Automating the Sales Reports

The month-end scene is familiar: the sales numbers get gathered by hand from three different places (the till, the Instagram orders, the bank statement), assembled in Excel, the report finishes at midnight; next month, all over again. Report automation is that ritual's abolition: let the data flow itself, let the table assemble itself, and let your work be reading and deciding. The good news: no corporate BI budget is needed for it.

This article builds the road in three stages: joining the data flows, the live panel and the AI commentary layer; at the end, the typical mistakes.

Stage 1: Join the data sources

The condition of the automatic report is automatic data; a flow must be built for each source: the sales channels (the till system/e-commerce platform export; most have automatic passing to a sheet or an API), the CRM data (the enquiry-to-sale funnel numbers), the ad accounts (the spend data; the report's "what did we spend on" side) and minimising the hand-sources (for what is still written by hand: a 30-second simple form on the phone; from the form to the sheet automatically). As the collection point, Google Sheets is the small business's ideal centre: free, it connects with everything, it processes with formulas. The binding tissue is the integration tools (Make/Zapier/n8n): the "write what comes from the till here" rules get built once.

Stage 2: The live panel; the report's new form

The tool choice by tier: Sheets' own charts (fully enough for the start), Looker Studio (free, it connects to Sheets, a shareable handsome panel; the small business's "BI") and, when the need grows, the specialised BI tools. One rule in the panel design: one screen, 8–10 numbers; the panel "showing everything" shows nothing. The report's delivery automates too: every morning at 9 the panel link/PDF drops into Telegram; the report turns from a thing "opened and looked at" into a thing "that arrives".

Stage 3: The AI commentary layer; from the number to the sentence

The new possibility: a layer that comments on the numbers. Its structure is simple: the sheet data + the prompt ("analyse this week's sales data: 3 main trends, 2 worries, 1 recommendation") → a ready text summary every Monday. Its value: a readable pulse for the team members with no time to look at numbers + catching the patterns that slip the eye ("the Wednesday sales have been weak 4 weeks straight"). Know its boundaries too: for calculation trust the numbers must sit ready in the sheet (let the AI not compute but comment), and the critical decisions must come from the panel, not the summary. This layer is the safest kind of the agent flow: it reads, it writes, it touches nothing.

The typical mistakes: the anti-list

• The report theatre: the 12-page automatic PDF nobody reads; 8 read numbers are worth more than 12 pages.
• The data cleaning skipped: the channels write in different formats (the date formats, the AZN/qepik), and the merge produces garbage; the input rules must be standardised in advance.
• The hand-points remain: "it's all automatic, only the Instagram sales I write myself" = the system still depends on a person; the last hand-point too must be closed with a form.
• The neglect: a channel changes (a new payment type), the flow breaks, nobody sees; the monthly "is the data correct" check (the reconciliation with the bank statement) is the system's health card.
• A tool collection instead of measuring discipline: the panel is for decisions; where the weekly 15-minute review ritual (the KPI rule) is missing, the prettiest panel is wall decoration.

Sources and further reading

Where to verify the source

The panel tool documentation:

• Looker Studio: the free panel builder

Continuing the topic

This line's neighbouring articles:

• The Sheets foundation
• The KPI choice
• The finance view
• The agent layer
• Other articles on this topic

The first step is one source: this week, pipe your biggest sales channel's data into a sheet automatically. Once one flow runs, adding the rest is easy; and the month-end midnight ritual goes into history's archive.