Catalog / Everyday Automation & AI Micro-Tools / AUTO-028

Absence handover & back-from-leave catch-up summaries

Before leave, AI drafts a handover note from open threads, tasks and files; after return, it summarises what happened in mail, chats and projects.

Quick win: proven, manageable effort, first results within weeks, low regulatory risk.

Parts needed to run it yourselfAgents & automationLanguage model

In small teams every holiday, sick leave or parental leave leaves a gap: open threads nobody picks up before leave, and hundreds of mails and chat messages waiting afterwards. The AI built into Microsoft 365 or Google Workspace can draft the handover note from mail, tasks and files and turn the backlog on return into a short briefing. The person stays the author and checks the result.

Problem & value

  • Today: the handover is a hurried mail on the last day. The deputy misses what was only in someone's head; on return the person scrolls through inbox and channels, often for a day.
  • AI value: a structured draft in minutes, nothing forgotten because the AI starts from the actual threads and tasks, and a "read this first" list on return. No independent benefit figures exist for handover use; measure it in your pilot.
  • Boundary: long-term knowledge capture belongs to KNOW-014, onboarding to AGENT-009.

Variants

  • Before leave (outbound handover): note for the deputy, plus an auto-reply that names the deputy.
  • After leave (inbound catch-up): summary of mail, chat channels, meeting recaps and changed files.
  • Private use: summarising a mail backlog after a holiday, see AUTO-002.

How it works

  1. absent person lists topics + deputy
  2. AI reads allowed sources (mail, calendar, chats, tasks, recent files)
  3. draft handover note (topic / status / next step / deadline / contact / link)
  4. absent person edits and approves
  5. deputy on return: sources of the absence period
  6. per-source summary
  7. "what needs me first" list with links to originals
  • Core component: a general LLM with access to the user's own data through the suite. Prompts demand a fixed structure and links to the original items.
  • Human-in-the-loop: the absent person approves the handover; on return the person opens the original for anything decision-relevant.

Data preparation

Which data: the person's own mailbox, calendar, chats, task lists and recent documents (Exchange Online or Gmail, Teams or Slack, Planner/Jira, SharePoint or Drive). No training data.

Typical gaps: topics tracked only in private notes, tasks not in any tool, local files, and too broad sharing rights.

Preparation steps: 1. Agree on one task place per team (1-2 hours). 2. Clean up oversharing before enabling an assistant (PREP-023), a few days for SharePoint/Teams. 3. Set a rule for what may go into which tool (PREP-024). 4. Provide a handover template as a saved prompt (AUTO-023).

Rough share of effort: low, 10-20 percent. See GUIDE-003.

Implementation paths

Which path for whom

  • SMB — Path A: the assistant in Microsoft 365 or Google Workspace is a mature product at SMB prices, the process is standard and no integration is needed.
  • Enterprise — Path A: a commodity office function; use the suite assistant already licensed. A shared AI platform is not needed for this alone.
  • Developer — Path C: for teams on self-hosted Nextcloud, Mattermost or Jira, n8n plus a local model keeps the mail and chat data inside.
  • Public — Path C: staff mail often contains citizen data, so sovereign hosting is preferred (for example openDesk or Nextcloud plus Ollama); the staff council is involved.
  • Microsoft 365 Copilot: Outlook offers a "Summary by Copilot" at the top of a mail thread, and Copilot in Teams chats and channels answers prompts such as "Summarize what I've missed" [1][2]. Prompt the channel by its exact name; thread summaries need a minimum amount of text [2].
  • Gemini in Gmail and Workspace: Gemini can summarise an email thread (eligible Workspace plan required) [7]. Channel prompts for Teams are described in [3].
  • EU aspects: Microsoft states Copilot is an EU Data Boundary service for EU customers, but LLM requests can be processed in other regions depending on setup, so check the tenant settings and the Anthropic-model exclusion note in the admin centre [4]. For Google, check which Data Regions tier covers Gemini prompts and responses before rollout [5]. Sign the AVV via the vendor terms.
  • Choose A when: the team already uses M365 or Workspace and licences are affordable. Without a Copilot or Gemini licence, copy-pasting into a free chatbot is not acceptable for customer or staff data (GUIDE-007).

Path B — Build on cloud APIs

  • A Power Automate or n8n flow collects the absence-period items (Microsoft Graph or Gmail API), sends them to an EU-hosted model (Mistral La Plateforme, Azure OpenAI in an EU region, IONOS AI Model Hub, STACKIT AI Model Serving) and posts the digest to a private chat or mail.
  • Choose B when: the data sits in several systems the suite assistant cannot see (DATEV, CRM, ticket tool). Effort: 2-5 days.

Path C — Self-host (open source)

  • n8n (self-hosted), Ollama with a 7-14B instruct model such as Mistral or Qwen, reading IMAP/Nextcloud/Mattermost.
  • Hardware: HW1 works slowly on CPU; HW2 is comfortable for several users. Ops effort: 1-2 days to set up. Choose C for very sensitive mail (law, health, HR).

Path D — Custom training

Not worthwhile. A prompt template covers the need.

State of the art (2026)

  • Suite assistants summarise threads with citations linking to the source mail, the key feature for checking [1].
  • Limits: summaries can drop a critical sentence, cannot see private notes, calls or paper, and read only what the account can access. Teams chat summaries default to a 30-day history [2], so long absences exceed it. The template forces deadline and link fields.

Shared resources & combinations

Effort, cost & KPIs

  • Path A: hours for a pilot, about 1 day with permission clean-up; cost is the per-user licence.
  • Path B: 2-5 days plus small API costs. Path C: 1-3 days plus a server.
  • KPIs: time to write a handover and clear the inbox (self-reported), missed-item incidents during leave, deputy satisfaction.

People & change

Involve early: the department lead (who defines the handover standard), the end users, IT (licences, permissions) and the DPO. Management decides the AI policy (GUIDE-007).

Works council: depends. Summarising someone's own mail is not the aim, but M365 Copilot and similar tools log user activity and under the Federal Labour Court's view, objective suitability for monitoring is enough for §87(1) no. 6 BetrVG. §90(1) no. 3 requires timely information about AI-supported work processes. A works agreement should state that the summary is for the user only and not used for evaluation. Public bodies: staff council (GUIDE-004).

Data protection: the summaries process colleagues' and customers' personal data, so an AVV (Art. 28 GDPR), the DPO and a legitimate-interest assessment are needed; a DPIA is normally not required for a personal summary function but check it if the tool also analyses behaviour (GUIDE-005). Deputies must not read the absent person's mailbox without a documented rule.

AI literacy: Art. 4 AI Act requires measures that support staff's AI literacy. A 60-minute session on checking summaries against sources is enough (GUIDE-006).

Communication and pilot: present it as relief on the last and first day. Start with 5-10 volunteers incl. one AI champion, run two leave cycles, then roll out (GUIDE-008, GUIDE-009).

Risks, regulation & compliance

  • EU AI Act: minimal risk; only AI literacy (Art. 4) applies. Do not extend it to performance rating, which could become high-risk (Annex III duties apply from 2 Dec 2027) [6].
  • GDPR: staff, customer and third-party data; check data location, prompt retention, AVV and employee information.
  • Confidentiality: professional secrecy (§203 StGB: lawyers, doctors, tax advisers) needs a confidentiality commitment, or Path C.
  • Security: prompt injection via incoming mail; keep the assistant read-only (GUIDE-013).

Getting started

  1. Pick a team and agree on a handover template (topic, status, next step, deadline, contact, link).
  2. Clean up the sharing rights of the shared drives (PREP-023) and confirm the assistant's data location and AVV.
  3. Save the prompt in your prompt library: "Create a handover note from my open threads from the last 30 days, use the template, add links".
  4. Run it for two leaves, compare with the person's checklist, count missed items.
  5. Add the return prompt for inbox and channels.

Sources

  1. Microsoft Support - Summarize an email thread with Copilot in Outlook - https://support.microsoft.com/en-us/outlook/copilot-pages/summarize-an-email-thread-with-copilot-in-outlook
  2. Microsoft Support - Use Copilot in Microsoft Teams chat and channels - https://support.microsoft.com/en-us/office/use-copilot-in-microsoft-teams-chat-and-channels-cccccca2-9dc8-49a9-ab76-b1a8ee21486c
  3. Microsoft Tech Community - Summarize your channel conversations with Copilot in Teams - https://techcommunity.microsoft.com/blog/microsoftteamsblog/prompt-like-a-pro-effectively-summarize-your-channel-conversations-with-microsof/4139059
  4. Microsoft Learn - Data, Privacy, and Security for Microsoft 365 Copilot - https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-privacy
  5. Google Workspace - Data covered by data regions - https://knowledge.workspace.google.com/admin/compliance/data-covered-by-data-regions
  6. Regulation (EU) 2024/1689 (AI Act) - https://eur-lex.europa.eu/eli/reg/2024/1689/oj
  7. Google Workspace Updates - Gemini summary cards in the Gmail app - https://workspaceupdates.googleblog.com/2025/05/gemini-summary-cards-gmail-app.html