Process every incoming doc before you see it.

What the AI will do

• Builds an automatic document reviewer around your actual email, messaging and file setup.
• Prepares each document with project context, your next action and additional information that would help; it also handles interesting material outside active projects.
• Can place relevant preparation with an upcoming calendar event where the calendar setup and your permission allow it.

What you can provide

• A few typical incoming documents or access to where they arrive, in whatever usable form you have.
• Your active work and broader interests; the AI will investigate existing records before asking you to fill gaps.

Where you'll take part

• Choose where prepared documents should appear and how different kinds of reading should be treated.
• Correct inferred project links and decide whether meeting preparation should be attached or linked to calendar events.

What you'll get

A tested document-preparation tool with useful review packets and verified intake for the connected channels. Unsupported channels, incomplete reads or calendar integration remain visible rather than appearing processed.

Activity context

An attachment arrives. Before you open it, imagine having the explanation you need beside it: what it is, why it matters now, what changed, what you need to decide and the missing context already looked up. An interesting article can receive the same attention, even when its value is a new idea rather than a project task.

For the user

Share a few typical documents or name where they arrive. The AI will investigate your setup and build a preparation process with you. You choose how project work and broader reading should appear, including useful links from meeting entries.

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

Build an automatic reviewer that prepares incoming documents before I encounter them in my chosen review workflow. Understand the document and my context, investigate what would help me use it, and deliver a source-linked preparation packet beside the original. Include project documents and worthwhile reading outside existing projects.

Understand my setup and reading needs

Inspect and try access to relevant email attachments, chat and other messages including WhatsApp where accessible, shared folders, document links, project boards, calendars and meeting transcripts. Search recent correspondence and project records to establish active projects, current decisions, responsibilities, versions and upcoming meetings. Use concrete project names, people and dates from the material to follow references rather than asking me to reconstruct the whole work history.

Inspect representative document arrivals and what I do with them now. Identify supported read, event or polling, storage, inbox and calendar capabilities. Discuss whether preparation should appear within email or messaging, in a standalone app, in a digest, or through a combination that matches my working habits. Show a small example and ask about the consequential experience choice before building it. Verify actual integration behavior through current official documentation and small tests; do not promise to intercept every channel just because one inbox is connected.

Build a current project and interest map. For projects, retain my role, current stage, known decisions, open questions and upcoming events with sources and dates. For broader reading, look at newsletters, shared articles and past discussions for candidate interests such as industry ideas, AI education or another topic I care about. Ask me to correct and extend that map and choose the treatment I want: a brief explanation, connection to existing thinking, a deeper analysis or an item held for later. A topic inferred from one article is provisional until confirmed.

Investigate connection and export routes before requesting specific user action. Supplied PDFs, links, pasted text and rough notes can support a useful pilot; state which arrival channels remain unconnected. If a document cannot be read, retain it visibly as unprocessed and explain the minimum useful access step.

Define the document preparation packet

Read the actual document, including relevant tables, appendices or linked material, in bounded sections where needed. Preserve the original and its source location. Keep a coverage record for material too large or unreadable to inspect completely. A filename or summary is not the document itself.

For each arrival, prepare:

• What it is: document type, author or sender where established, date, version, main point and source.
• Where it belongs: an existing project, a possible new project, a broader interest or an uncertain match, with reasons and links.
• Where we are: the relevant current project stage, decisions and commitments, including later amendments.
• What I need from it: the particular information, review, decision or action my role requires, distinguishing an explicit request from your inference.
• What happens next: a supported next action, actual deadline if any, unresolved question and any useful prepared work.

Treat an unfamiliar project as a candidate to clarify, not a newly accepted obligation. For an interesting PDF or article unrelated to a project, explain its relevance to an agreed interest and offer the chosen depth of processing; do not invent an action item. Include the reason an item may be worth reading even when it requires no response.

Compare revised documents with the relevant prior version where accessible. Distinguish changed substance from formatting and preserve contradictory or unresolved accounts. The newest file timestamp does not itself establish an approved change. Link claims in the packet to the document or other supporting source so I can inspect the exact passage.

Find the context that would make review useful

Use the document’s actual issue, named entities and my current decision to identify useful missing information. Search related email and chat threads, earlier documents, decision records, task data, calendars and transcripts first. Seek external original sources when facts, terminology, company developments or current capabilities need checking. Explain briefly why the additional information helps with this document rather than creating an unrelated research report.

For a proposal, this might mean the original request, an earlier rejected alternative and the budget assumption now in dispute. For meeting material, it might mean what we promised last time and which decision is expected now. For an AI article, it might mean the original evaluation and the difference between what it demonstrated and what the article implies. These are illustrations, not universal additions.

Keep document content, contextual evidence and your analysis distinct. A preparation packet does not certify that every claim is true. Check the claims consequential to my use, flag material unresolved contradictions and ask a focused question when my answer would change the work. Continue unaffected preparation. Do not suppress a time-sensitive original while waiting for complete investigation; show its status and the available preparation.

Build the intake and review experience

Implement the agreed tool with arrival detection, document retrieval, project/interest lookup, preparation, source-linked storage and the chosen review interface. Each record should retain a stable source identifier, original link, version, processing state, packet location, processing time, evidence dates and unresolved items. Make it possible to correct a project match or processing preference and apply that correction to later arrivals.

Preserve original messages and documents. Detect repeated attachments and changed versions without silently discarding new information. Define what happens when retrieval fails, a file exceeds supported size, a link needs authentication or processing is incomplete. Show these cases in an exceptions view with a useful next step. Keep a coverage count so a successful run on one channel is not reported as all incoming documents vetted.

If a packet belongs with an upcoming meeting, identify the actual calendar event from project context, participants and dates. Agree whether to link the packet in an event description, attach it where supported or create a private preparation reminder. Inspect who can see the event and packet; use a private alternative if the existing event would expose preparation outside its intended audience. Apply calendar writes only within the agreed permission, preserve existing event details and verify the link and event afterward. An ambiguous meeting match requires clarification before changing an event.

Keep the setup adaptable. A document might need a short email note today and a fuller packet later; use the preferences we agreed instead of forcing every arrival through the same long format. Verify current platform capabilities for the chosen mechanisms. Build a functional batch or standalone version if live intake is blocked and name the exact remaining connection; do not call it automatic until the trigger works.

Test the preparation from arrival to use

Test representative arrivals: an existing-project revision, a document for a genuinely new project, an interesting article, a duplicate, a long or unreadable file, an ambiguous request and meeting-related material. Hold back some examples until after initial configuration. Compare each packet with the full available source and project record: correct identity, current context, faithful main point, useful additional research, justified next action and accurate processing status.

Test both initial arrival and a repeated or updated arrival. Verify trigger behavior, exception visibility, original retrieval and any authorized calendar linkage. Check that a partial read never displays as complete. Have me use a few prepared documents and correct what would have made them more useful.

For substantial volume, divide complete documents or related document families among agents with the project/interest map, source references and common packet requirements. Size assignments to actual reading capacity, track coverage and reconcile shared project changes centrally. Use manageable waves or saved passes where independent agents are unavailable.

Give an independent reviewer the actual packets, source documents, project records and intended review tasks without your explanation. Ask it to find misleading summaries, invented obligations, irrelevant research, missed context and confusing presentation. Repair material findings and rerun affected cases.

Deliver a working way to read

Deliver the tool and simple access instructions, representative prepared documents, the editable project/interest preferences and the verified channel and calendar coverage. Show how to correct a classification, request deeper processing, open the original and find an unprocessed item. Distinguish tested ongoing intake from manually processed samples and identify remaining access dependencies. Finish with the agreed preparation experience working through its supported scope, rather than handing me a proposed architecture or claiming that unread sources have been vetted.

Done when

Incoming documents in the connected scope arrive with useful, checked preparation and their originals intact. Project and broader-interest handling work, exceptions are visible, and any promised calendar linkage or recurring intake is verified or explicitly unfinished.

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.