The answer
Mostly. Pulling numbers from each tool, pasting them into a sheet and working out week-on-week changes is almost fully automatable today. In our model about 75% of the reporting time can go to software. What stays human is agreeing what each metric means and explaining why a number moved, which is the part leaders actually read.
The tasks, and how much of each is automatable today
Six tasks make up a weekly KPI report. Exporting, cleaning and calculating are 85 to 100% automatable because they are pure data work. Chasing people for missing numbers is about 70% automatable with scheduled nudges. Writing commentary is partly automatable with AI drafts. Agreeing definitions and judging anomalies stays almost entirely with a person.
| Task | Min / week | Automatable |
|---|---|---|
| Export last week's numbers from each toolEvery major CRM, ad platform and accounting tool has an API. The 5% left is fixing a broken connection when a password or permission changes. | 120 | |
| Paste, clean and reconcile numbers in the spreadsheetRules handle date ranges, currency and duplicates. A person still looks when two systems disagree by more than a set margin. | 90 | |
| Calculate week-on-week changes and progress to targetPure arithmetic once the numbers are in one table. No reason for a person to do it. | 30 | |
| Chase team members for numbers that live outside any systemA scheduled Slack form collects them. The remaining time is chasing the person who ignores the form. | 30 | |
| Write commentary on what changed and whyAI drafts a factual summary of the biggest moves. The why, such as a lost client or a holiday week, needs someone who knows the business. | 45 | |
| Agree metric definitions and review anomaliesDeciding what counts as a lead or a sale is a management decision. Software can only flag that a number looks odd. | 30 | |
| Whole process | 345 |
Hours saved per week
For a 15-person service business pulling numbers from four tools, the Monday report takes about 6 hours of a manager's week by hand. Automating the tasks above returns about 4.5 hours a week, a little over 200 hours a year. The time left goes on reading the numbers and deciding what to do about them.
Modelled on: A 15-person service business where the operations manager builds a Monday KPI report from the CRM, two ad accounts, the accounting software and a shared spreadsheet.
- Numbers come from four systems, each needing a login, an export and a date filter.
- About 90 minutes a week goes on pasting, cleaning and reconciling numbers that disagree between tools.
- Two or three team members are chased each week for figures that live outside any system.
- Commentary is a short paragraph per area, written by the manager.
Tools that do the job
You need connectors that read each source system, a single place where numbers land, and a way to deliver the report. n8n, Make or Zapier pull from the CRM, ad accounts and accounting software into Google Sheets or Airtable. Looker Studio draws the charts. An AI agent drafts the commentary and Slack delivers it.
- n8n
- Runs the Monday schedule, calls each system's API and writes the numbers to one table.
- Google Sheets
- Holds the metric table with one tab of fixed definitions everyone can read.
- Looker Studio
- Draws the week-on-week charts from the sheet without anyone rebuilding them.
- AI agent
- Drafts a short commentary on the biggest changes for the manager to edit.
- Slack
- Collects off-system numbers with a form and delivers the finished report to the team channel.
- HubSpot
- Source of leads, bookings and deals, read through its reporting API.
A simple workflow
The workflow runs on a schedule, early Monday morning, so the report is waiting before the first meeting. It pulls last week's numbers from each system, writes them to one table with fixed definitions, flags anything outside its normal range, drafts a short commentary and posts the finished report to the team channel.
- TriggerRun at 06:00 every Monday for the previous Monday-to-Sunday week, in the business's own time zone.
- CollectPull leads, bookings, sales, ad spend, cash received and any form-submitted figures into one table.
- CheckCompare each number with its 8-week range. Flag anything outside it, and anything missing, for a person to look at.
- DraftCalculate changes against last week and target, then have AI write three to five plain sentences on the biggest moves.
- DeliverPost the charts and draft to the manager first. Once approved, publish to the team channel.
Keep these human
- Deciding what each metric means. A report is only automatable once 'lead', 'booked call' and 'sale' have one written definition.
- Explaining why a number moved and what the team should do next week.
Where it goes wrong
- Definition drift. Two tools count a 'lead' differently, and the automated report quietly shows a number nobody agrees with. Write definitions on a tab in the sheet and check them quarterly.
- Silent connector failure. An expired login returns zero, and the report shows a collapse that never happened. Treat zero or missing values as errors, not data.
- Nobody reads it. A report that posts itself can become wallpaper. Keep it to five numbers and one paragraph.
What most guides miss
Most guides compare dashboard tools. The real failure point is definitions. When a report moves from hand-built to automated, nobody is checking each number anymore, so a changed CRM field or a new lead source silently redefines the metric. A written definitions tab, plus range checks that stop the report, matters more than the tool.
The evidence
- Knowledge workers spend 58% of their day on 'work about work', including chasing status updates and searching for information.Asana, Anatomy of Work Global Index 2023 (2023)
- Collecting data and processing data have technical automation potentials of 64% and 69% respectively.McKinsey Global Institute, A future that works: Automation, employment, and productivity (2017)
- 62% of people say they struggle with too much time spent searching for information.Microsoft, Work Trend Index 2023: Will AI Fix Work? (2023)
“There is surely nothing quite so useless as doing with great efficiency what should not be done at all.”