesn't mean it's useless. So how do you overcome this problem? You need to to make sure you fit the right context so that it answers a more trustworthy answer. And you can be think of getting the answers and then double checking with generative AI as well these answers. For instance, one use case that was very interesting that I found out so detecting new markets. Okay, you can use JNAI to come up with ideas, but understand that these ideas may make no sense. And then you ask the generative AI to raise questions about these ideas that invalidate these ideas. And what are the hypotheses behind these ideas? So I ask a generative AI, come with ideas, then what's the hypothesis after behind the idea? And what questions should be asked to validate to accept or disregard this idea? And then get the scratch, scrape the internet with JNAI as well to validate the to get the answers to these questions. And when you have the answers, you feed all that again in a bigger prompt and ask, provided this find this initial ideas, this hypothesis, and these answers, is it a good strategy or not? And then you get everything is drills down to more elaborate prompts and cascading these prompts and putting the right context. And then you improve your chances of success. And a key thought I came in touch with recently is that the AI cannot guarantee that an answer is correct or that you don't have risks. But it can help you to find risks if there are. So they can't guarantee there are no risks. But if there are risks, they can be used to detect these risks. For example, I have a product that detects cancer nodules cancerous nodules in magical exams images. I can ask the AI, is there any risk? If I this to I send it to to a doctor, it may come up with ideas of risks that may happen. And that could be really useful. If you send this to a doctor and the doctor does this brain bias is something like that. This doctor can lead can lead the doctor to ever what you cannot do is to make it guarantee that there are no risks. So you ask them, is this solution 100% safe? It may answer yes or no, but it doesn't matter because there may be risk that the AI is not seen. Understand what I I see, I see. So that's how you ask AI, you have some ideas, they ask what potential risk could be involved in your certain ideas, so that you have deeper understanding of the outcome and make your own decision after you understand into a bigger picture. Am I right? Yeah. So it's like, instead, the AI is good for finding like dots in a white page. But maybe it's not good to guarantee that a whole page is white. So it's something like that, you know. Gotcha. So it's a little bit crazy, I know. But you start getting a feeling of that using AI and improving on your promoting techniques. Going on the courses we mentioned as well. The original question was, what are use cases we should pursue not? So what are the use cases we should pursue? Yeah, so you should pursue cases that if you get very bad results, if you use out of the box solution, for instance, one example would be I'm Nicholas and I want to learn about what would be a good, a good next step on my career. If I go to chat tpt and ask that, what would be the next the best next move on my career? That the answer will be very generic useless. Yeah. But so if you get bad results, and then you elaborate a little bit more, you're prompt, and you start getting much better results we feel for, like, I fit in my resume, I fit in my goals, I fit in what I like to do what I don't like. I use the chat tpt addons to allow it to scrape internet and do a market research on that. So when when you start to perceive that as you feed more information and make your prompt more elaborate, your result becomes much better because it's more aligned with the specific situation. This is a great date for an AI product solution. Gotcha. Actually, we did something similar. I have a separate course inside the inner circle and it is free for many people feel free to go to the website right here to check out our inner circle where we thought of one hour course, we're gonna hey, how would you use chat tpt write a perfect personalized cover letter and wishes based on feeding into existing template of my cover letter and my resume, different kind of company, you actually cater to the company. And to my background to my existence style of writing and then you have a perfect example. Yeah, just like what you described about make a very specific teaching on the AI. Nicholas, you told us different examples of use cases idea. Let's talk about the use cases we shouldn't pursue. We shouldn't pursue cases that look so obvious that big that are general and that are useful in many different domains. For instance, an assistant that helps you schedule things. This is something that is useful regardless if you're a lawyer or if you were sales person or anything. So probably a big type of sales force or some company that already work on space, we just get their already existing solution, add some AI, some more AI sauce, and then come up with a much better solution. So you shouldn't pursue pursue things that look that look like a next step for what a big company would already do. So what's important as well is to think if you are pursuing a real a real business case, or you are just surfing the gen AI way as a product manager that needs to hire or at saving the children. If this organization needs to hire a solution, it's very important that the solution is trustworthy that the company is not going to fail six months from now. So we'll be looking a lot if the company already has a history in in this field for instance, health tech or is already doing something and he's using AI just to get to the next level. It's much different than a company that's a period two months ago and it's just surfing the gen to be a lot of startups trying to do that. By the way, I'm glad you pointed out and you're coming to be around for 100 years on trying to really use it to create the next generation of product instead of some startup just starting your company. Let's do something about generated AI, which was coming trying to do this. So that's why people need to be cautious regarding type of company want to join and also the type of company you want to become as well. Awesome. So thank you so much for sharing all the tips and advice today, Nicholas. And for everybody, make sure to check out our last video Nicholas filmed and regarding how he teach us regarding his career path become a head of AI and save the children international and make sure to go to the wait list regarding our upcoming AI for product management courses. And we'll be the first person who know once we make it available and make sure to share like and subscribe to the channel and keep tuned regarding our future AI product management training for free. Nicholas, thank you for joining us so much. Hey guys, this is Dr. Nia CV from T.M. Exciter Diome. See you in my next video right here.