AI Prompt: Build a Handover Doc From Your Inbox

When you leave for parental leave, change roles, or exit a job, the hardest part of the handover is not the process documentation — it is what lives only in your inbox. A vendor who will renegotiate the moment they sense you are absent. A client expecting a deliverable you committed to in paragraph four of a 20-message chain. A relationship with someone three levels above your replacement who only trusts you personally. None of that ends up in a Notion page.

The prompt in this guide asks an AI to read your threads and produce a handover document organised around risk: what is outstanding, who owns each relationship, what the recurring calendar of obligations looks like, and — the part most templates skip — which conversations are most likely to go sideways in the next 90 days if nobody watches them. That last section is what makes a cover feel prepared rather than exposed.

This guide gives you the short answer first, then the full prompt, the platform differences, common failure modes, and a note on doing this automatically from inside your inbox rather than by copy-pasting threads into a chatbot.

What should an email handover document include?#

Before you run any prompt, know what output you are aiming for. An email-derived handover document should cover five areas that a standard process doc cannot see: the open threads, the relationship map, outstanding promises, recurring obligations, and a risk register.

Most handover guides list the recurring meetings and the current project status. They skip the informal commitments scattered across threads, the specific people who need careful handling, and the fragile negotiations that could break without context. Those are what the prompt below is designed to extract.

• Live threads with open action items: what you owe someone, what someone owes you, and what is waiting on a third party.
• Relationship map: who each key contact is, what matters to them, and the current status of the relationship in plain language.
• Outstanding promises: deliverables with dates you committed to, whether explicit in a thread or implied by a reply.
• Recurring obligations: the weekly update, the monthly review call, the quarterly reporting cycle — and who to loop in for each.
• Risk register: the two or three conversations that are fragile right now, and what would cause each one to go wrong.

How do you build a handover document from your inbox using AI?#

Gathering the right threads is the most time-consuming step. The search is not the bottleneck; deciding which threads are relevant is, because relevance is a judgment call the model cannot make until you have already made it. Work backwards from the last 60-90 days: your sent folder for commitments you made, your inbox for open threads with active vendors, clients, and stakeholders.

Once you have the threads, paste them oldest-first below the prompt, separated by === markers. Include your departure date so the model can flag time-sensitive items accurately. Run the prompt twice — once on your inbox threads and once on your sent folder for the same period — because the model cannot surface what you did not paste.

1. 1

   Search your sent folder for commitments

   In Gmail, use 'in:sent after:2026/01/01 ("I'll" OR "I will" OR "by Friday" OR "I'll send")' to surface messages where you made explicit promises. Gmail's full operator reference is at support.google.com/mail/answer/7190. In Outlook, try 'from:me received:2026-01-01..today' in the Sent Items folder and scan for action language. The goal is every thread where you told someone you would do something.

2. 2

   Search your inbox for open threads

   Look for threads where the last message is from someone else and contains an expected deliverable or a question you haven't answered. Prioritise active vendors, clients with upcoming milestones, and anyone you communicate with more than twice a month. Sort by last sender to find the threads still waiting on a reply from you.

3. 3

   Export each thread oldest-first as plain text

   Open each thread, expand the full conversation including any quoted history, and copy as plain text. Paste oldest message first — models reason more reliably when events arrive in the order they happened. Strip email signatures, legal footers, and repeated quoted blocks before pasting to save context without losing any facts.

4. 4

   Run the prompt and review the output

   Paste all threads below the prompt using === as a separator between threads. After the model returns the document, check every commitment against its source thread. A handover document is exactly the kind of high-stakes output where a hallucinated date or invented owner does real damage — verify specific deadlines, contact names, and amounts against the original messages.

5. 5

   Add the parking-brake section manually

   Append a short list of decisions that must not happen without a specific person's sign-off: contract changes, approval commitments, pricing concessions. The person covering you will not know what they do not know; this section is the safety net for the things that would be catastrophic to get wrong.

What does the handover document prompt look like?#

Use the prompt below in ChatGPT, Claude, Gemini, or any capable model. Adjust the role details to match your situation. Each thread goes below a === delimiter, oldest message first.

How does this differ across Gmail, Outlook, and other clients?#

What do you do when the output is too vague or wrong?#

A weak or inaccurate output is almost always a problem with the input or the prompt structure, not the model's capabilities. The failure modes below are the ones that come up most often and each has a direct fix.

• Output is too vague: Add the five section headers explicitly to the prompt as shown above. Models given named output sections produce far more consistently structured results than those given a single open-ended instruction like 'write a handover document.'
• Model invents a commitment or contact: Add 'use only facts in the threads; write Not stated rather than guess' as an explicit rule, and verify every specific date, dollar figure, and contact name against the source message before sharing the document.
• Thread is too long to paste: Summarise each long thread separately first — five bullets covering the current state, open items, last decision, next steps, and any sensitive context — then run the handover prompt on the collection of summaries.
• Important threads are missing from the output: Run the prompt on your sent folder separately. Your outgoing messages are where your commitments live; your inbox shows only what others sent you. Running both is not redundant — they produce different sections of the handover.
• Risk register is empty or too generic: Add 'flag any thread where a relationship or negotiation is at a sensitive stage, even if nothing looks alarming on the surface.' This prompts the model to surface items it would otherwise treat as routine closed threads.

A faster way: letting your inbox surface the threads for you#

The manual workflow works, but the gathering step is slow because deciding which threads are relevant requires reading them first. It is easy to miss the quiet negotiation sitting in a sub-folder, the client relationship where the last message was a mild complaint you meant to follow up on, or the vendor thread that went quiet because you were waiting on them and both sides forgot.

AI Emaily is an AI-native email client that runs semantic search across your entire inbox — you can ask 'which vendors have open commitments in the last 90 days?' and get a filtered view without writing a single search operator. The search reads meaning rather than substrings, so a thread where you wrote 'I'll get you the revised proposal by end of week' surfaces even when it contains none of your search keywords. Once the threads are identified, the AI assistant summarises each one and extracts the open items — the same work the prompt above asks a chatbot to do after you have already done the gathering by hand. We build AI Emaily. Try it on your real inbox with a 7-day free trial at aiemaily.com, or see plan details at aiemaily.com/pricing.