Every Friday someone spends 30-60 minutes clicking through boards to write "what happened this week". A scheduled flow reads the task tool, an LLM drafts the status in a fixed format, and the team lead reviews and sends it. The result is more consistent updates and a chore that disappears from the calendar.
Problem & value
- Team leads and project managers in small teams compile status by hand from Planner, Trello, Jira or a shared list, then paste it into mail or chat. The text is often late, uneven, or skipped.
- AI helps because the source data is already structured (status, assignee, dates, comments); the LLM only has to condense and phrase it.
- Honest limit: the status is only as good as the tickets. If nobody updates tasks, the AI summarises stale data. Third-party commentary on Jira's native AI makes the same point [3].
- No sourced time-saving figure is quoted here; measure your own baseline (minutes per week) before and after.
Variants
- Team status to colleagues (Teams/Slack channel) versus management summary (email, one page).
- Per-project reports for several boards merged into one portfolio mail.
- Sprint or release digest for software teams (Jira).
- Private use: a weekly summary of your own to-do lists (Todoist, Notion), local model.
- Out of scope: KPI dashboards and report narratives over BI data (see DATA-002).
How it works
- schedule trigger (Fri 14:00)
- fetch tasks changed in last 7 days
- filter and group (done / open / blocked / overdue)
- LLM prompt with fixed template
- draft to team lead
- approve
- post or send
- The deterministic steps (query, grouping, counting) happen in the flow, not in the LLM. Pass only grouped task titles, so numbers are never "invented" by the model.
- The prompt demands a fixed structure and forbids adding facts not in the input; include last week's status as a style example.
- Human in the loop: the draft goes to the team lead first (email or approval card). Fully automatic sending only after a few clean weeks.
- Existing examples: n8n templates for Jira weekly reports with Azure OpenAI and email [1] and for Trello board summaries with Gemini to Slack [2].
Data preparation
- Sources: Planner/To Do in Microsoft 365, Trello boards, Jira projects, Excel or SharePoint lists, sometimes Nextcloud Deck.
- Typical gaps: tasks without assignee or due date, status stored in free-text comments, inconsistent buckets/labels, boards that nobody updates, several tools per team.
- Steps (about 0.5-2 days): agree on 4-5 status values, clean up buckets and labels, make "blocked" a label, name one owner per board, and grant the flow a service account with read-only access.
- Pseudonymise if useful: report by task title and team, not by person, unless the team agrees otherwise (see People & change).
- See GUIDE-003 for a data readiness check.
Implementation paths
Which path for whom
- smb: B. Tools are usually mixed (Planner plus Excel, or Trello plus Jira), so a small n8n or Power Automate flow with an LLM step fits better than a single vendor feature. If everything lives in one tool with a native feature, A is enough.
- enterprise: B. Central automation platform plus the approved LLM endpoint; native A features suit single-tool teams.
- developer: C. A script or n8n self-hosted with a local model; this is cheap and keeps task data in-house.
Path A - Buy (off-the-shelf)
- Microsoft Planner/Copilot: Microsoft ships Power Automate templates for Planner digests (daily summary by bucket, outstanding-task digest) [4]; these can be switched to weekly. Microsoft's Planner Agent can generate status reports on demand (Reports tab) for shared plans and needs a Microsoft 365 Copilot licence to generate them; rollout was announced for late October 2026 (Message Center MC1506735). Scheduled automatic delivery is not documented, so build it with a flow; check your tenant.
- Atlassian Rovo for Jira: a weekly-progress-report prompt template for Rovo Chat (Confluence and Jira) [3]; Rovo's access to Jira change history has been reported as limited, so check current documentation, plan and availability.
- Marketplace apps for Jira status reports exist; check vendor hosting and AVV.
- Choose A when the whole team uses one tool and the native text is good enough.
Path B - Build on cloud APIs
- Orchestration: Power Automate (included in many M365 plans for standard connectors), n8n Cloud (EU hosting) or self-hosted n8n, Make (EU region).
- Model: Azure OpenAI or Mistral (EU) or an EU region of a hyperscaler; a small or mid-size model is enough, and cost per weekly report is typically cents.
- Integration: native nodes for Jira, Trello and Planner (or Microsoft Graph via HTTP request), delivery via Outlook, Teams or Slack.
- Effort: 1-3 days for the first flow including prompt tuning, 2-4 hours per additional team.
Path C - Self-host (open source)
- n8n self-hosted (Docker), Ollama with a 7-14B model (e.g. Mistral, Llama, Qwen class), delivery through your own mail server.
- Hardware: HW1 is enough for short summaries (slow but fine for a weekly job); HW2 is more comfortable. A server that already runs Whisper or Paperless-ngx can share the Ollama instance.
- Choose C when task titles contain customer names or trade secrets, or when the Baukasten effect applies (same n8n for AUTO-007, AUTO-009, AUTO-012). Ops effort: patching and backups, about 1-2 hours per month.
Path D - Custom training
Not needed. Fine-tuning for tone is not worth it; a good template and one example are sufficient.
State of the art (2026)
- Agentic features inside work-management tools (Rovo in Jira/Confluence, Planner Agent) now draft status updates natively, partly on demand rather than on a schedule [3][4]; cross-tool reports are still best done with a flow.
- Current LLMs summarise structured input reliably; failure modes are omitted blockers, over-positive tone, and fabricated detail when the input is thin. Counter with fixed grouping, "only use given items", and a visible task list under the text.
- Limits: tools cannot judge real progress; a task "in progress" for six weeks stays "in progress".
Shared resources & combinations
- Same workflow engine and LLM endpoint as AUTO-009 (feed digest), AUTO-024 (daily planning), AUTO-004 (chat catch-up), AUTO-028 (absence handover).
- Complements DATA-002 (BI report narratives) for numeric KPIs.
Effort, cost & KPIs
- A: licence only, hours to enable. B: 1-3 days build, LLM cost in cents per report, flow licence (Power Automate premium connectors or n8n subscription). C: 2-4 days incl. setup, hardware largely shared.
- KPIs: minutes saved per week and team, share of weeks with on-time status, edits needed by the team lead (aim: decreasing), reader feedback, number of overdue items found.
People & change
- Department lead owns the template and approves the draft; end users must agree what is reported and how task hygiene looks. Management defines who receives the report. IT sets up accounts, secrets and the AVV; DPO checks the purpose.
- Works council: depends. A report that names people and counts their completed or overdue tasks can be suited to monitor performance; per BAG case law, objective suitability is enough for co-determination under §87(1) no. 6 BetrVG. Report at team level, avoid per-person rankings, and inform the works council in good time (§90(1) no. 3); an AI expert may be called under §80(3). Only companies with an elected works council; public sector: Personalrat (BPersVG/state laws). See GUIDE-004.
- Data protection: task data with names is personal data; sign an AVV with every processor, a DPIA is normally not required for team-level summaries, but ask the DPO. See GUIDE-005.
- AI literacy: Art. 4 AI Act (since 27 Jul 2026) requires measures that support the development of staff AI literacy, without guaranteeing a level; a 30-minute briefing on reviewing AI drafts is enough here. See GUIDE-006.
- Communicate that the report supports the team and does not rate individuals; start with one pilot team for 4 weeks (GUIDE-008, GUIDE-009).
Risks, regulation & compliance
- EU AI Act: minimal risk; no high-risk use as long as the output is not used for employment decisions (that would touch Annex III). Art. 50 transparency duties (applicable since 2 Aug 2026) target chatbots and synthetic content; an internal text draft reviewed by a human is normally not affected, but marking AI-assisted text is good practice if it goes to external recipients.
- GDPR: purpose limitation (status reporting only, not performance review), EU region or self-hosting, AVV, data minimisation. See GUIDE-014 for deadlines.
- Risks: leakage of confidential project names to non-EU services, prompt injection via task comments written by third parties (GUIDE-013), silent errors when an API change returns nothing (add an "empty result" alert), creeping use as a control tool.
Getting started
- Pick one team and one tool; list the five status values and clean the board (half a day).
- Start from an existing template: the Power Automate Planner digest [4] or an n8n Jira/Trello template [1][2].
- Add the LLM step with a fixed output format and last week's status as example.
- Run in draft mode for four weeks; the team lead edits and sends.
- Agree with works council/DPO on team-level reporting, then roll out to further teams.
Sources
- n8n - Generate daily Jira summaries and weekly reports with Azure OpenAI and Gmail - https://n8n.io/workflows/9683-generate-daily-jira-summaries-and-weekly-reports-with-azure-openai-and-gmail/
- n8n - Summarize Trello board activity with Gemini and send updates to Slack - https://n8n.io/workflows/13873-summarize-trello-board-activity-with-gemini-and-send-updates-to-slack/
- Atlassian - Draft a weekly progress report with Rovo - https://www.atlassian.com/software/rovo/use-cases/draft-a-weekly-progress-report
- Microsoft Power Automate - Create a daily summary of Planner tasks by bucket - https://powerautomate.microsoft.com/en-US/templates/details/df5797658f954281b54bad8bf9cc8246/create-a-daily-summary-of-planner-tasks-by-bucket/