When Spreadsheets Meet AI
1. When Spreadsheets Meet AI
So far, we've looked at how AI can streamline a consultant's research tasks. Now, let's shift gears to look at how AI can be used to analyze the data gathered during the research phase.2. Structured Data
In this video, we'll start with structured data, which is data collected into rows and columns. We'll look at unstructured data like user feedback and audio transcripts in the next video.3. Example: Electronics Sales Performance
Imagine you're consulting for a global consumer electronics company trying to optimize sales across different regions. You're analyzing sales performance data to identify top and bottom performing regions and products. You know from the research phase that overall sales have increased across all regions, but there may be some products whose sales are decreasing in a particular region. Investigating these products and regions will open up more sales growth for the company.4. Example: Electronics Sales Performance
First, we take the dataset compiled during the research phase. We're only showing a few rows here to get familiar with the data.5. Analyzing Step-by-Step
Conversational AI enables consultants to analyze structured data without deep technical skills. The important thing to keep in mind is to have a clear goal and build the analysis iteratively in small incremental steps. In practice, this means breaking the work down into clear goals, context, and limitations at each stage of the analysis.6. Example: Analyzing Sales Performance
First, we upload our file to Copilot and write a prompt containing our goal and any context about the dataset. We then ask Copilot to visualize the sales trends across each region over time in a line chart. From this first view, we can confirm that the total sales trend has increased across quarters for all four regions. But is this growth driven by all products or just a few? Let's break it down even further.7. Example: Analyzing Sales Performance
In the same thread, we ask it to visualize a multi-line chart, where each product is a line. Here we can also use natural language to filter the products based on a particular feature.8. Example: Analyzing Sales Performance
For example, if we're interested in per-region significance, we could ask the model to visualize only products that represent at least 15% of total regional sales. Or, if we're trend-focused, we could ask it to show only products whose quarterly sales grew at least 5% from Q1 to Q4. In this case, we see that the 15% threshold is too restrictive, so we can ask the model to suggest a more appropriate filter to make the visualization less cluttered.9. Example: Analyzing Sales Performance
In this case, the model suggested a 10% threshold, which worked much better with our data. By looking at the visualization, we can clearly see that sales of the Delta Watch are rapidly decreasing in the West region. That's a great insight!10. Example: Analyzing Sales Performance
To confirm what we observed in the visualization and identify any other products showing declining sales, we can ask the model to return the items that have decreased either overall or between consecutive quarters. Finally, we can see how the model identifies the correct products, generates a table, and provides the requested explanation and visualization.11. Summary
We've seen that AI tools can help us analyze data much faster than traditional spreadsheets, quickly detecting complex patterns, and letting consultants interact with data through natural language, making analysis more accessible and intuitive.12. Summary
However, these systems also have important limitations: their internal reasoning can be opaque, they may produce confident but incorrect answers, and they require careful handling of sensitive data for privacy and compliance.13. Let's practice!
Let's put this into practice!Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.