AI Email Prompts for a Chief of Staff: 16 Prompts

A chief of staff does not manage one inbox. They manage the flow of correspondence around an executive who cannot read all of it, and their job is to compress raw email into information the executive can act on. That requires a specific kind of reading: not skimming, but extracting — pulling out what was decided, what was merely floated, and what remains unresolved, with the source attached to each line so the executive can verify any claim without re-reading the thread.

The sixteen AI email prompts in this guide are built for exactly that extraction. Each names the output format, requires source attribution, and separates the three categories a chief of staff needs separated: decided, proposed, and open. They work with any capable AI assistant — ChatGPT, Claude, Gemini, or Copilot — and are organized by output type so you can reach the right one without scanning all sixteen.

When do chiefs of staff reach for AI email prompts?#

Most email AI tools are designed for people managing their own inbox. A chief of staff manages a different problem: synthesizing correspondence from multiple workstreams, identifying what requires the executive's judgment, and converting the rest into a format the executive can absorb in ninety seconds.

The specific moments these prompts earn their keep: preparing the executive for a meeting from threads they did not read; turning a stalled negotiation into a clear account of where things stand; compiling a board update from a dozen project email trails; or writing a cross-team nudge that carries the executive's authority without sounding like a threat. The attribute-every-claim requirement distinguishes a CoS briefing from a standard summary — it is what makes the output verifiable, not just readable.

What does a strong chief-of-staff AI prompt contain?#

Every reliable CoS briefing prompt has five parts. Skip any one of them and the output degrades in a predictable way.

Prompt 1: One-page exec briefing from a thread#

The core output: everything the executive needs on one page, in three labeled categories, with sources named. The 350-word cap matters — without it the model produces a structured document that is still too long to read in ninety seconds. Trim repeated quoted history from the thread before pasting; it can double the length without adding any information.

Prompt 2: Separation with full source attribution#

The precision version of Prompt 1. Where Prompt 1 produces a readable briefing, this produces an audit-ready log. Use it when the stakes are high — a contract decision, an org change, a sensitive negotiation — and you need a record that can be checked against the original thread item by item.

Prompt 3: Three-bullet status update for the executive#

When the executive has two minutes between meetings: current state, what changed, and what if anything they need to do. The 20-word cap per bullet is what makes it usable as a handoff rather than a document to read.

Prompt 4: Catch-up brief after the executive was away#

Coming back from a board retreat or an off-site, the executive cannot read forty threads. This produces a return brief — what changed, what was decided without them, and what awaits their input — with the most consequential thread first. Set the absence dates inside the prompt.

Prompt 5: Turn a thread into a decision memo#

A decision memo is the formal artifact: situation, options considered, recommendation, decision taken, and next steps. Use it when a thread has produced a real decision and you need a shareable document — for wider distribution, for the record, or for a board packet. The source-attribution instruction on the decision line catches the most common failure mode: a model marking a proposed option as the settled decision.

Prompt 6: Extract alternatives and their reasoning#

Before a decision is made, the thread often contains multiple options with advocates and objections. This prompt extracts each option, who raised it, and the reasoning — including options dismissed quickly. The rejected options often contain the clearest statement of the constraints the final decision had to work within.

Prompt 7: Format a decision for the record#

After a decision is made, you need a one-line record: what was decided, who made it, when, and on what basis. This is what prevents relitigation six months later. Use the one-record-per-decision instruction when the thread contains more than one distinct decision.

Prompt 8: Cross-team nudge with borrowed authority#

A chief of staff does not just analyze; they move things. A nudge from the executive's office carries different weight than one that doesn't invoke it. The keys: reference the specific commitment by date, name what is needed and when, and attribute the follow-up to the executive's office clearly. Keep it under 100 words — longer nudges get read as lectures.

Prompt 9: Escalate a stalled item#

When something has been unresolved long enough to require escalation, the email has to do three things: name the block, say what cannot move until it is resolved, and set a specific deadline. Make the framing factual rather than frustrated — the model reproduces the emotional register you give it.

Prompt 10: Multi-team status check from threads#

When you are managing a cross-functional initiative and need a single view of where each workstream stands, this prompt synthesizes multiple threads into one status table. Blocked items appear first, making them immediately visible without scanning.

Prompt 11: Board update from project threads#

Compressing project email trails into a board update requires four specific sections that board members expect. The source-attribution instruction on risks is important: a risk that has already been resolved should not appear as a current risk, and a model without attribution instructions frequently conflates the two.

Prompt 12: Pre-read for a board agenda item#

A pre-read is different from a summary: it is designed to be read cold by someone with other priorities, so they arrive ready to discuss rather than ready to be briefed. Five minutes to read. One specific question for the board to answer. The 400-word cap is what keeps it a pre-read rather than a report.

Prompt 13: Four-bullet board-adjacent situation summary#

For situations that might reach the board but have not yet — a significant commercial event, an investor inquiry, a regulatory item — this produces the four-bullet format that can be included in a board update or shared with the CEO in thirty seconds.

Prompt 14: Triage — what needs the executive, what doesn't#

The most valuable thing a chief of staff does is absorb what does not require the executive. This prompt returns two lists: threads requiring the executive's judgment or authority, and threads that can be handled without them. The framing matters: assess on whether the executive is needed, not on whether the topic is important. Important topics that do not require executive judgment are the things the CoS handles.

Prompt 15: Vendor or partner situation report#

When a vendor relationship has produced weeks of back-and-forth and the executive needs the current state in a few minutes, this produces a structured situation report from the email trail — including what each side owes the other, which is often the clearest indicator of who is responsible for moving things forward.

Prompt 16: Map who owns what across a complex thread#

In a multi-party thread with action items scattered across dozens of messages, this builds the accountability map: every commitment, pending item, and completed obligation — with firm commitments separated from soft ones. Overdue items appear first so you can identify what needs a nudge immediately.

The hard version: incomplete or contested source material#

The hardest briefing situations are the ones where the source material is incomplete or contested. Incomplete sources — threads where key conversations happened over a call, in a side chain, or via a messaging app — produce briefings that look complete but are missing the substance. The fix is to tell the model to flag the evidence level for each item: 'Mark each item as COMPLETE (clear in thread), PARTIAL (referenced but not detailed), or INFERRED (not in thread — note this as an assumption).' A briefing that shows its gaps is more useful than one that fills them invisibly.

For contested decisions, do not ask the model to resolve the conflict. Add: 'If the thread contains conflicting versions of a decision, report both with their sources rather than choosing one.' That is the chief of staff's job, not the model's. And for politically sensitive threads, redact names and identifying details before pasting — the prompts work equally well on a thread where a real name becomes 'Stakeholder A'.

What to do when stakeholders don't reply#

When a cross-team nudge goes unanswered, the second attempt needs a different frame — not a louder version of the first. Use this follow-up prompt after two to three business days with no reply.

How AI Emaily handles this for chiefs of staff#

The prompts above work in any AI chatbot. The structural cost is that every briefing starts with a copy-paste session: open the thread, expand the history, copy, switch tabs, paste, and re-describe the executive's preferences — repeated for each thread each morning. For a chief of staff handling eight threads a day, that overhead compounds.

We build AI Emaily, an AI-native email client with a living brief feature that turns the executive's real inbox into an auto-updating briefing — decided, proposed, open, with source attribution — without a paste. Your context and client profiles are set once in the product, not re-described in every session. When you need to act on a thread, Copilot mode drafts the nudge or update for your approval before anything goes out. Start with a 7-day free trial at aiemaily.com, and see the full plan details at aiemaily.com/pricing.