Q: What are the key steps to solving this problem? A:
- Create a central “Memory Document” to house all core preferences and system operating instructions.
- Export and consolidate existing memory from other AI tools (like ChatGPT) to quickly build context.
- Leverage NotebookLM to ingest massive amounts of contextual data, such as transcribed meeting recordings and video transcripts.
- Query the initial memory document to identify missing dimensions (e.g., brand voice, philosophy) and manually update for maximum completeness.
Q: What is itGenius? A: itGenius is an IT consultancy that helps small businesses scale effectively by providing affordable and effective technology services, specializing in Google Workspace support and strategy. We offer both transactional support and an “all-you-can-eat” Cloud Concierge subscription.
The future of AI is an all-encompassing companion that will have the context of your entire business life, drawing on every document, email, and conversation you have ever had [00:22]. While this reality is coming, small business owners using Google Workspace don’t have time to wait. Gemini’s automated memory feature is currently only active for personal Gmail accounts, leaving business accounts to rely on generic results [01:39].
The solution is to get ahead and proactively build a comprehensive, custom memory document that trains your Gemini companion now, turning it into a truly personalized “Biz Gem” [02:42]. This active approach ensures your AI always has the depth of context it needs to deliver superior, relevant results that help you scale effectively [05:57].
The Core Foundation: Building Your Memory Document
To start, create a single, centralized document that will serve as the source of truth for your AI companion [03:20]. This document should contain detailed instructions on how you want Gemini to respond, including a set of system operating instructions [03:47]. This includes basics like localization preferences (metric vs. imperial) and a desire for shorter, more direct instructions [04:03]. As you begin to notice answers that you want to tweak, you can easily refine these preferences in this central document.
Consolidating Context from Existing Tools
A fast way to kickstart your memory document is to draw from existing AI tools you may have already been using. ChatGPT, for example, is automatically building a dossier of memory in the background based on your conversations [01:16]. You can easily ask ChatGPT to “Export every note in my memory into one summary” to get a full “memory dump” [06:47]. Cut and paste this consolidated data into your Gemini memory document, saving you significant time in gathering initial context [07:08].
Leveraging Rich Data Sources with NotebookLM
For true depth, one of the best data sources is internal business information, specifically meeting transcripts. NotebookLM is described as a “smarter scrapbook” that allows you to gather data from multiple sources and run powerful queries on it [08:34]. An excellent strategy is to use NotebookLM to import all of your transcribed Google Meet meetings [08:45]. Since it is critical to record and transcribe every meeting in your Google Workspace account, you may have hundreds of hours of data that reveal your leadership style, philosophy, and business principles [09:23]. You can then query this immense data set—for instance, by asking, “From all of these meetings and conversations where I have spoken, can you tell me what you know about me that would be useful information to put into a memory document for Gemini?” [10:19].
The Iterative Process of Refinement
After gathering your initial context, the critical next step is refinement. Use Gemini to query your own memory document by asking it to analyze the existing context and highlight any missing areas that would benefit from manual updates [07:22]. It may flag key dimensions like your brand voice, business unique strengths, or team culture performance as currently empty [07:43]. This iterative process of putting in context and then querying the AI to refine it is what ensures your Gemini memory document becomes truly complete and effective [07:55].
Watch: Creating Memory for Gemini in Google Workspace [Guide Part 1]
Ready for Expert IT Support?
Our mission is to give you control over your technology strategy. If you’d rather have an expert handle this for you, we have a few ways we can help:
Cloud Concierge Membership: Get unlimited, “all-you-can-eat” tech support for you and your team with our yearly subscription. The ultimate peace of mind. Click Here: https://www.itgenius.com/concierge/
One-Off Projects: Unsure if your Google Workspace is set up for maximum security and efficiency? Let our experts review your configuration and provide a clear roadmap to protect your data with our Comprehensive Audit.
Quick Fix: Have a small, urgent tech issue that’s slowing you down? Our is designed for instant IT support. click here:https://www.itgenius.com/quickfix/
Book a Free Consultation with Our Team to Find the Right Fit here https://www.itgenius.com/products/consultation/
Ready to move beyond the basics and unlock the full power of AI for your business? Join the conversation in our Gemini AI Community to get direct access to the latest strategies and tools before anyone else.
Full Video Transcription
this is from Sam Ultman he basically said the vision for chat GBT and AI tools in general is that they really want to have AI be this all-encompassing companion that goes with you everywhere you go and it basically has all of the context of your whole life so effectively what I’m saying here is plan for a future where every bit of data that you have is going to be added as context to the language model tools that we’re talking to every email, every text message that’s in your phone, every Google doc you’ve ever created, every pretty picture you’ve drawn in Canva. You will consent to this, remember this will be opt-in, and you will consent to this because it will be so convenient and it’ll just make the experience of these tools so much better. So what are we trying to do today by building our Gemini memory? We’re trying to get ahead of the game and create this new reality. Now OpenAI has already done a pretty good job of doing this automatically inside Chat GBT. If I open my Chat GBT it will automatically start creating memory for me. What Chat GBT does is it automatically picks up conversations that you have and it’s building this memory in the background automatically. It’s building this like kind of dossier on you based on the things that it finds in the conversations that you have. Gemini has this feature but it’s only switched on for Gmail accounts, personal Gmail accounts. It’s not yet switched on for your business account, so we’re missing this feature right now for business accounts. It will eventually be added, but it’s not there yet. The other limitation of this is this is pretty basic. I wouldn’t say this like deeply, deeply knows me yet, and it’s got a limit on how much memory is actually collected here. It is useful but I think we can still do a bit of a better job. Remember, in 6 months time, 12 months time, 18 months time, every bit of data from every app that you’re using or have used is going to be fully included in the context. They don’t have the computing power for that yet, but they will, and so at one point in the future, all of this is just going to be basically available for you as context. But for now, we’ve got to add that context, so that’s what we’re going to build together. What the finished result looks like is it’s called “My Biz Gem” and anything I put into “My Biz Gem” is going to use the context of my memory that I’ve created. My memory has all been added to this Biz Gem, and anytime I have a conversation with Gemini, I’m actually opening up this Gem to have the conversation rather than just going “new chat” which is just a fresh one. I always start with this one, and you might want to call it “Memory Gem.” Any conversation I have here has got all of the context that I’ve added automatically, and that’s what I’m going to show you how to build. The basis of this is based on one document of Gemini memory. What I’ve got here is a document which has got as much information as I can gather on me and it’s in bullet form. Number one, I’ve set my preferences for chat, so I’ve said this is how I want you to respond to me in chat. I’ve manually given a set of system operating instructions, things like even though I’m based overseas, I might want specific metric or imperial type preferences—all those basics of localization. There is for me a desire to have shorter, more direct instructions, so I put those in here. This is just a set of preferences. If I notice I’m getting answers that I don’t really like or I want to tweak them, I tweak these preferences. Then we’ve got all of the different kind of like titles and headings here: information on business, information on the company, lots of those things. Next, I have some pretty good data sources of information on myself. I’ve got a couple of YouTube channels, and they’ve got lots of journal-style videos, and from those, I can export those. I used NotebookLM, and I actually got a lot of information into documents as well, like my business journey, my kind of philosophy, guiding principles on business, you know, how I think. These came from a couple of interviews, so if you have a resource like that, even if it’s not a public video on YouTube, that is very, very valuable. You can go to any YouTube video if it’s a public channel, you can go to the channel, and you can basically snip that channel into NotebookLM. If it’s not a public channel, that’s okay. You can take any private video, you can put it into NotebookLM, and you can have it transcribe that video and get the insights from it. Then leadership style, so I took this from meeting transcripts, and then business strategy was another query that I added from business strategy, and these are all just different dimensions of context of Peter Moriati. What I’ve tried to do is think of like how can I bring in as many different angles and dimensions of Peter Moriati that would be useful to train my AI companion and have it have lots of detail on my preferences so it really, really understands me differently in lieu of having every single conversation ingested, which eventually we’ll get there. I’ve created some headings and I’ve got some basic prompts headings here, and we’re going to use Gemini now to build out these headings. The first thing we need to do is we need to go and find our data sources. I think the easiest thing to do first up is to go to Chat GBT because that’s probably already got some memory on you if you’ve been using it. Very basic prompt: I said, “Export every note in my memory into one summary.” That’s the prompt. Basically, you’re asking Chat GBT to give you a memory dump. This has basically given me a dump of everything in my memory in Chat GBT, and what we’re going to do is we’re going to cut and paste that and we’re going to put it into our Gemini memory. Now we’ve got some headings here, and your Chat GBT memory might not have covered all of that so far. What you could do is you might ask Gemini, “Using these headings, can you analyze this document and tell me if there’s anything I’ve missed and should update manually to add more context to this document?” And then I’m going to add the headings in there. It’s going to tell me if there’s anything that I should add: philosophy, brand voice, and my business unique strengths currently empty that would benefit from manual updates, and team culture performance could be updated as well. We’re asking and we’re querying, “Hey, how do I make this more complete?” This is an important part of this process: we put in context, but then we want to query, “Okay, how do I refine this? How do I make it better?” Now, one of the best data sources that I’ve found for this is NotebookLM. When we go into NotebookLM, what this lets us do is to gather lots of data from lots of different sources in one place, and then we can query that data. NotebookLM is like a smarter scrapbook; it’s a bit of a digital scrapbook where you can bring in lots of different sources and then you can run queries on those sources. I’ve been going to my NotebookLM, I click the Google Drive button, and then I search for the word “transcript,” and that is going to let me import meetings that I’ve had, and ideally internal meetings, but here I can import meetings from the business. I’ve got 130 meetings in here, so this is like literally hundreds and hundreds of hours of data of conversations that we’ve had and every word transcribed. I’ve shared a number of times how important it is to record and transcribe every meeting in your Google Workspace account. You can switch this on to happen automatically in your admin panel. If you want auto-record switched on and auto-transcribe switched on, just flick a message to our team via chat or via email and say, “Hey, can you switch on auto recording?” As soon as you jump into a meeting, it’ll automatically start recording and transcribing, and it just puts it all in your My Drive. Once you do that, when I go to add a document to Google Drive here and I search for the word “transcript,” I get hundreds and hundreds and hundreds and hundreds of contextual pieces of information that I can bring into a notebook here. From here, I can query this, and I want to make a memory document. Let’s query it and say something along the lines of, “From all of these meetings and conversations where I have spoken, can you tell me what you know about me that would be useful information to put into a memory document for Gemini?” A simple query like that, and this again should give us a long output of information. It’s impressive when it scans a 10-page document and gives you a summary. It’s super impressive when it reads 200 10-page documents and then gives you summaries. That’s the end of part one.
Creating Memory for Gemini in Google Workspace [Guide Part 1]
itGenius 🤓 Biz Tech Experts · 555 views
