Lasting Principles and Use Cases for Google Gemini |

It can feel unproductive to learn an AI tool when the underlying model might become obsolete the next day. Therefore, this article focuses on principles, use cases, and tips for Google Gemini that are designed to last. Let's get started.

Unlocking a Hidden Setting

Diving right into a setting most people don't know about, you actually need to go into your Gmail—yes, Gmail—to get the most out of Google Gemini.

Go into Settings.
Under the General tab, use Command/Control + F to search for "smart feature."
You need to turn both of these on: Smart features and Google Workspace smart features.
Toggle them on and click Save.

You'll see how this works in a little bit.

Next, head over to the actual Gemini web app. Under Settings, go to Apps. It's a good practice to only keep the Google Workspace and YouTube app/extensions enabled to avoid wasting time triggering other apps by mistake if you don't use them.

Applying Expert Knowledge with Gemini Apps

Speaking of Gemini apps, let's start with a powerful use case. The process involves several steps:

Share a link to an online article or video where an expert shares practical tips relevant to you.
Ask Gemini to extract takeaways and frameworks from that content.
Ask Gemini to apply the learnings to your own work.

For example, using the @ command in Google Gemini, you can bring up the YouTube app. You could paste a link to a resource on viral storytelling and enter a prompt like: extract the top three takeaways and explain them like I'm five. This provides actionable tips. For instance, when telling a story, you should zoom into one tiny, exciting part because focusing on the small stuff makes a story feel real. It can even share a storytelling framework, such as T-A-E-D:

Thoughts
Actions
Emotions
Dialogue

Next, you can share a rough draft of a presentation from work and ask Gemini to apply the T-A-E-D storytelling framework, requesting over 4+ actionable recommendations. The reply can be incredible. For example, one recommendation might be: instead of starting directly with OKRs and summaries, begin with a core challenge or insight that drove the plan. Gemini can even provide an example of a powerful opening for the presentation.

As for the other Workspace apps, their agent capabilities are still developing. While it's technically possible for Gemini to create a calendar event, a Keep note, or search Gmail, it's often faster to do it manually.

Note: A key exception is using the mobile version of Google Gemini to create tasks with specific due dates using your voice, which is very efficient on the go. If you're using an Android phone, you can also ask Gemini to open specific settings, saving you from navigating through menus.

Chat Models vs. Reasoning Models: Knowing When to Use Each

An extremely important topic for all AI users is knowing when to use chat models versus reasoning models. The model numbers change too quickly to be a reliable guide. What's more important are the descriptions: one for everyday tasks and another that's best for complex tasks.

Diving into an example, a pinned chat in Gemini might use the "everyday tasks" (or Flash) model—a basic chat model. This could be for a recurring task like converting simplified Chinese characters into traditional Chinese characters. The task is simple, requires a fast result, and doesn't need the AI to think through complications or nuances.

Compare this to a situation where you receive an email in formal business Chinese and need to translate it for English-speaking colleagues. In this case, you would use the Pro model (a reasoning model) to analyze the email in detail, extract implications, and share nuances that English speakers might miss. Although it's much slower, it's worth letting Gemini take its time for a more thorough response.

The distinction can be broken down like this: - Chat models usually require several back-and-forths to reach a satisfactory answer. - Reasoning models are given a goal, and the model itself plans the steps, self-corrects if needed, and then provides a final response.

A good rule of thumb is to default to the Pro (reasoning) model, as many AI tasks are generally more complex. Only switch to the Flash (chat) model when the request is simple and you need a very quick answer.

Here’s a quick overview of the two model types for some major players: * Chat Models: * Gemini: Flash models * ChatGPT: GPT-[Number] models * Deepseek: V models * Reasoning Models: * Gemini: Pro models * ChatGPT: O-[Number] models * DeepSeek: R models

Some might wonder about the middle option, which often mentions "advanced reasoning." To massively oversimplify, this is a hybrid model mainly geared towards developers looking for a balance between speed and cost. For most users, it's best to ignore it. Default to the Pro model and switch to Flash for faster response times.

Creating Interactive Content from Documents

Speaking of the Pro model, when preparing a presentation on AI agents, one could ask Gemini to turn a dense AI agent white paper into an interactive game to test knowledge. The process is straightforward: upload the PDF to Google Gemini, select the Pro model, enable the canvas feature, and ask Gemini to turn the paper into an interactive game where questions start easy and get progressively harder.

Gemini will spend a minute or two writing the code, running it, and fixing its own errors. The end result can be mind-blowing: a fully functional game that looks good, complete with elements like a working progress bar.

To be clear, both ChatGPT and Claude have equivalents of the canvas feature, but Gemini currently tends to deliver a more complete result after just one prompt. This means you might need to follow up a few times with the other tools to get something Gemini can produce in the first attempt.

The Power of Workspace Integration

What makes Google Gemini unique isn't just its underlying model—the technology is evolving rapidly everywhere. Its key advantage is the deep integration within the Google Workspace ecosystem via the Gemini side panel, which is accessible after enabling the smart feature settings in Gmail.

For instance, let's say you receive an email asking you to be a speaker at an offsite. You want to say yes but are busy, need more context, and want to suggest a later time. Your reply won't be straightforward. In situations like this, it's much easier to open the Gemini side panel, use a text expander app to bring up an email reply prompt, input a bunch of unstructured thoughts, and let Gemini output a coherent, professional response that you can paste directly.

Although the side panel is a paid Gemini feature, for those in the Google Workspace ecosystem, it's often worth the cost for the decreased friction of accessing AI in the tools used every day.

For example, in Google Sheets, after agreeing to the workshop, you could ask Gemini to create a table to track registrations based on the invitation channel. Instead of creating columns and rows from scratch, this rough draft serves as a great starting point. You can then remove unnecessary columns and values, make a few adjustments, and it's ready.

While in Sheets, consider this: if you have collected a bunch of feedback for an app, instead of using a traditional formula, you can use the AI formula.

For example, to categorize feedback from column D into buckets like 'Price', 'UI/UX', 'Privacy', or 'Others', you can type the following formula: =AI("categorize this feedback under Price, UI/UX, Privacy, or Others", D1) After entering the formula, you can drag it down the column and click 'Generate and Insert'. Gemini will then categorize the feedback accordingly. It's quite powerful.

Similarly, you can translate text. =AI("translate this feedback into simplified Chinese", D1) Essentially, if you don't remember a specific formula, you can just describe the end result you want.

Moving over to Google Docs, let's say you're preparing team updates. You can open the Gemini side panel, use the @ symbol to find the Google Sheet with the feedback, and type in a prompt like: Received a lot of feedback. Write a short paragraph explaining what we're going to do. After it runs, with just one click, you can insert the output into your Google Doc. All of this is achieved without leaving the page.

Pro Tip: For lengthy Google Docs like meeting notes or product requirement documents, you can use the @ command to insert an AI summary block at the top. This provides an at-a-glance overview, and whenever new content is added, you can simply click the refresh button to get the latest summary.

And of course, we can't forget about Google Drive. Here, you can right-click on a project folder and ask for a summary of all the files within it. Then, you can ask Gemini to draft a detailed, clear, and concise report on the project to debrief stakeholders and senior leadership.

To quickly recap, Gemini's integration within Google Workspace is its biggest advantage, as it helps minimize context switching and fragmented workflows.

Leveraging the Massive Context Window

Speaking of advantages, many have heard about Gemini's massive context window, but what does that mean in practice? For example, when trying to add lengthy industry reports into another AI tool, it might quickly exceed its capacity. A 250-page document could be too large for some tools to process.

Gemini, on the other hand, can handle a much larger volume of information. Even with multiple large documents, it may not come close to its context window limit. This means for tasks requiring the AI to ingest a massive amount of data, like analyzing industry trends, Gemini currently has a significant edge.

A Note on Sensitivity

At this point, it's important to address what could be considered Google Gemini's biggest downside: its sensitivity. There are instances where a genuinely curious and objective question on a potentially sensitive topic might be blocked. For example, asking for an explanation of a historical figure's legacy might be refused multiple times because the topic is deemed too sensitive. In contrast, other AI tools might provide objective and unbiased answers to the same question.

To be clear, AI safety principles are incredibly important. However, if too many legitimate requests are blocked, users may default to other AI applications. This is a balancing act that all AI developers are currently navigating.