Last week, I shared a guide for how to decide when to use market research specific AI tools versus a business subscription versus a consumer subscription to an AI model to do market research for a customer. One of my readers went and created a flowchart infographic of that particular article. Huge shout out to Dean Stephens for creating this and being willing to allow me to share it with everyone else!

This week, I want to talk about Gemini Notebook (formerly known as NotebookLM) and some use cases for it, specifically for the insights community. It was going to be about the projects across ChatGPT, Claude, and Microsoft, along with Gemini Notebook, but Claude just made a major update to their Projects (*checks watch*) yesterday, and I have yet to understand if the changes impact how it might be used for insights work. Gemini Notebook hasn’t (yet) made updates, so here we go.

You may have heard about Gemini Notebook, where people upload a bunch of documents, websites, and other files into a “notebook” and then use AI to summarize it all podcast-style.

But there are certain key features that I think make this a useful addition to an insights tech suite.

First, it’s more than just a place to upload lots of information and then generate a podcast summary. You can also generate:

And you can simply write questions that are answered using only the information in the notebook.

Any query or item generated within a Notebook only uses the files that are in that Notebook. You can even select specific individual files that should be used for any of the tasks. This means it reduces the risk of hallucination because it limits the data used to execute the task, as well as giving users a way to take information and make it more engaging for stakeholders or any audience that you need to present to.

The other way this reduces hallucination is that each answer given references back to the file it came from. You can click the link to open the file and verify the answer.

Let's say you've been asked to do research on a topic that you're not as familiar with as you'd like to be. You can search for authoritative documents about that topic and then upload them to a Notebook and use the Notebook to generate summaries to help you become more informed before you go write the study.

You could even take that information and generate a couple of slides that you could put into a report as supporting data for your findings from primary research.

Literature reviews also get easier with Gemini Notebook. Upload a variety of reports and data sources, then see what Gemini Notebook does to create a summary of the information. Even better, if you’re doing the literature review with a hypothesis in mind, you can write that hypothesis and ask Gemini to pull any relevant data in what you’ve uploaded that supports and refutes that hypothesis!

Once you've created reports, it can be difficult to repackage it into things that can be more engaging. Now, given the caveat from last week's note on questions that you should ask before using either a business account or consumer account to process any information, you could use this to upload data from a report or upload the report itself and supplement with secondary data to create a more engaging way for your stakeholders to engage with the information from the report.

One of the things that's difficult to do is keep up with newsletters and articles that you come across or that people send you.

Notebooks are a great way for you to upload PDF versions of those articles and newsletters, and keep adding to it as your own personal knowledge base. Go back to it any time to get a refreshed audio or video summary on the latest news, or query it for the most recent trends from the items you’ve uploaded!

Now you have a knowledge base of your own that you can keep adding to, with sources you trust, and, when you query against it, knowing the answers are based on the sources you’ve uploaded and not being hallucinated by the AI reaching outside of the sources to try to generate an answer. I tried to get it to answer questions that I knew didn’t have answers in the sources, and it let me know the sources didn’t have an answer, but offered to do external research on the question.

As we covered last week, qualitative data can have information that makes it possible to trace it back to who gave the answer. So this becomes a place to be careful when using this for qualitative analysis. I'd recommend that if you are exploring this use case, you go back to the flowchart linked at the top of this article to review the first layer of questions you should be asking before deciding whether or not to use Gemini notebook for this particular use case.

I'd also recommend that you be very careful in reading the terms and conditions and the privacy policies associated with using this tool because Google is one that can flag a conversation for human review and, even if you delete the conversation, it remains flagged and can still be reviewed by a human.

But if the type of subscription you have allows for it, and the type of data that you are using allows for it, and the customer has approved it, this would absolutely be a tool to consider for qualitative analysis because of the links that it provides back to the data source and the fact that it bounds answers to the files that you have uploaded into the notebook.

ChatGPT and Claude each have Projects, which can be dedicated workspaces with files and conversation history attached. However, each needs to have project-level-memory turned on to keep the memory bound to the project, else conversations will pull on data from outside the project. Not helpful when you want to keep answers to a certain set of files.

Claude Projects does better. You can set the memory to be “separate” which puts the boundaries on conversations to be just that project, and you can tell it to only limit answers to the files within the project. The memory setting applies from the moment you set it, and there have been too many instances of Claude not following user-set instructions for me to think it’s guaranteed it won’t go to the web to find answers to questions even after telling it to only look for answers in the files attached to the project.

If you pay for a business Google Workspace account, you might already have access to Gemini Notebook among the apps you can access. It might be worth playing around in there to familiarize yourself with how it works and what it can do, just to know what it does.

An insights professional today needs to be aware of the various tools available, including what they would be best suited to support. Since Google Workspace is a go-to for many smaller organizations due to the affordability, knowing that you can access Gemini Notebook as part of your subscription can be helpful when you need to do a literature review or pull a lot of information together and get up to speed on a new topic area.