FlowsBook a callSign inCreate free accountSign up
Business and places data

How to automate a weekly report that people actually read

To automate a weekly report, decide what the week is compared with, who reads it and in what form, then let a schedule, a data step and an AI summary do the rest.

Part of the guide Location intelligence automation

By the xMap team · · 5 min read

To automate a weekly report, set a timetable for the day and hour the reader wants it, gather only what happened in the last seven days, compare it with the week before so the numbers mean something, and have an AI step write a short summary on top of the figures. Then send it in the form the reader already opens: an email for a manager, slides for a Monday meeting, a PDF for a client.

Most weekly reports that get automated are copies of a spreadsheet someone used to fill in by hand. They arrive on time and nobody opens them, because a list of everything is not a report. This article is about the decisions that make a weekly report worth reading, a worked example, the mistakes that make one ignored, and how to set it up as an AI employee. For the ideas behind automating any recurring job, see the workflow automation guide.

Four decisions to make first

  • When it lands. A report read in a Monday meeting must arrive before the meeting, in the reader's time zone. A report that plans the week belongs on Sunday evening or early Monday; one that reviews the week belongs on Friday afternoon.
  • What the week is compared with. Twelve road closures means nothing alone. Twelve, against five last week, most of them on the ring road is a finding.
  • What is left out. Decide the three or four things the reader acts on, and put everything else in an attachment or a link.
  • The form. An email body for quick reading, a one-page PDF to forward or file, or short slides with one idea each for a meeting screen.

What a weekly report should hold

A good weekly report is short and has the same shape every week, so the reader learns where to look. A reliable shape is a summary of three to five lines at the top, a handful of figures with last week's beside them, the few items that need action, and then the full list for anyone who wants it. Keep the order fixed: when the shape changes from week to week, readers stop trusting that they have seen everything.

The summary should be written from the figures, never alongside them. An AI that writes the summary must only word what the data says. Tell it so in plain words: use only the figures given; do not add numbers of your own. That one line stops the most damaging kind of error in an automated report, a confident sentence nobody can check.

A worked example

This example is made up. A regional logistics manager covering Singapore, Kuala Lumpur and Bangkok wants a road disruption report every Sunday at 07:00, so her planners can route around the worst areas on Monday. She starts from the template Weekly road disruption report.

  1. A Schedule runs every week on Sunday at 07:00, in Singapore's time zone.
  2. An Area of interest lists the three cities.
  3. Traffic incidents finds the closures, roadworks and other incidents in each city.
  4. The Place classifier puts each one in a group: road closure, roadworks, accident or other.
  5. An xMap LLM block set to work on the whole list writes one page: how many disruptions in each city, the worst areas and which routes to avoid.
  6. Email sends it to the planners with the report at the top and the list below, and a map shows every incident.

Each Sunday the planners get the same shape: the page of advice, the counts by city, and a map they can zoom into. Because traffic incidents are what is on the roads at the moment of the search, she reads the report as a snapshot of Sunday morning's planned works and closures, not as a total for the week, and the summary says so.

This week against last week

The comparison is the part people most often skip, because it needs last week's numbers kept somewhere. Two ways handle it for you. If the report is about what your other AI employees found, the Team findings block counts their results this week against the week before, by city and by kind: openings, closures, strong fits, companies hiring, road incidents. The template A weekly brief from your AI employees puts it before an AI step told to word the findings and never add a number, and a report step.

If the report counts new things, such as new businesses in a city, put Only new after the search. Each week's run then holds only what appeared since the last one, which is the week's change by definition.

Mistakes that get weekly reports ignored

  • Sending everything. Three hundred rows in an email body is a data dump. Keep the body to what needs action and the rest in the PDF or the map.
  • No comparison. Without last week beside it, a reader cannot tell a quiet week from a bad one.
  • A summary that invents. If the AI step is not told to use only the figures given, it may round, guess or add a trend. Check the first three reports line by line against the data.
  • The wrong hour. A Monday 09:00 report for a 09:00 meeting arrives during the meeting. Schedule it for the evening before or early morning in the reader's own time zone.
  • A shape that drifts. Ask the AI step for the same headings in the same order every week.

What lands, and where

Once deployed, the report runs as an AI employee with its own card on the Flows page. Each run's email lands at the hour you set, its map is saved under Results, and the run's history shows what it found and sent. If a run fails, the owner gets an email about it instead of a silent gap. To change the report later, write what you want in its Team inbox thread, such as add Jakarta, or edit a block. More in A written report about the whole list.

Questions

What day should a weekly report arrive?

When the reader acts on it: Sunday evening or early Monday for planning the week, Friday afternoon for reviewing it. Set the time in the reader's own time zone.

Can the report be a PDF or slides?

Yes. Make a report turns the results into a PDF with your logo, short 16:9 slides, or both, and emails them or saves them to a Google Drive folder.

How do I stop the AI summary inventing numbers?

Tell it to use only the figures given and not add numbers of its own, and check the first few reports against the data.

How does it compare with last week?

Team findings counts your AI employees' results this week against the week before. For new items, Only new keeps just what appeared since the last run.

Templates that do this

Read next

More on business and places data · Every article · Templates

Put it to work

An AI employee in xMap Flows does this on live data, on its own timetable. Start from a template or describe it in plain words.