AI in business intelligence and Excel
In the BI track you built dashboards that answer “what happened?”. AI adds two questions: “why did it happen?” and “what will happen?”, and lets you ask in plain language instead of building every chart by hand.
| Feature | What it does | Watch out |
|---|---|---|
| Copilot in Power BI | Summarises a report, suggests visuals, writes and explains DAX measures, and answers questions about the data. | Needs Microsoft Fabric capacity (F2 or higher) and an admin setting. Review any DAX it writes. |
| Forecast on line charts | Add a forecast for the coming periods with a confidence band from the Analytics pane. | Assumes past patterns continue. |
| Anomaly detection | Marks unusual points in a time chart and suggests possible causes. | Unusual isn't always wrong: year-end promotions, say. |
| Key influencers and Decomposition tree | AI visuals that show what drives a metric most and break it down step by step. | Influence isn't causation. |
| Excel Forecast Sheet | From the Data tab pick a date column and a values column, and get a forecast and chart. FORECAST.ETS powers it. | Needs regular dates without big gaps. |
Prepare your data for AI. Copilot understands your model from table and column names. A column called amt_x2 means nothing to anyone, human or machine; Sales Amount (SAR) is clear. Give tables and measures clear names and descriptions, hide technical columns, and set relationships correctly. It's the same good design you learned with ERDs.
The simplest forecast you can build yourself is a moving average: next month = the average of the last 3 months. Simple, yet sometimes it beats complex models, and it makes a good baseline to compare any model against.
sales = [120, 128, 135, 131, 142, 150] window = 3 last = sales[-window:] forecast = sum(last) / window print("Last", window, "months:", last) print("Forecast:", round(forecast, 1))