Last week's numbers, compared, every big move flagged.
Weekly ops brief is an AI agent that prepares the numbers for your weekly business review. It reads your weekly metrics table, compares last week with the week before and the four-week average, flags every metric that moved more than 10% and posts a short brief before the staff meeting. It works out every percentage itself.
- 08:00Schedule fired
- 08:00Read 9 metrics, 5 weeks
- 08:00Compared week on week and 4-week average
- 08:01Rechecked every percentage
- 08:01Posted brief · 4 flagged
Why does the weekly numbers review take so long?
Because someone rebuilds it every Monday: export the numbers, recalculate last week against the week before, decide what counts as a big move and write it up. The method changes with whoever does it, so a real jump in churn or ticket load sits in the sheet until the meeting.
Without it
- Monday morning spent recalculating last week in a spreadsheet.
- Percentages worked out differently each week.
- A big move nobody noticed until the meeting.
- A deck with every number and no signal.
With Weekly ops brief
- Every metric compared two ways.
- Only the moves over 10% get a line.
- The rest listed as steady, in one line.
- One brief, before the meeting.
What goes into a weekly business review?
Every metric you track, read two ways, so the meeting starts with what moved and spends no time on what didn't.
Week on week
Last week against the week before, for every metric in the table, as a signed percentage.
Against the average
Last week against the mean of the four weeks before it, so one odd week doesn't hide a trend.
Flagged moves
One line for each metric that moved more than 10% on either comparison, with both numbers.
Caveats
Missing values, fewer than four earlier weeks or a metric with no baseline, said plainly.
How does an AI weekly business review work?
Every run follows the same five steps. Your table is only read; the brief is written once, after every number is checked.
- 1
Finds the table
One row per week; it confirms the table before reading.
- 2
Reads the columns
Never assumes which column holds the week or a metric.
- 3
Reads the last five weeks
Last week, the week before, and the four weeks for the average.
- 4
Works out every change
Week on week and against the average, to one decimal, checked twice.
- 5
Posts the brief
The week's dates, one line per flagged metric, the rest as steady.
Slack
- Mondays at 8:00, before the staff meeting
- @weekly-ops-brief run last week's numbers
- Run when your export calls its webhook after the new week's row lands
From one sentence to Monday's brief.
Describe the job, let Arclyx draft the agent, test it once on your real table, create it, then read what lands in #ops-leads.
Paste the sentence. Arclyx picks the tools it needs and asks what it can't work out.
What should it do?
Every Monday at 8, compare last week's operating numbers with the week before and the 4-week average, flag moves over 10%, and post a short brief for the staff meeting to #ops-leads. The numbers are in our weekly metrics table.
- Tools it can use
- Slack
- Which column holds the week?
- week_start, one row per week
It writes the instructions, and you sign in to each tool once. Nothing runs until you say so.
- Reading your description
- Writing the instructions
- Picking the tools
Weekly ops brief
- Slack
Mondays, 8:00
Try it once as a dry run. It reads your real tools, and what it would post is shown to you, not posted.
- Finds the table
- Reads the columns
- Reads the last five weeks
- Posts the briefSlack · Post
Once it's created, the output lands where you said, with a source for every item.
5 vs 3 (+66.7% week on week); 4-week average 2.75 (+81.8%).
Moved more than 10% on both comparisons.
↗ example-a.com/weekly-metrics149 vs 126 (+18.3% week on week); 4-week average 126 (+18.3%).
Median first response rose too: 5.1h vs 4.0h (+27.5%).
↗ example-a.com/weekly-metrics9 vs 8 (+12.5% week on week); 4-week average 8 (+12.5%).
Over the 10% line on both comparisons.
↗ example-a.com/weekly-metricsNew customers 14, cash in $181,000, cash out $159,800, headcount 44, hiring pipeline 30.
Hiring pipeline is +9.1% on its average: close, not flagged.
↗ example-a.com/weekly-metrics
One brief, every lead.
The same Monday message answers a different question for each person at the table.
Operations leads
The meeting's numbers ready before Monday's coffee.
Founders
What moved last week, without opening a spreadsheet.
Finance
Cash in, cash out and overdue invoices, compared the same way weekly.
People and support leads
Headcount, hiring and ticket load in the same view.
How do you prepare a weekly business review with AI?
Keep your weekly numbers in one table, one row per week, and give an AI agent read access and a threshold. Each Monday it compares last week with the week before and the four-week average, flags the big moves and posts a brief. In Arclyx that's five steps, starting with one sentence.
“Every Monday at 8, compare last week's operating numbers with the week before and the 4-week average, flag moves over 10%, and post a short brief for the staff meeting to #ops-leads. The numbers are in our weekly metrics table.”
- 1
Describe the job
One sentence: which table, what counts as a big move and where the brief goes.
- 2
Upload your weekly table
One row per week with a date column. Add a row each week, or replace the file.
- 3
Pick its Slack channel
Such as #ops-leads, or the brief sent to you directly.
- 4
Watch a dry run
Optional. Read the brief on your real numbers before anything is posted.
- 5
Schedule it
Mondays at 8:00, or @mention it before the meeting.
Or start from one of these sentences
“On the 1st of each month, compare last month's numbers with the month before and the 3-month average, and flag moves over 15%.”
“Every Monday, compare last week's ticket volume and response times with the 4-week average and post the flags to #support-leads.”
“Every Friday, compare this week's cash in and cash out with the 4-week average and DM me anything that moved more than 10%.”
“Every other Monday, list every metric that moved more than 10% against its 4-week average, with both numbers, for the board update.”
AI weekly business review vs a spreadsheet.
A spreadsheet and conditional formatting each cover part of the job. Here is what each one does, and doesn't, do.
| A spreadsheet by hand | Sheets conditional formatting | Weekly ops brief | |
|---|---|---|---|
| Works out the changes | Formulas you maintain | No: it colours cells by your rules | Yes, both comparisons, every week |
| Flags big moves | If you look | Yes, cells that match a rule | Yes, over 10% either way |
| Writes the brief | You do | No | Yes, under 200 words |
| Says what's missing | Rarely | No | Yes, in its caveats |
| Where it lands | A sheet to open | A sheet to open | One Slack message before the meeting |
Source: Google Sheets help: use conditional formatting rules, checked 24 October 2026.
Is it safe to give an AI agent your business numbers?
It reads one table you upload and writes one message. It can't change your data. Every run is recorded and has a spend ceiling by default.
Every run is recorded
Every step, in order
Each run keeps a trace: the table it read, the query it ran and the brief it posted.
Every metric, accounted for
Flagged metrics get a line each, the rest are listed as steady, so nothing is silently dropped.
What each run cost
Tokens and cost are recorded per run and per agent, next to your plan's limits.
A spend ceiling, on by default
Each run has a dollar ceiling unless you remove it. A run stops before it goes over it, and tells you.
It can
- Read the weekly table you upload
- Post to one Slack channel you name, message you, or email you
It can't
- Change or add rows in your tables
- Fill in a missing number
- Read your CRM or finance system
- Email anyone but you
Weekly business reviews, answered.
What is a weekly business review?
A short weekly meeting where a team reads its key numbers for last week against recent weeks and decides what to look into. Weekly ops brief prepares the numbers part: every metric compared two ways, the big moves flagged, in one message before the meeting.
Where do the numbers come from?
A table you upload to Arclyx, one row per week, such as an export from your finance, support and HR tools. It reads only that table, and never fills in a missing value: a metric with no baseline is marked not comparable.
How does it decide what to flag?
It works out the change week on week and against the average of the four weeks before last. Any metric that moved more than 10% on either comparison is flagged, with both numbers shown. Change the threshold in its sentence.
Does it explain why a number moved?
No, unless the data says. Its instructions forbid guessing whether a move is good or bad, or why it happened, and any interpretation is labelled as one. The meeting decides what to look into.
How accurate are the numbers?
It computes every percentage from the rows it read, checks each one again before writing and shows them to one decimal. In our tests every number matched a separate calculation of the same table.
What if a week is missing?
It stops instead of guessing when the table is empty, the week order is unclear or the latest row looks stale. With fewer than four earlier weeks, it says how many the average uses.
Can it read HubSpot or our finance system directly?
Not in the version we tested: it reads one uploaded table. You can give an agent more tools in Arclyx, but this page claims only what we ran.
Do I need to write code?
No. Describe the brief in one sentence and upload your weekly table. Arclyx drafts the agent and picks its tools, and you can watch a dry run before you create it.
Agents that work well with it.
Build your weekly ops brief.
One sentence in, a working agent out. Join the beta and we'll set you up.