A long email can mix firm requests with background, preferences, tentative ideas and questions. Ask an AI assistant to extract everything, and it may turn a passing suggestion into a commitment—or leave out a small but important output. The safer approach is to define what counts as a deliverable, ask for evidence from the message, then compare the result with the original. This gives you a useful working list without quietly expanding the sender’s request or treating the assistant’s interpretation as the final word.
The useful bits
- Separate requested outputs from ideas, conditions and open questions.
- Ask the assistant to show the exact wording behind each proposed deliverable.
- Check the list against the original email before agreeing to any work.
Decide what belongs on the list
A deliverable is an output the sender asks someone to provide: for example, a revised document, a summary, or a set of recommendations. It is different from an activity (“please review this”) unless the email also names something to hand back. It is also different from context, such as why the work matters, or a preference about how to present it.
Before using AI, set that boundary in your request. Ask it to list only outputs the sender explicitly requests or clearly makes a condition of the work. Tell it to put suggestions, possibilities, and unresolved questions in separate categories. This reduces the chance that every sentence containing an action verb becomes a new task.
Read the wording, not just the topic
Look for commitment language: “please send,” “we need,” or “include” usually signals a request. Tentative language—“we could,” “perhaps,” or “if useful”—usually signals an option, not an agreed output. A question such as “Would a short overview help?” needs an answer; it does not necessarily request an overview.
Words such as “if,” “unless,” and “only after” can make a deliverable conditional. Preserve the condition rather than listing the output as unconditional. Also keep any stated format or recipient attached to the item: “send the revised table to the project group” is more specific than “revise table.”
Give the assistant a bounded task
Provide the email text and ask for a compact table with four fields: proposed output, status (explicit, conditional, suggested, or unclear), supporting wording, and any condition or deadline stated. Tell the assistant not to infer missing outputs, owners, dates, or formats. If it cannot point to wording that supports an item, it should label the item unclear instead of filling the gap.
For sensitive mail, consider whether you should share its contents with an AI tool at all. Use only information you are permitted to provide, and remove unnecessary personal or confidential details. AI assistance may be available depending on your BITTU Mail account and plan; availability can vary. Regardless of the tool, do not provide real passwords or one-time codes.
Try it on a mixed request
Suppose a sender asks for a revised schedule and a one-page summary by Friday. They add that a chart “might be useful,” and ask whether a presentation is necessary. The schedule and summary are explicit outputs. The chart is a suggestion. The presentation is an open question, not an assigned deliverable. A careful extraction preserves those differences instead of converting all three into tasks.
Prompt: “Extract only requested outputs from this email. Separate explicit, conditional, suggested, and unclear items. For each item, quote a short supporting phrase from the email. Do not infer owners, dates, formats, or extra work. Keep questions and suggestions separate.” Email: “Please send the revised schedule and a one-page summary by Friday. A chart might be useful. Do we also need a presentation?”
Check every item against its evidence
Read the original email alongside the assistant’s list. For each proposed deliverable, find the phrase that supports it and ask: does this phrase request an output, or merely mention an idea? Is the item conditional? Has the assistant turned a question into a commitment, or added a format, audience, or deadline that was not stated?
Then scan the email in order, including its closing lines and any quoted context you meant to include. This catches omissions and helps distinguish the sender’s current request from earlier discussion. If the wording remains ambiguous, keep it marked unclear and ask the sender to confirm rather than guessing.
Confirm the list before acting
A useful final list is short enough to assign and precise enough to verify. Keep explicit outputs separate from optional ideas and unresolved questions. Preserve conditions and stated deadlines, but do not add estimates or promises. If you need to acknowledge the request, restate the bounded list and ask about only the unclear points.
You can confirm the extraction worked by tracing each listed output to the email, checking that suggestions and questions have not been promoted to commitments, and confirming that no clear request was missed. Review the original message for names, dates, recipients, promises, and financial details: AI can misread them. The assistant’s list is a draft for checking, not a substitute for the sender’s wording.
- Every explicit output has supporting wording in the email.
- Suggestions and open questions remain separate from commitments.
- Conditions, formats, recipients, and deadlines match the source—or are marked unclear.
About this guide
Prepared with AI assistance for the BITTU Mail editorial library. Check important details against your original messages and current account settings. Product access depends on your role, plan and provisioning state. Provider interfaces and requirements can change.
View current plans · Security information · Suggest a correction