Author Says Simple Prompts Build Better Copilot Excel Dashboards
- The author argues that building Excel dashboards with Microsoft Copilot works best when users give the artificial intelligence a broad goal and step aside, rather than engaging in...
- The first testing approach involved feeding Copilot a workbook of 5,000 movie-viewing records.
- The figures showed that 82.4% of entries were repeat viewings, viewing volume peaked in 2023, and completion rates had virtually no relationship to personal ratings.
The author argues that building Excel dashboards with Microsoft Copilot works best when users give the artificial intelligence a broad goal and step aside, rather than engaging in conversational back-and-forth or detailed instruction.
Analytical Findings Shape the Initial Layout
The first testing approach involved feeding Copilot a workbook of 5,000 movie-viewing records. The instructions directed the assistant to analyze the data, identify patterns, and design a dashboard based on those findings.
That dataset covered 878 unique movies. The figures showed that 82.4% of entries were repeat viewings, viewing volume peaked in 2023, and completion rates had virtually no relationship to personal ratings.
Copilot generated a functional layout featuring six summary cards, six slicers, and four charts. Yet a distinct drawback emerged: the final dashboard reflected whatever the AI found most striking during its initial scan. Repeat viewing dominated the layout, pushing potentially more useful metrics aside simply because the algorithm flagged it as the primary story worth investigating.
Web Chatbots Introduce Overwhelming Density
A second method utilized the web-based Copilot chatbot to plan a comprehensive dashboard before passing those recommendations into Excel. The resulting blueprint called for detailed viewing patterns, platform breakdowns, genre metrics, ratings, heat maps, scorecards, and additional calculations.
When built in Excel, the resulting dashboard contained eight charts, two tabular visualizations, five slicers, five summary cards, and a current-filter insights area. This version became overwhelmingly dense.
Users had to zoom out so far that text readability suffered. Maintaining the elaborate layout introduced unnecessary verification work, ultimately defeating the core purpose of a productivity dashboard.
Simple Prompts Grant Maximum AI Freedom
For the final test, the author stripped the instructions back to basics. The approach provided only the raw data and a simple request for a useful interactive dashboard without prescribing specific steps.
This strategy aligned with updated prompting guidance from OpenAI. OpenAI suggests giving AI a clear goal while letting the model determine its own route.
Because Microsoft Copilot integrates underlying OpenAI models, the simpler directive allowed the tool to function more efficiently without overcomplicating the output.