AI Prompts to Catch Up on Email After Time Off

Do not start at the top and work down. Newest-first is how your inbox displays, not how a backlog should be read: message 1 references a decision that message 12 overturned, and message 19 cancels the thread. For an email catch-up after two weeks away, read recency-weighted instead. Collapse every conversation to its final state, find what was decided without you, find what is now late, then answer only the threads that need you.

The prompts below run in any capable AI assistant, including ChatGPT, Claude, Gemini and Copilot, before you touch a single reply. Each one tells the model to report where a thread stands now, not to retell its history. A 20-message negotiation that ended on message 17 takes one line when the model gives you message 17 instead of the whole chain. For a to-do list rather than a summary, there is an optional fifth prompt at the end of the list.

1. 1

   Collapse each thread to its current state

   Where does every conversation stand right now, with superseded positions dropped.

2. 2

   Find what was decided without you

   A list of settled decisions only, not everything that was discussed.

3. 3

   Surface what is now past due

   Every deadline and commitment checked against the day you came back.

4. 4

   Draft short re-entry replies

   One brief reply per thread that needs you, with no apology for being away.

5. 5

   Optional: turn the backlog into a to-do list

   Every action you owe, with its thread, due date and who is waiting.

Before you start: isolate the leave period#

Two minutes of setup saves thirty minutes of confused output. Give the model only the mail from your absence, oldest first, with the noise stripped out.

In Gmail, the search operators after:YYYY/MM/DD before:YYYY/MM/DD return exactly what arrived in a date range. Google's own search operators page has the full syntax, and our guide to using Gmail search operators covers the combinations that matter here. In Outlook, use the search filters to limit results by received date. Classic Outlook has an Advanced Find dialog for this, while new Outlook and Outlook on the web put date options in the search filter, and the exact labels change between versions, so check the one you run.

Most clients list threads newest-first, so reverse the order before you paste. Models follow a conversation most reliably in chronological order, and a thread pasted newest-first can make the model report the opening position as the outcome. Strip signatures, legal disclaimers, unsubscribe footers and the quoted history repeated under every reply, because none of it carries a fact you need. Then paste by thread, or in batches with a clear separator between threads. A model summarizing 200 mixed threads averages across them and buries the two that need you.

Prompt 1: Email catch-up, collapse every thread to its current state#

This is the core catch-up prompt. It asks the model to report only where a thread stands now and to discard everything that was revised or superseded along the way.

The phrase 'current state' does the work. Without it, a model recaps the whole conversation and gives equal weight to a position from day one that the group overturned on day four. For a thread with eight position changes, the difference is a 400-word narrative or a two-sentence update you can act on. Run it on every thread before you decide whether it needs a reply. If the question you are really asking is which threads need a reply at all, use the triage prompts for your everyday inbox, which are built for exactly that sort.

Prompt 2: Find what was decided without you#

A fortnight away means decisions moved forward without you. Some need action, some need acknowledgement, and a few may deserve pushback. To judge which, you need a list of what was settled, not what was discussed.

This prompt scans a batch of threads for decisions: a vendor was chosen, a scope was trimmed, a deadline moved, a budget was approved. The key instruction separates decided from discussed. Discussion that reached no conclusion is noise on a list meant to drive action, so the prompt tells the model to skip those threads.

Prompt 3: Surface what is now past due#

Finding what is late comes before everything else. A deadline missed during leave can still be salvaged. One you discover a week later usually cannot. This prompt scans for dates and commitments and checks each against your return date.

You tell the model when you came back, and it classifies every date as past due, due today, due this week or upcoming. The URGENT flag on past-due items changes how you spend the first hour of your first day: anything flagged needs a reply or an explanation before you move to anything else.

Prompt 4: Draft batch re-entry replies#

By now you know what the inbox contains and what is urgent. Next you have to reappear in it. A re-entry reply has one rule: address what was decided or requested, and do not apologise for being away. An apology moves the energy from action to sentiment and invites a reassuring reply that delays the actual business. Say you are back, confirm what you understand from the thread, and state your next step.

This prompt drafts a reply for each thread that needs one, in a single pass. The 60-word limit is deliberate: a 200-word re-entry reply suggests you did not find the essential point. If a thread needs nothing from you yet, the prompt skips it and says so, which is useful information too. Read every draft before it goes anywhere, because the model only knows what you pasted.

Prompt 5 (optional): Turn the backlog into a to-do list#

Many people searching for AI prompts to catch up on emails and to-dos want the output to be a task list, not a set of summaries. This prompt runs after Prompt 1 and pulls every action you personally owe out of the threads, with the date and the person waiting on you.

Keep the same guard as Prompt 3: dates come only from the thread. If one thread holds a long list of commitments from several people, run the dedicated prompts for how to pull action items out of a thread on it separately.

How do these prompts work on ChatGPT, Claude, Gemini, and Copilot?#

All four accept pasted email threads and can run these prompts. The practical differences are how much each takes in one paste, how closely it follows strict format rules, and whether it can read your mailbox without pasting at all. Product names and plan requirements change often, so check each vendor's own help page before relying on a row.

Can Copilot in Outlook catch me up on emails I missed?#

Yes, if your account and licence qualify. Microsoft's Copilot Chat catch-up page lists a prompt for exactly this: ask it to summarize messages from a colleague over the past two weeks and to call out deliverables, due dates and action items. Microsoft states that for Copilot to access work data you need a Microsoft Copilot add-on licence. A thread summary button is a separate route: open a conversation and choose Summary by Copilot, or Summarize, at the top of the thread.

Two limits are worth knowing. Copilot in Outlook supports work or school accounts and Microsoft accounts on outlook.com, hotmail.com, live.com and msn.com addresses, and Microsoft says other mailboxes opened in Outlook do not get its Copilot features. Microsoft's examples are per thread or per colleague, so the four prompts above remain useful alongside it: they give you a decided-without-you list and a past-due check across whatever batch you paste.

What to do when the prompts do not work#

Three problems come up most often on a real return-from-leave backlog.

The first is incomplete context. The summary misses a key thread or skips the most important decision. This almost always means you did not paste that thread, or it was buried in a batch the model weighted equally. Rerun the relevant prompt on the missing thread alone. Batches are efficient, single threads are precise, so use both.

The second is a hallucinated date or owner. The model invents a deadline or assigns a commitment to the wrong person. Add one line to any prompt that extracts dates or owners: 'Use only facts stated in the pasted text. Write not stated rather than guessing an owner or a date.' Then verify every flagged item against the original message. A confident, well-formatted output can still be wrong, and on this list the errors have real consequences.

The third is a thread whose current state is genuinely ambiguous, because the conversation is still live or the last message was not definitive. Add: 'If the current state is unclear, say so and tell me what question would resolve it.' An acknowledged ambiguity beats a plausible wrong answer.

A faster way to handle this inside your inbox#

The prompts work, and they take effort to run: isolating the date range, stripping boilerplate, batching pastes, checking output. We build AI Emaily, an AI-native email client for Gmail, Outlook and Microsoft 365, iCloud, Yahoo, Fastmail and other IMAP accounts, and it takes a different route to the same problem. Its Brief is a digest of your primary inbox: a caught-up line showing how many messages are unread and how many need you, then your top five important threads, each with a short AI summary and a drafted reply. It runs up to three times a day at hours you pick, and you can generate one on demand when you sit down after leave. It is a what-needs-you digest, not a report of everything that changed while you were away, so the past-due and decided-without-you prompts above still earn their place on a long absence.

Drafts use the Personal Context you set and your per-client profiles. In Copilot mode every draft waits for your approval before anything is sent, and compose has a 10-second undo window. The first sync covers about 60 days of mail, which is enough for a typical leave. AI features send message content to model providers under zero-retention terms, and your mail is not used to train models. You can see the features on the homepage and plan details on the pricing page. There is a 7-day free trial on Pro or Autopilot.