Rank incoming messages before you read them.

What the AI will do

• Learns what matters from your projects, commitments and relationships, then builds a tool that ranks incoming email and messages accordingly.
• Keeps that understanding current, alerts you when something cannot wait, and prepares replies you can review and authorize together.

What you can provide

• A rough description of the messages that overwhelm you, or access to the inboxes and messaging channels you want covered.
• Your work and personal priorities where the records do not tell the whole story; the AI investigates available connections and setup.

Where you'll take part

• Confirm which projects and people matter, what deserves an immediate interruption and where alerts should reach you.
• Compare a sample priority list with your own judgment, correct the tool and decide exactly which replies it may send.

What you'll get

A connected message-priority tool with an on-demand list, tested urgent alerts and ready-to-review replies. It shows which channels it covers; missing background access means you have a tested partial tool until that connection is enabled.

Activity context

Newest on top is a peculiar way to organize your life. A message that arrived eight seconds ago can bury the decision holding up your biggest project—or a note from someone you love. Build a tool that understands what matters and puts those messages first.

For the user

Start with the email and messaging apps you use and a rough sense of what gets lost. The AI will find relevant context, help establish access and build the tool. You will tune its judgment with real examples and choose how interruptions and sending should work.

Guidance for the AI

These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result.

Start with the activity and the user's context

Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear.

Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them.

Check the capabilities needed for this work

Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit.

When useful access is missing:

• Explain what you cannot reach and how that material would help.
• Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot.
• If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response.
• Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available.

Ask focused questions and respect the answers

Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful.

Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements.

Keep evidence and independent contributions distinct

Distinguish evidence, inference, and missing information. Check your work against the source material.

Use independent agents when your system actually supports delegation and independent investigation or review would improve the work:

• Give each agent a bounded brief and the context appropriate to its role.
• Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations.
• Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately.

Produce, check, and revise the result

Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them.

Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result.

Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized.

Your AI prompt

Learn what deserves my attention

Build a tool integrated with my email and messaging that reads new messages from the channels I authorize, ranks their importance using an evolving understanding of my life, alerts me to genuinely critical and time-sensitive items, and gives me a priority list whenever I ask. Prepare appropriate replies so I can authorize a clearly identified batch with an instruction such as “Send those responses.”

Start with what I have supplied. Inspect and try access to the relevant email and chat systems, including WhatsApp where accessible, before asking me to assemble a history. Search recent threads, calendar invitations, meeting transcripts, project plans, task and decision records, and messages involving recurring work and personal contacts. Use project names, upcoming events, deadlines and people from those records to extend the search. Find active commitments, who is waiting on me, consequential deadlines, unresolved questions and the people whose messages I particularly need to see. Read enough of a thread to distinguish a new request from an issue already resolved elsewhere.

Build a dated, source-linked working picture of active projects, priorities and important professional and personal people. Distinguish a documented deadline or relationship from an inferred priority. Frequent correspondence and a senior job title alone do not establish importance. Ask one focused question at a time about consequential gaps—particularly personal relationships and priorities the records cannot establish. Confirm inferred priorities before treating them as my preferences.

For substantial histories, assign bounded source or project slices to agents if available, with periods, source references and required findings. Maintain a coverage record, reconcile overlapping conversations centrally and use manageable sequential passes if agents are unavailable.

Establish the integration and operating rules

Inspect whether the actual environment can retrieve new messages, revisit changed threads, store the priority picture, run in the background, show an on-demand list, deliver alerts and create or send replies. Check the current official documentation for the tools actually being connected. Demonstrate access with small reads; a named connector or an open chat session does not establish continuous monitoring.

Choose the simplest supported implementation that fits my existing apps. Build the code, configuration and interface yourself. Investigate missing setup first, then ask for the specific connection or administrator action only another person can perform; offer to draft that request. An export can support an initial ranking and a tested local tool, but cannot supply incoming messages. Make the resulting coverage limit clear early and carry it into the final result.

Propose a short set of practical priority rules from the evidence. Consider consequences of delay, actual deadlines, an upcoming meeting, a blocked colleague or customer, personal importance and whether action is mine to take. Treat a sender's “URGENT” as a claim to evaluate. Show a few real examples and let me correct them. Establish what counts as both mission-critical and time-critical, where an immediate alert should reach me, and the treatment of quiet hours and exceptions. Establish which connected accounts and channels are in scope. Resolve consequential unanswered choices before enabling the dependent behavior; retain decisions I have already made.

Build the priority and response workflow

Implement the complete incoming-message cycle:

• Read new or updated messages from every connected source. Record the last successful read for each channel and visibly identify stale or failed sources. Reconcile duplicate copies and cross-channel references without losing the original links.
• Refresh the project and people context from substantive new decisions, changing deadlines, completed work and my corrections. Preserve the evidence and date behind a change. Suggest uncertain priority changes to me instead of silently turning an inference into a personal preference.
• Rank conversations by what needs attention now. For each item show who it is from, what they need, the relevant project or personal context, the reason for its position, any deadline, the next useful action and a link to the original thread. Separate action requests from valuable information. Surface uncertain classifications where overlooking the message could matter.
• Reconsider waiting items as deadlines approach or their context changes. Make “show me my messages by priority” work even when no new messages have arrived.
• Send agreed critical alerts to the chosen destination with enough information to act. Avoid repeated alerts for the same unchanged issue; escalate again only under the rules we set. Show the actual delivery state.
• Draft replies using the thread, relevant records and my established tone. Identify missing decisions or commitments a reply would make. Do not invent an answer, availability or promise. Keep drafts linked to the exact conversation and current message version.

Build a simple way for me to correct a ranking, mark an issue handled, change an important person's treatment, update a priority or pause the tool. Feed these corrections into subsequent runs without making one exceptional choice an unexplained permanent rule.

Make batch sending precise

Present each proposed reply with its account, recipient or channel, full text, attachments if any and source conversation. Give the current batch a clear identity and allow individual exclusions or edits. My approval of that displayed batch authorizes those exact replies; it does not authorize future replies. If I have already granted a specific standing sending rule, apply it within its stated bounds.

Before sending, check whether a new message or changed fact makes a draft stale. Hold affected replies for revision and renewed approval when their content or destination changes. Send the authorized remainder once, record outcomes and check the source system for actual sent messages. If a timeout leaves an outcome uncertain, inspect the conversation or sent record before retrying. Report partial success and any remaining items precisely. If sending access is missing, preserve usable drafts and label them unsent.

Test, repair and make it useful at my desk

Replay a representative set of real messages and controlled examples through the tool: an urgent project blocker, a time-sensitive personal message, a routine recent message, a resolved thread, a deadline approaching without new mail, conflicting priorities, a changed reply thread and an unavailable channel. Compare the ordering with my judgment and revise the rules. Check that alerts arrive at the intended destination and that on-demand results identify their freshness. Use controlled test destinations for sending tests; no production reply is sent merely to test the tool.

Give a fresh reviewer the priority results, source threads and agreed rules to check missed important items, unjustified interruptions and reply fidelity. Use a distinct review pass and disclose shared context if no independent reviewer is available. Repair material findings and recheck affected behavior.

Within the agreed scope, enable the recurring service and verify its first actual run and next trigger. Demonstrate an incoming message becoming a ranked item, an agreed test alert arriving, a corrected priority persisting and an approved sending test completing. Deliver the working tool, current priority list, drafts, source-coverage view and concise instructions for asking, correcting and pausing. If essential access or runtime is still missing, deliver the implemented and tested portion with the exact remaining setup; do not describe it as reading every incoming message or running continuously.

Done when

You can ask for messages by priority, receive verified critical alerts and review replies as a precise sendable batch, with the tool updating its project and people context. Its connected coverage and recurring operation are demonstrated; any missing connection or unsent reply is identified explicitly.

These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.