Building a Knowledge Base with NotebookLM (Step by Step)

if you've been trying to keep up with AI and wondering how you're going to be able to keep your Edge in the workforce then you are not alone many people are wondering what's going to happen once these AI tools become ubiquitous and everybody has access to them and everybody knows how to use them the one thing that I keep coming back to is data data around whatever you're good at around your audience around the trends and your industries the more you can collect this data and organize it the more Edge you're going to have as these tools get more powerful your knowledge in this area your knowledge base of data will continue to pay dividends and building these knowledge bases I believe is the number one thing that we should all be working on and that's what this video is all about this is a intro to building your knowledge base specifically around your audience and your customers although you might want to think of all these things that we're doing today in light of maybe just your industry Trends or maybe your brand itself and the uh expertise that your subject matter experts have those are all different ways to think about these knowledge bases on this channel we've talked a lot about custom gpts that go through a specific process I believe there's a whole second category of these custom gpts that are for knowledge based collection and interacting with you know particular personas and that's what we're going to dive into today if you're new to the Blazing zebra Channel I want to welcome you and thank you for joining me on my journey of helping entrepreneurs marketers and professionals all around the world learn to use these powerful new AI tools I've got a patreon membership that has cheat sheets available to every single video that I create including this one you can jump in there and get access to well over 80 cheat sheets with all sorts of resources there including how to build knowledge bases for your customers for your industry and for your expertise to give you an edge in this new world of AI jumping right into today's video we are going to look at using uh creating knowledge bases with notebook LM notebook LM is a new Google product it's free uh and it I think is the most basic way to build and test these knowledge bases there's a lot of ways to get fancy with them and we'll be covering those on future videos but just to get started and experimenting and learning I think this is a great place to start and it may be perfect uh for a lot of people going forward it's a powerful tool but first we're going to look at how to analyze and summarize the source material so organizing your Source material I think is critical you hear a lot of people talking about they're just going to dump everything in there that's going to get very confusing to the AI there is a term in um uh Technologies is garbage in garbage out so if you don't organize what you're putting in there you're not going to get great results so we're going to start with some best practices there I'm going to show you how to create a new notebook and add those sources and then we're going to test and play around with it a little bit so the benefits of building a knowledge base are many you can use it to tailor your messaging to specific audience needs I'm going to show you that it was really mind-blowing how much better the recommendations from this knowledge base were than from general you know chat GPT or other tools you can gain insights you can enhance user engagement and operational efficiencies these can be used for uh customer support knowledge bases um and there are a lot of I think ways that these can be used in the future for Revenue as you may be able to just lease these models out lease this data out to other folks who may need it so this is going to open a lot of doors in the future I think it's well worth your time to focus in this area and how to build these uh custom gpts that are wrapped around knowledge bases today we're going to be focusing on Notebook LM but in future videos we're going to be looking at how to build these same things in custom gpts and maybe even into uh code API access type uh chat Bots Etc so why notebook LM these are really powerful AI powered notebooks I consider them more than that um they are can be a central Repository for any types of information how you organize that information is is very critical I think people have this assumption that they can put all the information about them about their customers about their business about their family Etc all into one place no you want to get granular um for for me I think of for sure creating one around my audience with various personas in there as I get more and more data I might split that out into various different personas or if it's internal uh knowledge bases I might split that up into the various levels or areas that people in my organization are um knowledgeable at so you know you might want to create one all with just your brand standards and with best practices for creating blog posts you might want to create a separate one all about you know your product itself and the uh you know benefits to your customers Etc so thinking about how you organize these I think is critical and really this is more of an art than a science currently so feel free to be creative and think outside of the box with some of this stuff there's no right perfect way to do it uh but I think there is a wrong way to do it which is trying to dump too much into these uh knowledge bases notebook LM has some really cool features that are are going to be helpful you can think of this to uh really help aid against hallucinations a major concern with AI is that it makes up stuff which I think is getting better and better I haven't seen it really go off the rails and make up a ton of stuff recently as much as it used to but if you're adding all of you know trusted information to it so you know everything that you're uploading into these notebooks into these knowledge bases to be correct then you're just using the AI as a reasoning engine and not as a search engine it's referencing material that you know is legit that you know is true therefore you're going to be able to uh reduce any hallucinations and just focus on uh the core valuable input that you have there okay so tons of stuff you can use these for summarization question answering idea generation one really cool thing about notebook LM is it uh shows you the citations of exactly where in its knowledge base it's pulling that information from that can be really helpful for a lot of reasons as you're testing it so let's Dive Right In starting with analysis and summary of sources so the way I got started with this I have a a group call that I work with some of my patreons on Tuesdays at 2 p. m. I've been recording that call and I've just just dropped that into descript here to create this uh transcript so I just copied and paste everything from this transcript you don't have to use descript you can use any transcription service that you like as long as it's not summarizations I think we want to we want to get all of the actual words folks are using verbatim there's a lot of summary tools out there sometimes they over summarize and you lose the Nuance of the conversation so I'm really especially for use in marketing I want to make sure I'm you capturing the actual words that my audience that these PE folks that I work with how they're describing these problems in their own words can be very helpful as I'm trying to come up with you know titles for my videos or ways of explaining specific problems so grabbing everything out of this everything here out of the script and I'm going to drop all of that right into Claude I think Claude does the best with large sections of text like a whole call transcript and now I'm just going to start playing around with this transcript and seeing what I can pull out of it what I think might be best for a knowledge base I've got a cheat sheet for this video and all of my videos there's a link to that in the description this goes through everything I'm going through now but it has a bunch of extra prompts which are going to be specifically helpful for cleaning this data so you can see here I've got the core transcript preparation prompts and then many others going into focusing on action items and decisions that can be a big part you want to pull out of transcripts and pull into your knowledge base prompts focusing in on specific topics prompts for deeper analysis specific use cases general cleaning prompts depending on how messy your data is specific cleaning prompts and others for refinement categorization and separating things into documents so there is a lot in this cheat sheet that can help you specifically cleaning this information and organizing these uh transcripts or whatever you're dumping in there into something that the uh AI can use let's hop back into Claude here and see what it came up with I just asked it for um some ideas on how I can understand this first of all and it gave me a bunch of insights here directly based on uh this conversation this was a great onh hour call these are the calls we do Tuesdays at 2 p.

m. uh with uh folks that are really taking this AI stuff seriously so it's always great I always learn a ton from these calls and they are a valuable way for me to understand what's going on with my audience so I then asked it hey can you outline this transcript into a general flow of maybe the five to 10 major themes that we've discussed the reason I did this is because I want to separate this out into these various themes so I can focus in with Claude zooming into each one of those individually because the call was just too long to say hey give me a summary of all of the questions and all of the insights and all of the problems from this entire call I don't think that that's going to fit into to a typical response from Claude And I want to focus in and get a good response for each of these different areas so that I'm getting the most out of this rather than trying to do too much just with one prompt so in this way I'm splitting it up so I can focus in Focus the AI deeper into these specific areas so next here for the number two for that second section of its outline that one really jumped out at me they had some questions around custom GPT experience so I said let's dive a little deeper into that number two can you return the comments or questions my customers had verbatim like I said I want to keep their language intact this was a little bit wordy so I think that I wanted to clean it a little bit more which I'll do here in a second but first I repeated it for section three related to using AI for legal documents and contract negotiation and now I'm going through and asking it please edit this to make them more concise but keep the exact language used intact so now we're shortening things up a little bit here so that we're not dumping all of that in when I say the word verbatim Claude takes that very literally some of these I think are a little bit too long so I want to tighten this up a little bit before I'm using it in my knowledge base but before that I've asked it to do the same thing for section four and five getting a feel for what we covered now I've simply asked it please return as many of the questions asked in this transcript as possible this is another way to just get it to focus in on the questions themselves lot of good questions here and I thought it might be running out of its response I it to please continue here to see if there is any more questions and it says there are no more questions so that's a really good uh response there that we've got all the questions those are very helpful for me for what I'm doing with this stuff which is trying to figure out what questions folks have so I can create content around it now I did the same thing here please create a complete list of all the problems that were discussed on this call and here are the problems people are struggling with very helpful but it didn't include the actual words that folks were using so I wanted to pull that back in I said for each of these add a few words that the participant used to describe these problems and now we're getting the problem and then some very short concise words that are used to describe that problem this is gold for understanding your audience understanding your customers which is the number one problem that marketers face we're inherently uh coming at these things from a different angle than our audience and our customers are and it's very hard to get into their heads that's what this AI can really excel at now I've asked it to separate those different problems into different personas because I know that the folks on this call are very very different you know some are coming from marketing agencies some are just entrepreneurs so I wanted to separate that out the best that I could and it did a pretty good job here of pulling that out even though I didn't really give it very much information about the folks I gave it no information about the folks on the call it really just gleaned this from uh the the call transcript itself now I was moving on from the problems to the insights gained on this call so I asked it to please create a complete list of all the parts of this call that seem to resonate with the participants places where they seem to learn something please in include a few of the specific words they used in these instances so this is the flip side of the problems all the insights that they gained so that I can make note of that think about how I can incorporate that in my future trainings and with that I felt like I had a pretty good starting place to start incorporating this into the knowledge base into notebook LM here we are in Notebook LM it's just notebook lm. gooogle and you got to click try notebook LM if you run into issues you need to uh adjust some things potentially in your admin so you might want to float this uh link to the admin you can do a search for this or I have a link to this in the cheat sheet as well and it'll walk you through uh how to access this there are just a couple settings you need to flip inside of the admin section specifically in the apps additional Google services giving folks access to Early Access apps nothing too hard there just a couple Hoops you got to jump through I think it's well worth it to get into this tool and once you're in you're going to see something like this has some example notebooks and then it has this notebook that I created here I'll create a new one and I'm going to just copy this text straight in from Claude I've created this I call this the AI challenges discussed on the group call May 28th it does its cool little summary of it it gives it some key topics so there's actually AI that's already helping you organize this stuff embedded in here taking it one step farther than what you've done I've gone ahead and done the same exact thing from the insights so this was only two files I just took the challenges uh that folks were having and the insights so the good and the bad that came out of those calls dropped that in here and now I have the basics of my customer knowledge base and I'm going to continue to do this with every group call and every other uh potential appropriate call that I have pulling this stuff in here so I can get more and more Nuance understanding of what my audience is going through now we'll just kick the tires and give this a try I'm just asking it to please generate some content ideas for this audience so now it really knows a lot more than you could stuff into any one prompt about the audience and it's coming directly from my specific audience so uh getting through all of the noise that might be in the training data and it is generating some awesome ideas the one I'm working on right now the AI powered knowledge base uh it's recommending that which is pretty interesting um tons of cool stuff a practical guide to choosing the right AI tools for your needs and it kind of quotes why it's telling us this stepbystep guide to building your first AI model some really cool stuff in here directly related to those conversations reimagining your forms with AI creating userfriendly online experience that convert so this is a big thing that came out of that call and something I'll probably work on uh is how to replace these generic forms that folks have to fill out on your website with uh AI counter parts that can really help with those things both on the ways that the users are filling out the forms as well as what the folks who are getting the forms you know what they're seeing and how you can filter and manipulate that information there which was a really cool thing that came out of this call so tons of really cool stuff again this is another video I'm working on already so it is right on the money here uh as far as things that these folks were interested in and I'm going to be using this notebook LM quite a bit going forward there are I think next levels as I mentioned I'm going to experiment with dropping these two you know documents into uh the knowledge section of the custom gpts on the open AI side on the chat GPT side and then I'm going to probably down the road start playing around with API features and ways I can plug these in to um you know other projects that I'm working on there's just a ton you can use this stuff for so thinking outside of the box you can use these if you're thinking of raising your prices if you're thinking of rolling out a new software feature what features do you want to focus on the more info that you have in there that's highly organized the better so you're helping its memory helping it uh to find the things that you need notebook LM is pretty cool you can share it between teams you can use it as you know a knowledge base for different processes for the employee handbook you can create a you know FAQ type of Bot that you can post on your website uh you can allow other people to uh you know put up put new information in there and that way your whole team can be collaborating on building this so I hope you got something out of this again there is the cheat sheet that has tons of cleaning prompts to help you organize your knowledge-based documents it also includes step-by-step processes and a little bit more about notebook LM so check out out that cheat sheet also take a look at my patreon I've got a lot of coaching options in there there are these Tuesday calls Tuesday at 2 p.

m.