Most handover docs are a status report dressed up as a transition: long lists of “what” with no “why”, no named owner per open loop, and zero record of the decisions the previous owner already considered and rejected. Six weeks later the new owner re-runs those same arguments from scratch. That gap is expensive: consultancy reviews of failed transitions blame poor handover and weak knowledge transfer for a large share of projects that never deliver their expected benefit. The 11 prompts below force the parts people skip: context, decision rationale, in-flight risks, and the tacit knowledge that only lives in the outgoing person’s head. Pair them with the SOP drafting prompts so the repeatable work stops depending on you.
TL;DR
- Paste raw inputs (status notes, owners, risks, links) and let the model structure them. The value of these prompts is the output schema, not the writing.
- Write every handover in the second person (“You own X; escalate Y to Z”) so it reads as instructions to your successor, not a diary of what you did.
- The decision log (prompt 4) is the single highest-leverage section: it stops your successor from re-opening settled questions.
- Run the prompts inside a Claude Project, a ChatGPT Project, or a Gemini Gem so all your transition files share one context and you can iterate without re-pasting.
- Start two weeks before your last day, not the morning of. A handover written under pressure ships unread.
Best for
- Leaving a project
- Going on parental or medical leave
- Role transitions and promotions
- Vacation coverage
- M&A integration
Which model to run these in (as of June 2026)
These are document-synthesis tasks: you dump messy context, the model imposes structure. Long context and faithful summarization matter more than raw reasoning, so any current flagship works. Pick by how much material you paste.
| Tool (June 2026) | Context window | Best when | Tier |
|---|---|---|---|
| Claude Opus 4.7 / Sonnet 4.6 | 1M tokens | Long, nuanced handovers; faithful to source notes | Free (limited Sonnet) / Pro $20 |
| Gemini 3.1 Pro | 1M tokens | Pasting months of meeting notes or whole drives | Google AI Pro $19.99 |
| ChatGPT GPT-5.5 | ~320 pages in-app on Plus (full 1M only on $200 Pro) | Quick single-doc handovers; widest availability | Free / Plus $20 |
For a multi-document transition (status decks, contracts, incident logs), put the files in a workspace that persists context: Claude Projects, ChatGPT Projects, or a Gemini Gem. You write the instruction once and every follow-up question sees the same files.
A note on placeholders: in each prompt, replace the bracketed [like_this] tokens with your real values before sending. The original templates below use curly braces inside the code block so you can spot every fill-in at a glance.
1. One-page project handover
Draft a 1-page handover for project "{project}". Inputs: {paste status, owners, risks, links}. Output: project purpose, current status, in-flight risks, top 3 contacts, top 5 decisions made (with rationale). Write in the second person, addressed to the next owner.2. Role-handover document
I am handing my role to {name}. Help me write a role-handover doc. Output: responsibilities, recurring meetings (with purpose), recurring deliverables, stakeholders, open commitments, and "what I would do differently next quarter".3. Vacation-coverage handover
I will be out {dates}. Draft a coverage handover for {colleague}. Output: 3 things they must own, 5 things they can decline, who to escalate to for each category, and the 2 risks I worry about most.4. Decision log
Draft a decision log for project "{project}". Inputs: {paste recent decisions}. Format each row: decision, date, owner, rationale, what we considered and rejected, and what would trigger a revisit.5. “What only I know” knowledge dump
List the things only I know about {area}, broken into: technical context, political context, customer context, vendor relationships. Format each as a 2-line entry. Aim for 15 entries.6. Stakeholder map
Draft a stakeholder map for "{project / function}". For each stakeholder: name, role, their interest, their concern, the cadence I use with them, and the 1 thing not to do.7. Open-loops list
List the open loops in my current role that the next owner will inherit. For each: what is open, who is waiting on whom, what unblocks it, and what happens if it stays open another month. Name a recommended next owner per loop.8. Customer / account handover
I am handing account {customer} to {new owner}. Draft a 1-page handover: account history, contract status, last 5 interactions, open commitments, sensitivities, expansion opportunities, and watch-outs.9. Engineering on-call handover
Draft an end-of-shift on-call handover. Active alerts: {paste}. In-flight investigations: {paste}. Output: open incidents (with severity), pending decisions, who I left waiting on me, and runbooks I exercised this shift.10. M&A integration handover
We acquired {company}. I led integration for {area}. Draft a handover for the new owner. Output: state of integration, decisions already made, decisions deferred, cultural watch-outs, and regulatory commitments.11. Pre-resignation handover
I am resigning in {weeks}. Draft a thorough handover for my manager. Output: projects I will close before leaving, projects I cannot close, recommended next owners, training the team needs, and what I learned that the team should keep.How to make the output trustworthy
The model only knows what you paste. Two habits separate a handover that holds up from one that quietly invents details:
- Feed primary sources, not memory. Paste the actual status doc, the real contract dates, the verbatim incident timeline. If you summarize from memory first, the model amplifies your gaps.
- Verify every name, date, and number by hand. Current models will confidently fill a blank rather than flag it. Treat the draft as a first pass that you fact-check, not a finished document.
Then close the loop with a person: schedule one overlap call where your successor executes two or three real tasks while you are still reachable. That single rehearsal catches more missing context than any prompt.
Common mistakes
- Listing “what” without “why” so the next owner re-runs the rejected options from scratch.
- No decision log, so reversals look like progress to people who weren’t in the room.
- Skipping the “what only I know” tacit context: vendor quirks, internal politics, who hates whom.
- Open loops with no named next owner. Every loop without an owner becomes nobody’s.
- Handing over too late to be useful: written the morning of your last day, never read.
- Treating the handover as one-way, with no scheduled overlap call for the new owner to ask follow-ups.
FAQ
Which AI tool is best for writing a handover document? Any current flagship handles this well, because the task is structured summarization rather than hard reasoning. As of June 2026, Claude Sonnet 4.6 and Gemini 3.1 Pro both carry a 1M-token context window, which matters if you paste large incident logs or months of notes. ChatGPT on Plus is fine for shorter single-doc handovers. Choose by how much raw material you need to feed it.
How do I stop the AI from inventing details I didn’t give it?
Paste primary sources rather than summarizing from memory, and tell the model to mark anything it is unsure of as [CONFIRM] instead of guessing. Then read the draft against your source docs and verify every name, date, and dollar figure yourself. The prompt structures your knowledge; it does not replace your review.
When should I start writing the handover? At least two weeks before the transition, and ideally you maintain a running decision log from day one so the handover assembles itself instead of being written under deadline. The worst handovers are produced on the last morning and never read.
Can I run these without uploading anything confidential to a public chatbot? Yes. Strip names and figures, run the prompt on the structure, then fill the specifics back in offline. For sensitive transitions, use your employer’s approved enterprise tier (which excludes your data from training) or a local model, and keep contracts and customer data out of consumer chat.
What single section matters most? The decision log (prompt 4). Most wasted re-work after a handover comes from successors reopening questions the team already settled. Recording the rationale and the rejected alternatives is what stops that.