Automating reports and notifications: up-to-date data without assembling spreadsheets by hand

A manual report is usually produced the same way. Someone exports data from the sales system and the accounting system, combines it in Excel, fixes the formulas, makes a chart and sends it by email. When the board has a question, the whole procedure starts again.

The cost is easy to calculate: for each report, multiply the preparation time by the number of editions a year and the hourly labour cost, then add up the results. Suppose the sales director spends 2 hours a week over 50 working weeks on a report for the board, and an accountant spends 6 hours a month on the financial report. That gives 100 hours plus 72 hours, or 172 hours a year. At PLN 120 an hour, that is about PLN 20,600.

On top of this come costs that are harder to quantify. A report assembled by hand is sometimes late, and each author calculates things slightly differently. The board then argues about whose figures are right instead of making decisions.

In many companies, reports build up over years and nobody reviews them. Before you start automating, make a simple list of all recurring reports. For each one, write down a few details, as in the points below.

A simple test for reports of uncertain value: pause one of them for two weeks and see whether anyone asks for it. If not, it is not worth automating.

Many information problems in a company come from using the wrong format. The board gets a daily email with data that changes once a week, and learns about an overdue receivable from the monthly report. The table below helps you match the format to the need.

A company using Microsoft 365 has three basic routes, compared in the table below. The choice depends on where the data comes from and whether the company already has Power BI.

Power BI is Microsoft’s reporting tool that pulls data from systems on a schedule by itself. You build a report once, and after that it simply refreshes. Power Automate lets you set up a flow that runs at a fixed time: fetch data from a spreadsheet or system, compose a message and send it by email or to a Teams channel.

In both cases, most of the work goes into the first connection to the data, not into the report itself. Systems with ready-made connectors are quick to connect. Older accounting or production software sometimes requires an export to a file or help from the vendor. Some connections in Power Automate, for example to a database, require a Power Automate Premium licence.

A subscription is a schedule on which Power BI emails a snapshot of a report with a link to the full version. It can be set to run hourly, daily, weekly, monthly or after every data refresh. One report can have up to 24 subscriptions with different recipients and times.

To send subscriptions to other people in the company, a Power BI Pro licence is enough, costing USD 14 per user per month according to Microsoft’s price list. A few features require more, however, and are worth knowing about before you buy:

A good notification meets three conditions. It concerns a situation that needs a response before the next report. It goes to one person who knows what to do. And it happens rarely enough that someone still reads it.

The table below shows examples for a mid-sized trading or manufacturing company. The thresholds are illustrative – set them based on your own data.

In Power BI, alerts are set on dashboard tiles of the card, gauge or KPI type (a value compared with a target). Power BI checks them after each data refresh and can pass the matter to Power Automate, which sends a Teams message to the right person. Notifications from ERP or CRM systems, or from SharePoint lists, are built directly in Power Automate.

The most common mistake is setting up too many notifications at the start. Every department wants to know about everything, so after a few weeks people mute the Teams channel and miss the one important signal.

A few rules help avoid this. Each notification has one recipient and a defined response. Less urgent matters are collected into a single morning digest instead of separate messages. Once a month, you check which notifications did not lead to any action and switch them off.

It is also worth separating channels. Notifications that require action go to a specific person, while “for information” updates go to a shared channel that nobody has to follow in real time.

The board often does not need thirty charts, just a few figures and two sentences of commentary. Copilot in Power BI can prepare a description of what has changed in the data and answer questions about the report asked in plain language.

However, it requires Microsoft Fabric F2 capacity or higher, or Power BI Premium P1 or higher. A Pro or Premium Per User licence on its own is not enough, which for many mid-sized companies means extra cost. A simpler route is often a short comment added by the report’s author to the automatically sent snapshot.

Someone should always read AI-generated commentary before it is sent to the board. AI can misread the cause of a change, for example treating a drop in sales as a trend when it actually results from a large delivery being moved.

The best candidate for your first automation is the report that is produced most often and read by the board. Usually this is the weekly sales report. The effect is visible from the first Monday, and along the way differences in data definitions come to light. The plan might look like this: