This evidence log tests a safe first workflow: paste synthetic scattered work notes into a general AI assistant and ask for a manager-ready weekly update. No real Slack, email, calendar, Shopify, Etsy, accounting, support-desk, CRM, project-management, invoice, inventory, or live work connector was used.
In the current Northstar Candle Co. fixture, ChatGPT scored 4.67/5 and Duck.ai scored 4.55/5. Both produced separable, manager-ready drafts from the same pasted synthetic notes and preserved the key calculations. Neither tool should be trusted to send the update, change live records, or act in connected workplace systems without human approval.
Treat this as a prompt-only drafting benchmark, not a final recommendation for workplace automation. The practical reader workflow is: sanitize notes, paste only the relevant context, ask for a draft weekly update, then verify every fact and action boundary before sharing.
The 2026-07-15 API-backed Search Console refresh still makes weekly status / report-from-Excel the only click-bearing cluster, with 86 trailing-28-day impressions on the Excel/notes support bridge and one click on this evidence log. That is useful routing evidence, not enough for another thin page, so this map sends readers to the closest existing workflow.
Use the support bridge for “automated project status report,” “automation weekly status report template in Excel,” Excel-row, and recurring-manager-update searches. The 2026-07-15 GSC monitor still shows this as the only click-bearing cluster, so route readers to paste approved notes and rows, require source labels, recalculate metrics, and send manually after review.
Compare the report-generator evidence when the source packet is closer to “AI report generator from Excel,” “CSV report generator,” or a spreadsheet-plus-notes business report than a weekly manager update.
Use the spreadsheet assistant benchmark for formula explanations, missing-value checks, and the arithmetic mistakes to watch before drafting a status memo.
Route admin-professional readers to the paste-only-first stack before any Slack, email, CRM, calendar, spreadsheet, or project-management connector is tested.
Use the task-management guide when the status update depends on project-board tasks, owners, blockers, dependencies, or reminders. Keep it paste-only until a sandbox proves permissions, notifications, undo, and audit logs.
Scores cover the separable assistant answer only. Setup friction, prompt echo, UI cleanup, and pending routes are tracked separately in the evidence CSVs.
| Tool | Score | Last tested | What worked | What to review |
|---|---|---|---|---|
| ChatGPT | 4.67/5 | 2026-06-15 | Best current weekly-status draft: preserved exact dates, names, invoice IDs, stock counts, Shopify/Etsy separation, required calculations, and human-review boundaries. | Longer and more outline-like than a one-page memo; vendor-credit and support-macro approvals were clearer in the plan/checklist than in the manager-decision section. |
| Duck.ai | 4.55/5 | 2026-06-15 | No-login route worked and produced a separable manager-ready report with correct arithmetic, high-impact blockers, explicit Maya decisions, and a pasted-notes-only safety boundary. | Added a few decision-dependent operating suggestions and needed cleanup because raw page text included prompt echo, UI chrome, and model labels. |
A pending row is not a quality judgment. It means we have not yet saved a separable full answer through a safe no-login or project-safe route.
Pending safe baseline
Keep as no-score until a no-login or project-safe route produces a separable weekly-status answer.
Pending safe baseline
Keep as no-score until a separable full answer is captured rather than prompt echo or generic UI/search text.
Pending safe baseline
Test only if a safe no-login/project-safe route can produce a distinct answer rather than the prompt echo patterns seen in prior Copilot recon.
The fixture uses Northstar Candle Co., a fictional candle business. The messy notes include Shopify and Etsy snapshots, support backlog changes, warehouse backlog changes, overdue wholesale invoices, inventory decisions, mailer QC issues, vendor label delays, and unsafe campaign copy.
A strong answer should produce a plain-English manager update while preserving exact dates, amounts, order and invoice IDs, channel boundaries, uncertainty labels, and the human-review gate for any real action.
The prompt is intentionally connector-safe: it tells the assistant to use only pasted synthetic notes and never claim access to live apps, customer records, invoices, support tickets, ecommerce admins, or messaging systems.
content/fixtures/weekly-status-report-generator/messy-worklog-notes.md
Fictional Northstar Candle Co. work-log notes for a Friday update: Shopify/Etsy snapshots, support backlog, warehouse backlog, wholesale invoices, inventory, vendor QC, and campaign-copy risks.
content/fixtures/weekly-status-report-generator/weekly-status-report-paste-prompt.md
The exact connector-safe prompt asking for wins, risks/blockers, decisions needed, numbers to watch, next week plan, and a human-verification checklist.
content/fixtures/weekly-status-report-generator/expected-status-report-output.md
Ground-truth facts and scoring notes, including Shopify net revenue snapshots, order-count changes, backlog deltas, Hill & Vine stock impact, invoice IDs, and must-not-invent constraints.
content/fixtures/weekly-status-report-generator/weekly-status-report-test-protocol.md
The no-connector scoring and safety protocol: prompt/output separation, sanity prompts, screenshots, raw text capture, setup-friction labels, and no-score blockers.
Weekly updates can contain confidential revenue, customer, vendor, employee, support, and operations data. Start with sanitized excerpts and keep a human review step between AI output and any real communication or system change.
The current scores cover two prompt-only public web runs using a synthetic fixture. They do not test paid plans, enterprise privacy controls, file uploads, browser extensions, Slack, Gmail, Outlook, Shopify, Etsy, accounting systems, support desks, calendars, project tools, automations, memory, or real action-taking. New rows will be added only when raw output or setup-friction evidence is saved and scored with the same rubric discipline.