The Agents behind Google Gemini’s Data Analysis

Leon Nicholls
8 min readMay 23, 2024

Ever wish your data could analyze itself, revealing hidden trends, customer preferences, or the magic formula for skyrocketing sales? With Google Gemini, that wish just got much closer to reality. But it’s not just one hero at work; it’s a dynamic duo of agents who make the magic happen.

Let’s face it: Spreadsheets can be a mixed bag. They are packed with valuable insights yet often as cryptic as ancient hieroglyphs. Some of us may dabble in coding, but few are Python pros. And even then, who wants to spend countless hours wrangling data when you could focus on the bigger picture?

Good news! Google Gemini is here to change the game. Consider it your data analyst team minus the hefty consulting fees. Gemini’s secret weapons are its agents, a powerful pair working behind the scenes to unlock the secrets buried in your spreadsheets. Let’s take a closer look at how they make it all possible.

Note: This article spotlights techniques for the Google Gemini Advanced chatbot (a paid service). While these concepts also apply to the free version, we’ll focus on the enhanced capabilities offered by the Advanced subscription.

What’s the Big Deal?

Google recently supercharged Gemini’s data analysis capabilities with a major update, making it even faster and easier to uncover hidden gems within your spreadsheets. This breakthrough is powered by a dynamic duo: the large language model (LLM) at Gemini’s core and its trusty sidekick, the data analysis agent.

While the LLM excels at understanding your language and questions, the data analysis agent is the math whiz, tirelessly crunching numbers and uncovering patterns in your data. Together, they transform your messy data into beautiful insights — no coding expertise is required!

But wait, there’s more! The recent update streamlines your workflow, allowing you to upload spreadsheets directly from Google Drive or your device. Whether you’re working with Google Sheets, CSVs, or Excel files, Gemini has you covered. Go from data upload to insightful analysis in a snap.

This powerful combination of LLMs and agents allows Gemini to process, explore, and analyze your data faster than ever. It’s like giving your data a shot of espresso! The result? You’ll get those “aha!” moments and presentation-ready visuals in a fraction of the time, so you can finally say goodbye to those late-night spreadsheet marathons.

Note: While exploring the ins and outs of Google Gemini’s data analyst agent, it’s worth noting that Google still needs to spill all the technical beans. So, while we’re basing our insights on the latest industry buzz and what we’ve seen firsthand, consider this a sneak peek into the potential of this exciting new feature.

How the Agents Work

Let’s peek behind the curtain and meet the two agents that make Gemini’s data analysis magic happen:

  1. The Data Analyst: This agent is your strategic guide. It examines your uploaded spreadsheet, identifies patterns and relationships within the data, and decides the best action to answer your query. Think of it as the project manager figuring out what needs to be done.
  2. The Coding Agent: Once the Data Analyst has a plan, it calls upon its coding partner. This agent is a technical expert, proficient in Python and other programming languages. Following the Data Analyst’s instructions, it generates the necessary code to slice, dice, and transform your data.

These agents work together seamlessly. The Data Analyst understands your question and the data, while the Coding Agent executes the technical analysis. It’s a tag team that delivers the insights you need.

But what if your data could be more pristine? No worries! Gemini’s agents are persistent problem-solvers. They keep going if they encounter hiccups, like missing values or inconsistencies. Instead, they cleverly refine their code and try again until everything runs smoothly.

“Show Code” Sneak Peek: If you’re a curious cat (or a budding coder), you can even take a peek under the hood with the “Show Code” feature. This lets you see the different versions of the Python code Gemini generated, along with the output from each try. It’s a fantastic way to learn how Gemini thinks and pick up a few coding tricks.

The Power of Combining LLMs and Agents

What makes Google Gemini’s data analysis prowess tick? The answer lies in the dynamic duo of LLMs and agents. Think of them as a dream team, each with unique strengths that, when combined, unlock a whole new world of possibilities.

LLMs, like the one at the heart of Gemini, are language virtuosos. They understand your questions, generate human-like responses, and even summarize complex information. However, LLMs can sometimes fall short regarding tasks requiring specialized knowledge or number-crunching.

That’s where agents come in. These specialized sidekicks are designed to tackle specific tasks with laser focus and precision. In Gemini’s case, the data analysis agent is a math whiz, adept at identifying patterns, making calculations, and extracting meaningful insights from your data.

Gemini creates a team that can handle even the most complex data challenges by combining LLMs' language understanding with agents' analytical power. The LLM acts as your interpreter, understanding and formulating your questions in a way the agent can understand. The agent then dives deep into your data, performing the necessary calculations and analysis to uncover hidden trends, correlations, and insights.

This powerful partnership enhances Gemini’s capabilities and addresses some of LLMs' inherent limitations. For example, while LLMs might struggle with complex mathematical reasoning or detailed knowledge of specific domains, agents can fill those gaps, ensuring accurate and reliable results.

The result is an incredibly powerful tool that’s also remarkably easy to use. You don’t need to be a data scientist or a coding expert to harness Gemini's power. Simply ask your questions in plain language and let Gemini’s agents do the rest.

Real-World Example: Detecting Teia NFT Marketplace Shenanigans

Let’s take this out of the theoretical and into the real world. Imagine you’re a curious observer of the Teia NFT marketplace and suspect some fishy business might be going on. You’ve got a hunch that some folks might be manipulating prices, artificially inflating volumes, or even engaging in outright wash trading. How do you confirm your suspicions? Enter Google Gemini’s data analyst agent.

Here’s how you could use it to investigate:

  • Gather your data: You’ve downloaded a JSON file containing the sales metadata for the past three days.
  • Feed it to Gemini: You upload the file and give Gemini a prompt like this (this is a shortened version; see the real prompt):

“Analyze the following list of sales metadata in JSON array format about the last three days’ sales on the Teia NFT marketplace.

[Your JSON data here]

Given the sales transaction history in the JSON metadata, provide an analysis for only today, 17 May 2024, covering these topics:

* Wash trading: Identify patterns in which the same NFTs are repeatedly bought and sold between multiple wallets to increase prices.

* Volume spikes: Detect spikes above 100% in today’s trading volumes. Ignore the sales activity of NFTs that have only a single sale transaction today.

* Price anomalies: Detect over 100% of price deviations above today’s sales prices. Ignore the sales activity of NFTs that have only a single sale transaction today.

* Whale activity: Monitor significant buying or selling of NFTs by an individual account. Avoid reporting every transaction; only display the total sales amount. Avoid reporting total sales of a whale account that is below 10 tez.

* Unusual Trading Patterns: Find unusual transactions indicating collusion or other illicit activities.

Format your final report using Markdown headers and tables, where relevant. “

  • Let Gemini do its thing: Gemini’s data analyst agent will analyze your data and look for suspicious activity that matches your criteria. It will generate code to churn through the data, identifying patterns, flagging anomalies, and calculating relevant statistics.
  • Get your report: In a matter of moments, Gemini presents you with a neatly formatted report, highlighting any potential wash trading, volume spikes, price anomalies, whale activity, or unusual trading patterns it found. You now have concrete evidence to support (or disprove) your suspicions without writing a single line of code yourself.

This is just one example of how Google Gemini’s agent can be used to uncover insights and make sense of complex data. The possibilities are endless, and the best part is that you can be something other than a coding expert to take advantage of this powerful tool.

Note: The coded analysis is pretty basic for this example, but more detailed prompts allow the agent to conduct a much more advanced study.

A Glimpse into the Future

Hold onto your hats, folks, because the data analysis and coding agents are just the tip of the iceberg. Google Gemini’s future is bursting with potential, and a whole squad of specialized agents could be on the horizon, each ready to tackle unique challenges and transform how we work, learn, and create.

Imagine this:

  • The Scientific Research Agent: Your trusty lab assistant, sifting through mountains of research papers, pinpointing crucial data, and brainstorming new experiment designs. It is a game-changer for scientists working on everything from life-saving medicines to eco-friendly technologies.
  • The Financial Modeling Agent: Your personal Wall Street guru, analyzing market trends, building risk models, and offering investment advice tailored to your goals. Think of it as a robo-advisor but with Gemini’s language skills and nuance.
  • The Legal Research Agent: A legal eagle at your fingertips, digging through case law, statutes, and regulations to find the exact information you need. Lawyers, rejoice! This could mean more efficient case preparation and faster access to justice for everyone.
  • The Medical Diagnosis Agent: A second opinion in your pocket, analyzing your symptoms, medical history, and the latest research to help doctors make faster, more accurate diagnoses. Imagine a world where healthcare is more accessible and personalized, all thanks to Gemini and its agents.
  • The Creative Writing Agent: Your brainstorming buddy, sparking ideas, helping you flesh out characters, and even critiquing your writing style. Whether you’re a novelist, a screenwriter, or a social media guru, this agent could be your secret weapon for crafting compelling stories.

And these are just a few examples! As agent technology evolves, the possibilities are endless. We could see agents specializing in everything from architecture and engineering to music production and fashion design. The key is that these agents will address the areas where LLMs like Gemini might stumble — precise calculations, deep domain knowledge, and interactions with external systems.

With this army of specialized agents at its disposal, Gemini could become a genuinely versatile AI assistant, seamlessly integrating into our lives and empowering us to achieve more than ever before. The future is bright for Gemini and its agents, and we can’t wait to see the incredible innovations they’ll bring to the world.


So, there you have it! Google Gemini’s agents are your secret weapon for conquering data chaos and turning it into actionable insights. It’s like having a data scientist on speed dial, minus the fancy degree and the expensive hourly rate.

With Gemini, you can finally say goodbye to those endless hours of manual data wrangling and hello to a world of possibilities. Whether you’re a business owner, a researcher, or someone who loves to tinker with data, Gemini empowers you. It’s time to unleash the power of your data and see what stories it has to tell.

Check out my reading list of other Google Gemini articles.

This post was created with the help of AI writing tools, carefully reviewed, and polished by the human author.