# AI weekly reports and prompt templates for a small office

An n8n template sends a weekly summary of Google Sheets rows through OpenAI to Gmail and lists 10 minutes of setup, but gives no accuracy figure. Anthropic and OpenAI agree a prompt should state the task, the format and examples. OpenAI is shutting down reusable prompt objects on 30 November 2026. Read 2026-11-06.

Short answer, from the sources below: a weekly report written by an AI model is a four-step job that a no-code tool can run on a timer: read the rows, send them to a model with a prompt, label the result with a date, and email it. The one template read for this page lists 10 minutes of setup and says nothing about how often the summary is wrong. The prompt is the part an office controls, and the two largest model vendors' guides agree on what goes in it: clear instructions, the output format, and a few examples. Search results for "AI weekly report generator" and "AI agent prompt template" are mostly vendors' own tools and lists of templates; this page reads a workflow template and two vendors' prompting guides instead.

## What a weekly-report workflow does

n8n, the workflow tool covered on [n8n for accounting](/research/n8n-for-accounting), publishes a template called "Send weekly AI summaries of Google Sheets data via Gmail using OpenAI". Its page lists five steps. It "runs automatically on a weekly schedule (every Monday at 9am by default)", "reads all rows from a Google Sheets spreadsheet", "aggregates the data and sends it to OpenAI to generate a concise summary report", "formats the report with a date-stamped subject line", and "emails the summary to your chosen recipient via Gmail."[1]

The setup list is seven items: set the schedule, connect Google Sheets and name the sheet, add OpenAI credentials, set the recipient, connect Gmail, create a "Report Log" sheet with the columns "date, subject, sent_at", and switch the workflow on. The page gives "Estimated setup time: 10 minutes."[1] It was last updated "8 months ago" when read.[1]

Three things the page does not give: the text of the prompt sent to OpenAI, any measure of how accurate the summaries are, and the cost of the model calls. Anyone using it would be sending every row of the sheet to OpenAI each week, which is the data-handling question covered on [OpenAI data terms and chatbot liability](/research/ai-privacy-and-chatbot-liability). How often summaries state something the data does not is the subject of [testing AI agents and hallucination rates](/research/testing-ai-agents-and-hallucinations).

## What a prompt template should contain

Anthropic's prompting guide for its Claude models gives this test: "Show your prompt to a colleague with minimal context on the task and ask them to follow it. If they'd be confused, Claude will be too."[2] It treats the model as a "brilliant but new employee who lacks context on your norms and workflows."[2] For a weekly report, that means saying who reads it, what to include and what to leave out, rather than "summarise this".

The guide's list, with the wording it uses:

- **Output format and constraints.** "Be specific about the output format and constraints you want."[2]
- **Reasons, not only rules.** "Providing context or motivation behind your instructions... can give Claude more information about your goals". Its example explains that a reply will be read aloud, so ellipses should not be used.[2]
- **Examples.** "A few well-crafted examples (known as few-shot or multishot prompting) improve accuracy and consistency", and they should be "relevant", "diverse" and wrapped in `<example>` tags.[2]
- **Separate parts.** XML tags "help Claude parse complex prompts unambiguously, especially when your prompt mixes instructions, context, examples, and variable inputs."[2]
- **Data first.** "Put long-form data at the top", above the question and instructions. The guide says putting the query at the end "can improve response quality by up to 30 percent in tests, especially with complex, multidocument inputs."[2] The guide does not say how many documents or what task those tests used.

OpenAI's guide is shorter. It says to "put overall tone or role guidance in the system message; keep task-specific details and examples in user messages", and to keep few-shot examples in "a concise YAML-style or bulleted block so your team can scan and update them."[3] It adds: "Run your prompt tests and evaluation cases every time you publish."[3]

Both guides describe a prompt as something to keep, test and change on purpose. Neither supplies a ready template for a weekly report.

## Where to keep the template and one change on 30 November

OpenAI's guide says to "treat prompts as application code", kept in "named modules" and reviewed like other changes.[3] It also says "OpenAI is deprecating reusable prompt objects in the API", that prompt creation "will be de-emphasized beginning June 3, 2026" (a date now past), and that "v1/prompts is scheduled to shut down on November 30, 2026."[3] The advice for new work is not to create reusable prompt objects and to keep each production prompt in a versioned file instead.[3] An office that stored its report prompt as an object in OpenAI's dashboard would need to move it by that date. A prompt written into a no-code tool's node, as in the n8n template, is not a reusable prompt object, though the template page does not say where its prompt lives.

## Not yet proven

- **Accuracy of AI-written weekly reports.** No source read reports a measured error rate for summaries of spreadsheet data.
- **The template's prompt.** The n8n page shows steps, not the prompt text.
- **A vendor-neutral template library.** The search results for prompt templates were vendors' own lists; none was read for claims.
- **Cost per report.** No page read gives the model cost of one weekly run.

If you run a weekly AI report in an office and can say what it gets wrong, I would like to compare notes: mackenzie@mackenziebowes.com.

## Sources

1. n8n, workflow template: Send weekly AI summaries of Google Sheets data via Gmail using OpenAI. Read in Rotunda 2026-11-06. <https://n8n.io/workflows/13404-send-weekly-ai-summaries-of-google-sheets-data-via-gmail-using-openai/>
2. Claude Platform Docs, Prompting best practices. Read in Rotunda 2026-11-06. <https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices>
3. OpenAI API docs, Prompting. Read in Rotunda 2026-11-06. <https://developers.openai.com/api/docs/guides/prompting>

---

Published 2026-11-06. Web page: https://www.mackenziebowes.com/research/ai-weekly-reports-and-prompt-templates
