How to Get a Weekly Business Summary Emailed to You by AI

Decide the five numbers you want every Monday, get them calculated in one spreadsheet tab, then have a scheduled automation read that tab, ask an AI step to write a short commentary and email it to you. Zapier or Make does this reliably. Assistant features such as ChatGPT scheduled tasks or Copilot scheduled prompts suit summaries of your inbox instead.

Let formulas do the arithmetic and let AI do the words. A model asked to total 300 raw rows will sometimes get it wrong, and it won't tell you. A model given "tickets closed 212, last week 187, four-week average 195" will reliably write a useful paragraph around those figures. Most of the effort in a good weekly summary goes into the numbers tab, not the prompt.

Follow me on Instagram@sagnikteaches

Decide what the email must answer

Start with questions, not data. What would you want to know on Monday morning if you'd been away all week? For most small businesses it's five things: how much work came in, how much went out, whether customers were served well, whether cash is flowing, and one thing that needs attention.

Connect on LinkedInSagnik Bhattacharya

For an illustrative 14-person IT support firm, those became:

Subscribe on YouTube@codingliquids

Resist adding more. A weekly email with 15 numbers gets skimmed; one with five gets read. If you're unsure which measures deserve a place, choosing KPIs worth tracking helps you narrow the list.

Three ways to get it built

The first route is the one to build for a numbers summary, and it's covered in detail below. The assistant routes are useful for a different job, a digest of your inbox and meetings, covered after it. Many owners end up with both: an automated numbers email on Monday and an assistant's inbox summary on Friday afternoon.

Build the weekly numbers tab first

Create a Google Sheet or Excel workbook with raw data tabs (one per source, pasted or appended automatically) and one tab called Weekly numbers. Each row of that tab is one metric, calculated by formula from the raw tabs:

The formulas are ordinary counts and sums over date ranges. With the week's start date in cell B1 and ticket dates in column C of a Tickets tab, for example:

Tickets opened this week:
=COUNTIFS(Tickets!C:C, ">=" & $B$1, Tickets!C:C, "<" & $B$1 + 7)

Tickets opened last week:
=COUNTIFS(Tickets!C:C, ">=" & $B$1 - 7, Tickets!C:C, "<" & $B$1)

Cash received this week (payments tab, date in A, amount in D):
=SUMIFS(Payments!D:D, Payments!A:A, ">=" & $B$1, Payments!A:A, "<" & $B$1 + 7)

If formulas aren't your thing, an AI assistant can write them: paste the column headers and a few sample rows and ask for a COUNTIFS or SUMIFS for each metric. Test each one against a number you know. Gemini in Sheets can also help build and explain formulas in place, as covered in practical uses of Gemini in Google Sheets.

Two cells matter more than they look. The Note column lets a person add context the data can't show ("two engineers off sick"), and the Data last updated cell lets the automation refuse to report stale numbers. How the raw tabs get filled depends on your apps: some can email a scheduled export, others can append rows through Zapier or Make as events happen. Stopping retyping data between apps covers the options.

Filling the raw tabs without blowing the budget

There are two ways to feed each raw tab, and they cost very different amounts. Event-driven appending, where a Zap adds a row every time a ticket closes, keeps the sheet always current but costs a task per event. At the IT support firm's volume of roughly 850 closed tickets a month, that's 850 tasks just to feed one tab, far more than the summary itself. A weekly export costs nothing in tasks: the helpdesk emails a CSV report of last week's tickets every Monday at 06:00, and a person (or a two-step Zap that saves the attachment to a folder) drops it into the Tickets tab.

The firm used exports for high-volume data (tickets, payments) and event-driven rows only for low-volume items (new client enquiries, about 20 a month). A good rule: if a source produces more than a few hundred events a month, export it weekly rather than streaming it row by row.

Route A, step by step in Zapier

1. Trigger: Schedule by Zapier, Every Week. Choose the day of the week and the time of day. The time uses your Zapier account's time zone setting, not the Zap's, so check it under your account settings first.
2. Google Sheets: Get Many Spreadsheet Rows (Advanced). Point it at the Weekly numbers tab. It can return up to 1,500 rows, far more than you need.
3. AI by Zapier. Map the rows into the prompt below. The Standard tier is enough; it counts as one task.
4. Gmail: Send Email (or Outlook's equivalent) to yourself and anyone else who should read it, with the AI's output as the body.

That's three tasks a week, about 13 a month, which fits easily inside any paid plan's allowance. It does need a paid plan, because the free plan only allows two-step Zaps and doesn't include AI by Zapier.

In Make, the same build is a scenario scheduled Weekly: Google Sheets Search Rows on the Weekly numbers tab, a Text aggregator to combine the rows into one block of text, a Make AI Toolkit Simple text prompt module, and Gmail Send an email. Expect a handful of credits per run.

A variation for event-driven data is Zapier's Digest by Zapier, whose Append Entry and Schedule Digest action collects items all week (each new large order, each escalated ticket) and releases them on a weekly schedule. It suits a "list of notable things" section alongside the numbers.

One formatting tip saves a confusing first test. If you map the sheet rows into the prompt raw, the AI may receive them as one long run of values with no labels. Build a single text block first, one line per metric in the form "Tickets opened: this week 231, last week 204, 4-week average 210, note: none", using a Formatter step in Zapier or the Text aggregator in Make. Clearly labelled lines are what stop the model mixing up columns.

The prompt that writes the summary

You write a short weekly summary email for the owner of an IT support
business. Below is the Weekly numbers table.

Rules:
- If "Data last updated" is more than 2 days before today's run,
  write only: "Weekly numbers not updated since [date]. Check the
  exports." and nothing else.
- Use ONLY the numbers in the table. Do not calculate new totals.
  You may describe changes as up, down or flat, and quote the figures.
- Call a change notable only if this week differs from the 4-week
  average by more than 20%.
- Never guess causes. Mention a cause only if it's in the Note column.
- Format: one-sentence headline, then 5 bullets (one per metric
  group), then "Needs attention:" with at most 2 items.
- Under 180 words. Plain British English.

Table:
[rows from the Weekly numbers tab]

An illustrative email from the table above:

A busy week, with response times slipping and overdue invoices up.

• Tickets: 231 opened and 212 closed, both above the 4-week averages (210 and 195).
• Response-target breaches: 9, against an average of 5. Note: two engineers off sick Tuesday to Wednesday.
• New client enquiries: 6, up from 3 last week.
• Cash received: $48,300, below last week's $61,900 and the $52,400 average.
• Invoices over 30 days overdue: $17,650, up from $12,100; the largest share is three invoices from one client.

Needs attention: overdue invoices (up 50% on the average); response breaches (nearly double the average).

What was fixed while testing: the first version, without the 20% rule, called a 4% dip in cash "a significant decline", which is how summaries start crying wolf. It also offered a cause for the breaches, "likely due to the higher ticket volume", which was plausible but wrong; the real reason was in the Note column. The "never guess causes" rule stopped that. Keep a handful of past weeks' tables and run each new prompt version on all of them before switching.

One summary, two readers

If the owner and the operations manager both receive the email, they probably want different emphasis. Rather than one compromise, add a second AI step to the same Zap with a different instruction ("focus on tickets, response times and workload by engineer; leave out cash") and send each version to its reader. It costs one more AI task and one more email task a week. Keep the numbers tab shared, so both versions quote the same figures and nobody argues about whose number is right.

Route B: an assistant's scheduled prompt for your inbox

Numbers are only half the week. The other half is what happened in email and meetings: client escalations, promises you made, invoices that arrived. That's where an assistant's scheduling feature fits, because it can read your connected mailbox.

ChatGPT scheduled tasks run an instruction once or on a repeat. With Gmail connected under Settings, then Apps, a task can read from it, and you choose push, email or both for task notifications under Settings, then Notifications. Paid plans allow exact delivery times; the free plan allows a recurring task at most once a day. Each plan caps how many tasks can be active at once. An illustrative Friday task:

Every Friday at 16:00: from my Gmail, summarise this week's emails
from clients (not suppliers or newsletters). List: (1) any complaint
or escalation, with the client name, (2) anything I promised to do,
with the date I said, (3) any email still waiting for my reply after
2 days. Maximum 12 bullets. If nothing fits a heading, say "none".

Microsoft 365 Copilot scheduled prompts do the same across Outlook email, calendar and Teams for Microsoft businesses. You hover over a prompt you've run, choose to schedule it, set when and how many times it runs, and can ask for an email notification when the response is ready. You can have up to 10, and a paid Microsoft 365 Copilot licence is required; the free Copilot Chat doesn't include them. Copilot Business lists at $21 per user a month on annual billing, with an $18 promotional price on annual plans until 31 December 2026.

Gemini scheduled actions allow up to 10 active at a time and deliver into the Gemini app with push notifications on mobile. Google notes that content is prepared in advance (within the hour before delivery on a Google AI plan, up to several hours before without one), so fast-changing data won't be current, and that they work best for daily summaries and wrap-ups.

The rule for all three: treat any numbers they give you as approximate. They're reading messages, not your systems of record, so "you received about 14 invoices" is a prompt to look, not a figure to report.

Check it for four weeks before you trust it

Run the summary alongside your usual routine for a month. Each week, compare every number in the email with the source system, and note anything the email said that you wouldn't have. Watch for four red flags:

• A number that isn't in the tab. The model calculated something. Tighten the "use only" rule.
• A reason that isn't in the notes. The model guessed. Strengthen "never guess causes".
• The same numbers two weeks running. An export didn't run. Your Data last updated rule should catch this; if it didn't, check the cell's formula.
• No email at all. The Zap or scenario was switched off or errored. Set up alerts and a heartbeat check as described in stopping automations breaking silently.

The IT support firm's Monday email, a month in

Before the automation, the illustrative firm's owner spent about 90 minutes on Sunday evenings pulling reports from the helpdesk, CRM and accounting software into a spreadsheet and reading them. Setting up the numbers tab, the formulas and the Zap took about four hours over two sessions, most of it on getting the helpdesk export to land in the sheet automatically. The firm already had a Zapier plan, so the running cost was about 13 tasks a month.

In the first four weeks, the email arrived at 07:00 every Monday. The owner's reading time fell to about five minutes, plus follow-ups. In week three, the helpdesk export failed over the weekend; instead of a confident summary of stale figures, the email said "Weekly numbers not updated since Monday 3rd. Check the exports." That single rule did more for trust in the email than any wording change. By week four, the operations manager had started adding notes on Friday afternoons, which made the commentary noticeably more useful.

Other versions worth building

The same pattern adapts easily. A print shop might track jobs completed, reprints, on-time dispatch percentage, quotes sent and quotes won. A translation agency might track words delivered, jobs delivered late, new clients and average margin per job. An events company in peak season might switch to a daily version covering enquiries, confirmed bookings, deposits received and supplier payments due.

Once the weekly email works, two extensions follow naturally. The same numbers tab, charted, becomes a dashboard, the subject of building a KPI dashboard without a data team. And the monthly equivalent, with profit and loss, balance sheet and variance commentary, is a bigger job covered in automating monthly management reports.

Weekly summary questions

Can the AI pull numbers straight from my accounting or helpdesk software?

Sometimes, through the automation tool's connectors, but it's rarely worth it for a first version. Many apps can email a scheduled export or already have a Zapier or Make integration that writes rows to a sheet. Getting the raw data into a spreadsheet first, then letting formulas calculate the numbers, is easier to check and keeps working when an app changes its connector.

Is it safe to send business figures through an AI step?

For most small businesses, weekly totals and trends are low-risk, but check the terms of the service processing them. Use a business plan or an API connection where data isn't used for model training by default, and keep customer names and personal details out of the numbers tab. The AI only needs totals and trends to write the commentary.

What day and time should the summary arrive?

Whenever you'll read it with ten quiet minutes. Many owners prefer Monday at 07:00, covering the previous Monday to Sunday, so the week's figures are complete. If your data comes from exports that someone runs on Monday morning, schedule the email for late morning instead, or it will report last week's stale numbers.

Further reads

• How to Turn Meeting Notes Into Tasks Automatically With AI — Turn the actions from your summary into tracked tasks.
• Copilot in Outlook: How to Triage Email and Draft Replies Faster — More ways Copilot can work through your inbox.
• How to Chase Late Payments With AI Reminders That Sound Human — Act on the overdue invoices your summary flags.
• How to Build a 13-Week Cash Flow Forecast With AI Help — Look forward as well as back each week.
• How to Manage Your Inbox With AI: Triage, Drafts, and Follow-Ups — Keep the inbox the summary arrives in under control.
• How to Add AI Steps to Zapier: Classify, Summarise, and Draft — Detailed setup for the AI step inside the Zap.
• Is It Worth Automating a Task You Only Do Once a Week? — The payback sum for weekly tasks, the factors beyond time that change the answer, and three worked cases: one yes, one partly, one no.
• AI Assistant vs Virtual Assistant: Which Should You Pay For? — A side-by-side comparison, a food truck owner's week sorted task by task, real costs of each and when a small business needs both.
• How to Track Competitors' Prices and Offers With AI — Set up page monitors, a newsletter inbox and an AI-built price log so you see rivals' price changes and offers weekly, without drowning in alerts.
• Best AI Business Intelligence Tools for Small Businesses (2026) — Ten reporting and analytics tools compared on what their AI really does, list prices, the tier each AI feature needs and which small businesses each suits.
• How to Speed Up Month-End Close in a Small Business With AI — A day-by-day close calendar, four AI prompts with sample outputs, and a farm shop that went from a nine-day close to four without skipping checks.
• Budget vs Actual: How to Explain Monthly Variances With AI — A materiality rule, price and volume splits, a context-notes template and an AI prompt, worked through a craft brewery's month line by line.
• Capacity Planning With AI: Know When Your Team Is Full — Find the role, day or machine that limits your next booking, with a practical weekly capacity calculation and checked AI scenarios.
• Opening and Closing Checklists: Build Them With AI in an Hour — Turn a walk-through of your premises into opening and closing checks that staff can follow, record and hand over between shifts.
• How to Use AI to Understand Your Profit and Loss Statement — Five steps to get a plain-English reading of your P&L from AI, with prompts, sample answers, a tour operator's year and the sums you must check yourself.
• AI Workflow Automation Examples: 20 Processes to Automate First — Twenty processes ranked by a 25-point score, from inbox sorting to credit packs, each with a worked example, rough costs and what to watch.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI help pages on scheduled tasks in ChatGPT; Gemini Apps Help on scheduled actions; Microsoft Support and Microsoft Learn on Copilot scheduled prompts; Zapier help articles on Schedule by Zapier, Google Sheets and Digest by Zapier; Make help pages on scheduling. Checked September 2026.