 Hello, everybody. Thank you for joining today. My name is Ajay Swamy. I lead product for the customer segments team for AWS Solutions at Amazon Web Services. And today, we'll be talking about how to use GenAI to increase productivity in product management. As we all know, GenAI is the hot topic today. Everybody seems to be obsessed with it. But what I'd like to show you today is how you can increase some of your daily tasks, how you can automate some of your daily tasks, how you can use AI as your thought partner, and how you can use it for some of your creative inputs in product management. And you'll walk away today with some tips and tricks that you can start to use immediately. So here's the agenda for today. We're going to talk about the personas you can use GenAI for. I'll go through some of the GenAI tools that are really good for some of the tasks. And then we'll talk about how you can use GenAI across your product lifecycle management. Then I'll share with you a sort of a chart around which GenAI tools work best for which use cases. And then finally, we'll end with a wrap-up. Okay, so GenAI and the future of product management. This is actually a really good quote by Marty Kagan. Obviously, you know, we all need to recognize that AI is the next big wave, right? Like if you thought about the generational waves that are happening, it wasn't necessarily a Web3 or blockchain or crypto. AI is here to stay. And AI is going to fundamentally disrupt and alter the way we do our jobs, the way we learn, the way we communicate, all of the above. And, you know, you might be a little scared. You might be a little overwhelmed by trying to understand how to use AI or even learn AI. Number one, you know, the first step you need to understand about the AI is that look, you need to just accept it, right? Start using it in everyday life. Then you'll get comfortable with it. Read articles, take classes, recognize that it's here to stay. Number two, AI is not going to take away your job. You, as a product manager, you're there to use your instinct. You're there to use your judgment to understand customer problems, be empathetic, your cognitive empathy, your judgment, and all of the true inherent product manager skill sets are here to stay. AI cannot automate that or AI cannot read people's minds or AI cannot judge people's reactions. So, you're the glue that brings people together across a particular outcome and you'll always be important. What might change, however, is that your job will get more efficient, right? You might start to achieve a lot more with lesser time. So, that's one of the benefits of using Jenn and I today. So, my suggestion to you is instead of sort of looking at it as something scary or something overwhelming, start using all of these tools today. You know, go to ChatGPT and register online or subscribe to it or go to Claude and subscribe to it. You know, it's okay for you to sort of pay, you know, a monthly fee and then play around with these tools so you get to understand the value of it. And then the more you use it, the better you get at it. So, here are the main personas that you can use with Jenn and AI. I sort of look at these three personas and I think to myself like, hey, if I were to manage my tasks and my daily workload across these three personas, I think I can get a lot done. So, the first one is think of it as AI being your assistant, right? Like, for example, as a product manager, you're in meetings, you have to take notes, you have to follow up, do all that great stuff. Now, you know, you can use various Jenn AI tools like auto.ai that will take meeting notes, they'll summarize those notes and action items and you can share those back with your team to ensure alignment. Obviously, sometimes these tools do get it wrong in terms of what they picked up and what they summarized, but it will get you most of the way there. You can also use AI as an assistant to do some of your research-related work, right? Like, you can get data on a particular customer segment, you can get data about a particular market or a cohort, depending on which Jenn AI tool that you use. It will give you some pretty solid data. However, be skeptical sometimes, right? You might want to cross-check your work because if you use chat GPT 3.5, it's probably not completely up to date because I think it goes up to around June 2022, but if you use a newer chat GPT, which is 4.5 or chat GPT 4, I should say, that will actually give you up-to-date data. You can also use Jenn AI as an assistant to ask about topics you don't know about, right? Like, hey, tell me about Internet of Things and how would I start to use Internet of Things and where would I learn about it, right? Like, Jenn AI will actually give you topics and resources to consider and it can actually even put together a lesson plan for you to learn more about topics that you don't know about. And then lastly, look, if you're in a writing culture like I am, Amazon has a very, very good writing culture and you need to get really good at writing. Sometimes Jenn AI will also help you edit your notes, edit your writing and it'll actually give you various ways that can improve your writing. I, for example, use Notion, Notion.io. It's a great, great tool and they have actually induced Jenn AI within that product. So sometimes when I need to reformat a paragraph or sometimes when a paragraph is too technical or too complex, I'll turn to Notion and say, hey, can you make this more simpler? Can you make it more simpler from the perspective of a business leader who does not know anything about this technology? And it will do that for me and it'll get me 80 to 90% of the way there and it's a really good tool. So using Jenn AI as an assistant means you can probably forgo some of the hard manual tedious labor, laborious work that you need to do and again, automate them or even just use it to get you 80 or 90% of the way there and then focus on the most important tasks. The second persona, which I think is really powerful is that of a thought partner. You can use Jenn AI to play devil's advocate. What do I mean by that? So let's assume you are pitching your product idea to an investor, right? What you can do is you can go to Claude and Claude is really good at text-based summarization. You can upload documents into Claude and Claude will ingest that really quickly and you can ask it to say, hey, you are a venture capitalist. I have just uploaded my deck to you. It reads with the deck and ask me questions from a VC perspective. So I know to better prepare for the set of questions that are coming my way. You can also, for example, have better one-on-ones with your direct reports or with your manager, right? So let's assume you have a review coming up end of the year review and you have to provide feedback to your team. You can upload, for example, and make sure you remove any PII information but you can upload sort of a document that says, hey, this has been the performance of my direct report. How would I break the news to him that he has underperformed in these areas? Or maybe you're doing a salary negotiation. You could ask Jenny to be a talk partner and talk about negotiation techniques. You could say, hey, act as my boss where you will be negotiating for your salary and give me inputs and outputs that I should consider in order for me to make a really good argument. So the boss sort of buys my case for a promotion or for a bump in my base salary. So you can use all of these different ways to actually form a cohesive thought or an argument that'll actually get you further in terms of alignment across your stakeholders or even just getting you where you need it to be. And the last one is actually really fun. This is sort of the ideator creator persona. So you can use Jenny to actually generate product ideas. It's actually fantastic and it's actually pretty powerful. And I'll demonstrate some of these examples to you in the upcoming slides. But using Jenny, you can say, hey, give me a product idea for this particular pain point. Tell me about which customer segments it might be useful for. You can also use Jenny to create block posts, positioning statements. You can generate logos using stable diffusion. And by the way, here's a quick cheat sheet. You could actually generate ethics and user stories and acceptance criteria. If that is your D2J job as a product manager, this makes it so very easy. But again, I have to stress this. You always need double check your work, right? Like Jenny is good, but there are hallucinations. It might sometimes just give you completely wrong data or misinformation. So use it, but just make sure you double check your work. All right. So now let's jump in in terms of how you can use Jenny across your product lifecycle management. So here I'm using chatGPT. I'm using the free version, which is GPT 3.5. And as I said before, I think it has data that only goes up to June 2022. But for this webinar purposes, I think it's more than enough for me to show you how you can use some of the Jenny AI tools for this. So one of my first prompts, and by the way, if you want to get better at using some of these Jenny AI tools, you have to get really good at prompt writing. So I would highly suggest you take a class. There's so many different classes online. In fact, LinkedIn Learning has a class around how to use Jenny AI for product management or even project management. So look into those courses. It's free. It takes less than 20 minutes. And you'll walk away feeling like you've actually learned something and you can start to put it to use pretty quickly. So let's dig in. Like here's sort of the first part of the product lifecycle management. Like I want to come up with a new idea for a problem that I have, and I don't know how to solve for it. Or I'm looking for a creative way to solve for it. So my prompt is what I'm saying is I'm the head of a product at a tech company. I've been tasked with coming up with new ideas that solve for customer problem. Below are the details of the problem for the effect and the effective customer persona. Provide me with three product ideas to solve for the problem with pros and cons. And I also say decompose the ideas into epics and user stories. And I tell chat GPT to present them as a solution idea followed by the pros and cons and then epics and then user stories. And then I give it my customer problem persona and my problem statement, right? I'm saying I'm a busy professional. I always booked last minute trips. I never know where to travel. All I know is that I like beaches and very nice hotels. I don't mind spending money, but I just need to let work done for me in terms of my travel options. And typical travel sites like Expedia don't work because I can't just sit there and scan all of the different options for direct flights, hotels, safety of the region, et cetera. So help me do that. So I put that into chat GPT and it gave me three ideas. And I'm just showing you, and I just picked one of the ideas. The solution idea three that I suggested was an AI powered smart travel assistant app. And it gives me pros and cons, right? Like some of the pros are like it offers real-time updates and travel options, allows flexible planning, and it also gives you some cons. Cons being like, hey, look, since this is sort of an AI powered app, there might be a steeper learning curve for the users to get used to it. And then users might need connectivity. They need a stable internet connection because if you need real-time information, you need that constant stream to update your app. And then it goes into Epic City User Stories, right? Like the first Epic is we need to first build an AI powered recommendation engine. And it decomposes that into user story one and user story two. For example, it says as a user, I want AI to analyze my preferences and suggest personalized last minute travel options, right? And you can sort of break those down even further. Like I could have said, OK, can you break down user story one into bite-sized tasks, right? And it probably would do that for me. And then also it provided me some acceptance criteria, right? Like I basically said, you know, I like Solution ID 3. Can you please expand the user stories and add acceptance criteria? So it did that, right? For the user story, it says the app should prompt the user to input their travel preferences. The AI should use machine learning, the Algos to analyze those preferences real time and route for last minute travel suggestions based on user's profile, et cetera, et cetera. So you can see how powerful this is in order for you, you know, if you're struggling to get inspiration or struggling to come up with cool solutions to a particular problem, you can always go to something like chat chipping in 3.5 or even 4 and say, no, this is my problem. This is my customer persona. Help me come up with a solution. And it also provided more constraints, right? You could say, you know, things like I need it to be cheap. I need it to be quick. I need it to be easy to understand, et cetera, et cetera. And you'll have a very good starting point for you to actually come up with a very creative solution. This section is about using GNI for user segmentation research, right? So here, I want to say, look, I want to create a new SaaS product for invoice processing using AI. What does that really mean? It could mean a bunch of things, right? And I'm trying to dig in more and see if there really is something there. Like, is there really value here? Is there really a market opportunity here for me to build something or invoice processing? So I asked the, so I clarified from my prompt and I say, what customer personas would find this useful? Create a list of personas and prioritize them. And for each persona, provide the name, current job function, pain points and needs. And chat GPT comes up with some pretty cool stuff, right? It tells me, you know, there's, you know, a bunch of different personas that will find an AI-powered invoice processing SaaS application useful, right? So for example, first is the accounts payable manager. This accounts payable manager works at a medium-sized manufacturing company. Her pain points are the high volume of invoices that lead to manual data entry errors. She has a lot of time consuming approval workflows and it leads to delayed payments. If the payments are delayed, then the vendors are not happy and that's just, you know, that means like the company loses credibility, all of that stuff, right? And what are some of her needs? The needs are like, hey, she wants to automate data extraction and entry and she wants to streamline her approval process. So it's pretty cool stuff. So it seems like there is probably personas that can find it, that'll find us the SaaS application very useful, right? Then I say, you know, then I prompt Jenny a little bit further and I say, tell me about the first persona. What is their day-to-day job function? What does a D in the life look like? So this actually picks up on that through the context window and it says, hey, for the first persona, Karen Miller, the day-to-day function, the day-to-day job function is that she oversees and manages the financial responsibilities related to the payment of invoices. Her role is to ensure that all invoices are processed accurately, efficiently, and in compliance with company policies, et cetera. And it also goes in great detail about the morning routine, right? She starts her day by reviewing any urgent emails. She checks the dashboard of the AI-powered invoice tool to get an overview of pending invoices, et cetera, et cetera. So this is really, really powerful because not only do you have the persona and the job function, the pain points and the needs and how this person feels the anxiety of doing her job, you can also see the day of her life and you could actually map out that user journey. And I would even argue that you can incorporate the Hooked model into this. The Hooked model, if you don't know, is the way to turn product usage into behaviors. It's by this professor, this gentleman named Nira Ayal, where you sort of take an internal trouble, which is your anxiety of not being able to process an invoice. And you sort of turn that into a behavioral response where you get a reward that every time you automatically process an invoice, you keep coming back to that application to solve for it more and more. So through this user journey, you can really sort of map out her internal triggers, her anxiety, what makes her anxious, as to why she might want to use this application to actually achieve the automation that she desires. So let's take this a little further. So now that I have some sort of view into what this product is, who my user personas are, who might find it useful, let's validate that. So first thing I want to ask Chad GBT to do is, hey, provide me a name for this product. So it suggests a product. You can go back to it and ask for more different names if you wanted it to, and I'll come up with more names. But I just did the first one. I called it invoice Intel Pro, which I think is pretty cool. And then I say, hey, tell me about the market trends related to invoice Intel Pro. Like, who are the competitors? Is the market for invoice Intel Pro growing? Or is it stagnant? What is the market size for invoice Intel Pro? So it tells me a little bit about the market trends. It talks about the rise of automation. It talks about the capability and opportunities to integrate with ERP systems that is used by so many different enterprises today. It talks about the growth in machine learning and data extraction, which is actually right. Machine learning is on the rise and data extraction in terms of zero ETL. That's on the rise as well. Security features and with cloud, all of the stuff is so easy to build and roll out. In the comparative space, it kind of struggled a little bit. And that's because of the fact that the GPD 3.5 is not necessarily have all of the data that is up to date. But still, it does a pretty good job. And then I go a little bit further. I'm still validating this product idea, but I want more. The next thing that I ask chat GBTS, hey, what are some common monetization and pricing strategies for invoice Intel Pro? What are some of the competitors in the space charge? And it came up with some pretty cool monetization strategies. If you think about pricing, you have competitive pricing, you have value-based pricing, or you have cost-based pricing. So here it actually goes into all of those different models and it talks about how you might want to incorporate monetization strategy. And I really like the fact that it called out the freemium model. It also called out the custom enterprise level plans. It also gave me an option if I wanted to go transaction-based pricing, i.e. like, hey, for every invoice processed, I take $2. But it gives me all of these options that I could use to model out, or it can ask chat GBTS to model out based on 10,000 invoices, which pricing model really works well for me. And it also gives me a hypothetical competitor pricing insights. It doesn't have necessarily specific information, but it sort of goes into these different ranges of business plans, whether it's a small business plan, professional plan, et cetera. And you can see how this can be absolutely powerful for you, right? As a PM, it's trying to lead a pricing strategy discussion very early on in the stage. This can be really useful for you to sort of look at all of this, all of these options that are really, really different, and can combine them to come up with the best pricing strategy for your product. All right, so this is a pretty, so maybe the last part of your part of lifecycle management is to actually create the product, right? Create the MLP, your minimum lovable product. And maybe after that, you want to go into like, hey, how do I maintain? How do I sort of prioritize fast follow stuff like that? But I won't go into that. But I'm using this last part of the build, like the early part of the build process, and I'm using chat GPT to try and get an idea about what I need from an MLP perspective. So I asked chat GPT, hey, give me a rough prototype for invoice Intel Pro. Can you give me a list of user screens that are needed and what each screen will do? And it does a pretty good job, right? It tells me, look, I need a dashboard because you want to understand like where all of the invoices sit, what's approved, what's not approved, what the status of it are, whether it's paid, unpaid, any alerts and notifications, et cetera. Then it says the second screen may be like, you know, the ability to upload invoices in various formats. And then it gives you feedback on successful uploads. The third one is about processing queue, right? Like which ones are in the processing queue? Which ones might probably need like a four eyes approval, et cetera, right? And then basically the last one is maybe the invoice details. Maybe you extract all of that information from the invoice. You dump it into your database and then you present it on the screen to the accounts payable manager and they can manually check to make sure that everything is correct. It's pretty great stuff, right? Like if you were to write a product requirements document, I think this is doing a majority of the work for you, which is fantastic. And then if you want to get a little bit more crazy, you can actually ask chat GPT and prompted and say, hey, look, can you generate the code for me in Node.js and HTML for the first screen? I'll do that for you. Obviously, look, if you have a good engineering team, they probably will find this not useful because they're just not going to take code from a GNAI engine. But you can see how powerful GNAI can be. This gives you an idea after the capabilities of what it can actually do and how it can actually help you get past that hump very quickly. All right, cool. So I want to sort of tell you a little bit about which GNAI tools you should be using and for what use cases. So since I work for Amazon, I'm obviously biased, but I think Bedrock is like one of the great, great AI providers. It's secure. It's reliable. It has ethical AI built into it. And it offers multiple LLMs, right? You could use Claude. You could use Table Diffusion. Go here. You could use Metas, Llama. And then obviously you have Amazon's own Titan models. And it offers both text and image use cases. The rest of the suggestions here, they come from section school. Section school is a great school. I've taken a few classes there, by the way. This is like a free plug for section school. But one of the things that they actually talk about is how to get better at prompt writing. And they did a little study there and they ranked some of these other GNAI providers and what it's useful for. So for example, Claude, if you look at Claude, it's really good for processing large documents, organizing data. And what that means is like you can upload a document and you can get insights on the document or suggestions for improvement. It's almost like on-the-go rag, right? Like if you're not familiar with rag, rag is retrieval augmented generation, which basically means that you sort of kind of fine tune an LLM model using external data, right? And that's why you're sort of doing on-the-go here. GPT-4, as I said before, is really good. It has added capability of plugins. So code interpretation, data analysis becomes super easy. You could upload file with a bar chart and ask what's going on and it'll tell you, which is very cool. Claude, too, is really good for managing even larger documents. It can handle many file types, like I've uploaded PDF files, PowerPoint files, and it'll critique and it'll tell me what's good and what's wrong or improve this or improve why. It even looked at one of my decks and it said, hey, you could improve the flow here. Why don't you approach it from this storyline? Really powerful stuff. So you'll obviously have this deck after this webinar, but use this as a cheat sheet to think about like how you want to use what Genai provider for what use case. And then here, well, here's a bonus thing, right? Like I'm a big fan of Claude and Claude is really good for text-based interpretation, text-based analysis, summarization. It's good for being your thought advisor that I talked about earlier. It plays a very good devil's advocate. So one of the things I said, if you're in the job market right now and looking for a new job or if you're just trying to improve your resume, one of the things you could do is go to Claude and you can say, hey, you're an executive recruiter for top 10% of product managers in New York. Review my resume and suggest improvements. So I can get an executive level product management position at a leading tech company, provide examples, improve my heading, summary the work experiences, and I have a piece of my resume. Basically I took a copy of my resume and I just uploaded that. And it gives me pretty solid suggestions, right? It says my current summary focuses too much on my technical skills, which is actually true in retrospect. I am a former software engineer and I work for AWS, which means I'm really technical, but it says, hey, if you want a higher leadership position, you might want to focus on leadership and factor strategic vision. And it gives me a suggested edit, which is really, really great. And then number two, I also asked it like, hey, can you expand on my work experiences and also provide me specific examples of where I can improve this? So it says, look for Amazon web services, change some of your points to really focus on business related outcomes. When I worked at McKinsey, it told me that, hey, pick out some of these experiences that you have bulleted and then bubble that up and then talk about it in terms of outcomes and improvements. And that'll really tell the story from a business perspective. Again, I always say this, be careful, don't lie, don't over embellish because you will get caught, right? Like, to get better at product management is to improve your competence. You're not going to be able to do this just by improving your resume. But this will give you some really good strategic advice on how you should be thinking about experience and how you should be altering your resume from the perspective of the person who's going to read it. So it actually connects, actually lands your interviews and gets you through that door. All right, so that brings us to the end. So a few parting thoughts. Look, double check your work. No matter what LLM you use or what generic provider you use, these tools are there to accelerate your time to value. That doesn't mean they're always truthful, right? There will be hallucinations. So please double check your work. Be careful not to upload any proprietary data, PII data because that can land you in hot trouble very quickly. So be careful about that. That's one of the reasons why I really like Amazon's Bedrock because if you have an AWS account, it's all sort of within that account. It's all data, it's sort of sectioned off, right? Here, if you upload data to chat GBT or into cloud, you feel like it's going into ether. You have no idea where the data is going. But if you use one of these really, really good enterprise cloud providers, I think your data is safe. The third thing I would say is take a course on prompt writing. There's so many free courses out there. It's really, really useful. And that'll give you more confidence in communicating with some of these GNI tools. And look, there are no shortcuts, right? Like if you want to get better at AI ML, start with prompt writing and using these tools, but then also start taking courses in AI and ML. ML is sort of the foundational aspect of artificial intelligence. Go and learn. And then hopefully this was helpful. Hopefully this can help you accelerate your time to value for building new products or improving existing ones. Thank you. And if you have any questions, feel free to reach out to me. Have a great day.