The Journal of a Journey in Gemini

Here is where we track the learning experiences and results of ongoing tests conducted using Google Gemini and leveraging custom Gems created to take on specific roles to support various use cases. Follow along on the journey of developing a team of GEMSTONES to assist with everything from data analysis to strategy planning and beyond.

When I began this project on creating custom Gems in Google Gemini, I started with two distinct personas. Each had a specific purpose, and together they helped me test how these Gems would not only respond to my prompting, but also how they would collaborate to create a powerful workflow. 

The overarching theme of this project is based on Marvel Avenger characters. I initially prompted Chat GPT to provide me with 6 specific Gem personas outlined within the prompt. For the sake of this discussion, we'll start off with the two that are relevant to this update: Vision (Data Analyst) and Black Widow (The Writer's Ally). 

The first Gem I built was Vision, the Data Analyst. Vision’s first task was to take a dataset housed in MS Excel, conduct an overall analysis and present an overview of its findings. For the first test, I provided Vision an Excel file containing product sales and return related data for a small retail business. The goal was to collect insights, highlight any recognizable trends, and generate a clear analysis of the overall health of the business. 

Vision delivered quickly, but not without a concerning discrepancy. Its initial summary contained a miscalculation of the total sales dollars. This turned out to be a good thing because it allowed for the opportunity to test how resilient and responsive the Vision was. With just one follow up prompt asking Vision to review the data again and re-run the analysis, it corrected its own error. The second output was accurate, comprehensive, and detailed. It not only aligned all the numbers correctly but also produced a sharp overview, including valuable callouts and recommendations ideas to increase sales, especially within already high performing product lines. 

The second Gem I created was Black Widow, The Editor. Unlike Vision, Black Widow’s role isn’t about crunching numbers but about refining narratives. Its purpose is to take raw text and shape it into a clean and compelling story. To test the synergy between Vision and Black Widow, I provided Black Widow the analysis Vision had generated. I asked it to review and edit the narrative where necessary, and prepare a summary of Vision's output. Black Widow responded with an Editor’s Note, acknowledging that the original narrative was strong, and explained that it would focus on condensing the content while preserving its intent. This was a major positive for Vision because this response proved its ability to generate an insightful analysis. 

The beauty of Black Widow lies in its flexibility. Each output is designed to provide three distinct versions:

1. A clear and professional version. 

2. A short, high-level executive summary that gets straight to the point. 

3. An engaging, audience-focused version designed to captivate. 

Depending on the audience, I could choose one of the three or even mix and match elements to draft a final summary incorporating the perspective and context that can only come from a human analyst, into the story being delivered. This flexibility makes Black Widow especially powerful. 

This test proved the connectivity between the two Gems: Vision produced the analysis and Black Widow refined it into multiple delivery formats, while maintaining the essence of the analysis. Together, they formed a seamless workflow: data analysis to narrative clarity. 

This first exercise was a success. It allowed me to see how errors could be identified and corrected quickly, and how a single prompt could turn a flawed initial review into a detailed and reliable analysis. It should go without saying, but it is important to note that the analysis was conducted prior to this test and Vision's output was validated by a human for accuracy. I don't believe that I will ever arrive at a level of trust in AI to the point where it doesn't require human oversight. The name of the game is continuous improvement. 

With these first two Gems, the foundation feels strong. There are still four more Gems to create and test, but if this result is any indication, the Journey with Gemini has a promising and productive road ahead. The process is iterative, flexible, and full of potential. And from here, the collective team will only grow more powerful as it is assembled.